الذكاء الاصطناعي وتعلم الآلة من أجل الطاقة المستدامة: النمذجة التنبؤية، التحسين والأثر الاجتماعي الاقتصادي في الولايات المتحدة الأمريكية AI and Machine Learning for Sustainable Energy: Predictive Modelling, Optimization and Socioeconomic Impact In The USA

المجلة: International Journal of Applied Sciences and Radiation Research، المجلد: 2، العدد: 1
DOI: https://doi.org/10.22399/ijasrar.19
تاريخ النشر: 2025-03-14

الذكاء الاصطناعي وتعلم الآلة من أجل الطاقة المستدامة: النمذجة التنبؤية، التحسين والأثر الاجتماعي الاقتصادي في الولايات المتحدة الأمريكية

تشيديرا فيكتوريا إيباه أيوديجى أديجبولا جامعة هاريسبرغ للعلوم والتكنولوجيا، الولايات المتحدة الأمريكية*البريد الإلكتروني للمؤلف المراسل: victoriaibeh23@gmail.com – ORCID: 0000-0002-5247-785X باحث مستقل، المملكة المتحدةالبريد الإلكتروني: adegbola_ayodeji@yahoo.com – ORCID: 0000-0002-5247-785Y

تاريخ المقال:

DOI: 10.22399/ijasrar. 19
تاريخ الاستلام: 10 يناير 2025
تم القبول: 14 مارس 2025

الكلمات المفتاحية:

الذكاء الاصطناعي، تعلم الآلة، الطاقة المستدامة، النمذجة التنبؤية، تحسين الطاقة، توقع الطاقة، الأثر الاجتماعي والاقتصادي.

الملخص

تستكشف هذه الدراسة كيف يمكن استخدام التعلم الآلي والذكاء الاصطناعي لتعزيز كفاءة الطاقة، وتوقع اتجاهات استهلاك الطاقة، وتحسين أنظمة الطاقة في الولايات المتحدة الأمريكية. استخدمت هذه الدراسة مجموعات بيانات تتضمن استخدام الطاقة المنزلية، واتجاهات اعتماد المركبات الكهربائية، وتحليلات الشبكة الذكية التي تم الحصول عليها من مصادر عامة، وقواعد بيانات، وأجهزة استشعار إنترنت الأشياء. تطبق هذه الدراسة تقنيات التعلم الآلي المتقدمة مثل التعلم العميق، ونماذج الانحدار، والتعلم الجماعي لتحسين دقة التوقعات بهدف تحقيق تخصيص فعال للموارد. بالإضافة إلى ذلك، تحقق هذه الدراسة في توقع الأعطال في المركبات الجديدة (NEVs) وتأثيراتها على استقرار الشبكة وإدارة الطلب على الطاقة. كما تفحص البحث الأثر الاجتماعي والاقتصادي للسياسات الطاقية المدفوعة بالذكاء الاصطناعي وتبرز دورها في تقليل البصمات الكربونية، وتعزيز العدالة في الطاقة، وتعزيز النمو الاقتصادي المستدام. يتم تطبيق الشبكات العصبية التكرارية لتوقع اتجاهات استهلاك الطاقة ومعدلات اعتماد المركبات الكهربائية من خلال تحليل بيانات الاستخدام التاريخية. تُستخدم الشبكات العصبية التلافيفية والمشفّرات التلقائية للكشف عن الشذوذ في أداء بطاريات المركبات الجديدة والصيانة التنبؤية. كما تستخدم نماذج التعلم العميق بيانات أجهزة استشعار إنترنت الأشياء في الوقت الحقيقي لتعزيز كفاءة توزيع الطاقة في الشبكات الذكية. تُستخدم نماذج الانحدار الخطي لتوقع الطلب على الطاقة في المنازل والصناعات بناءً على عوامل مثل الطقس، والأسعار، والمتغيرات الاجتماعية والاقتصادية. كما يتوقع الانحدار الخطي اتجاهات استهلاك الطاقة في المستشفيات والمصانع. تُستخدم غابات العشوائية وXGBoost في توقع الطلب على الطاقة وتجمع استهلاك الطاقة. تُستخدم مقاييس تقييم الأداء مثل متوسط الخطأ التربيعي (MSE)، وجذر متوسط الخطأ التربيعي (RMSE)، وR-squared ( تُستخدم لتقييم دقة وفعالية النموذج.

1. المقدمة

1.1 الخلفية

مع التقدم في تقنيات الطاقة المتجددة، والمركبات الكهربائية (EVs)، وأنظمة الشبكات الذكية، شهد استهلاك الطاقة في الولايات المتحدة الأمريكية تحولات كبيرة. نظرًا لأن طرق التنبؤ التقليدية غالبًا ما تفشل في التقاط الديناميات المعقدة للطلب والعرض على الطاقة، يتم استخدام نماذج التنبؤ المدعومة بالذكاء الاصطناعي لتحسين الدقة واتخاذ القرار [1-12]. وفقًا لأحمد وآخرون (2025)، تم استخدام نماذج التعلم الآلي بنجاح للتنبؤ باتجاهات اعتماد المركبات الكهربائية، وتقييم استهلاك الطاقة في المنازل، وتوقع كفاءة الطاقة في المستشفيات، مما يوفر رؤى للسياسات الطاقية المستدامة [1]. يضمن الكشف عن الأعطال المدعوم بالذكاء الاصطناعي في المركبات الجديدة للطاقة موثوقية هذه الحلول النقل الناشئة مع تقليل هدر الطاقة [7]. أدت مجموعة الذكاء الاصطناعي والتعلم الآلي في أبحاث الطاقة إلى تحقيق اختراقات كبيرة في أنظمة الاستجابة للطلب، وتحسين الشبكات، وإدارة الموارد المتجددة [2]. يوضح حسين وآخرون (2025) أن الفوائد الاجتماعية والاقتصادية لهذه التقدمات تتجاوز مكاسب الكفاءة، حيث تساهم استراتيجيات الطاقة المحسّنة في الاستقرار الاقتصادي، وتوفير التكاليف، وتقليل الأثر البيئي [8].

1.2 أهمية البحث

يعتقد سومن وآخرون (2024) أنه مع تزايد الطلب على حلول الطاقة المستدامة، ظهرت الذكاء الاصطناعي وتعلم الآلة كأدوات قوية لمعالجة تحديات كفاءة الطاقة [13]. الطاقة التقليدية
غالبًا ما تعتمد أنظمة الإدارة على نماذج ثابتة تفشل في التكيف مع التقلبات في الوقت الحقيقي، مما يؤدي إلى عدم الكفاءة وهدر الطاقة. وهذا بدوره يدفع إلى استخدام الذكاء الاصطناعي وتعلم الآلة في أنظمة إدارة الطاقة. يمكن لمزودي الطاقة أيضًا استخدام نماذج مدعومة بالذكاء الاصطناعي لضبط أحمال الشبكة ديناميكيًا، وتحسين استخدام الموارد المتجددة، وتنفيذ آليات الكشف عن الأعطال في الوقت الحقيقي. تتيح التحليلات التنبؤية باستخدام الذكاء الاصطناعي وتعلم الآلة التنبؤ الدقيق بالطلب على الطاقة، واستقرار الشبكة، والتأثيرات الاجتماعية والاقتصادية، مما يعزز بدوره صياغة السياسات المستندة إلى البيانات لتحسين أنماط استهلاك الطاقة في الولايات المتحدة. علاوة على ذلك، تعزز تقنيات تعلم الآلة العدالة في الطاقة، مما يضمن توزيعًا عادلًا ويقلل من فقر الطاقة في المجتمعات المحرومة. نظرًا للحاجة الملحة لاستدامة الطاقة، والعمل المناخي، والمرونة الاقتصادية، يسلط هذا البحث الضوء على كيفية مساهمة الذكاء الاصطناعي وتعلم الآلة في الجدوى طويلة الأجل لأنظمة إدارة الطاقة الذكية.

1.3 الأهداف

تهدف هذه الدراسة إلى استكشاف كيفية تحويل الذكاء الاصطناعي (AI) وتعلم الآلة (ML) لطريقة إدارة واستخدام الطاقة في الولايات المتحدة الأمريكية. أحد المحاور الرئيسية هو تحسين توقعات استهلاك الطاقة، مما يضمن أن الأسر والصناعات والمستشفيات يمكنها التنبؤ بشكل أفضل والتخطيط لاحتياجاتها من الطاقة. غالبًا ما تكافح الطرق التقليدية لالتقاط تعقيدات الطلب على الطاقة، لكن النماذج المدفوعة بالذكاء الاصطناعي يمكن أن تحلل كميات هائلة من البيانات وتكشف عن أنماط خفية لتحسين الدقة. جانب آخر مهم من هذه الدراسة هو تحسين الطاقة المتجددة. مع الاعتماد المتزايد على الطاقة الشمسية وطاقة الرياح، من الضروري تطوير نماذج يمكن أن توازن بين هذه المصادر من الطاقة مع الكهرباء من الشبكة لتقليل الفاقد وتحسين الكفاءة. يمكن لخوارزميات تعلم الآلة أن تساعد في التنبؤ بتقلبات توليد الطاقة المتجددة وضمان إمداد كهربائي أكثر استقرارًا وموثوقية. بالإضافة إلى ذلك، تتناول هذه الدراسة اكتشاف الأعطال في المركبات الكهربائية الجديدة (NEVs). مع تزايد شيوع السيارات الكهربائية (EVs)، فإن الحفاظ على موثوقيتها وكفاءتها أمر أساسي. يمكن أن تحدد أنظمة التنبؤ بالأعطال المدفوعة بالذكاء الاصطناعي المشكلات المحتملة قبل أن تصبح مشاكل كبيرة، مما يساعد على إطالة عمر البطارية وتحسين الأداء العام. بالإضافة إلى التحسينات التقنية، تفحص هذه الدراسة أيضًا الأثر الاجتماعي والاقتصادي للسياسات الطاقية المدفوعة بالذكاء الاصطناعي. من خلال تحليل كيفية تعزيز الذكاء الاصطناعي للعدالة الطاقية، وتقليل البصمات الكربونية، وتقليل التكاليف على المستهلكين، نهدف إلى تقديم رؤى قيمة يمكن أن تشكل سياسات أكثر ذكاءً واستدامة. في النهاية، تسعى هذه الدراسة إلى استغلال قوة الذكاء الاصطناعي وتعلم الآلة لإنشاء مستقبل طاقي أكثر كفاءة وفعالية من حيث التكلفة وصديق للبيئة.

2. مراجعة الأدبيات

لقد حظيت تطبيقات التعلم الآلي والذكاء الاصطناعي في قطاع الطاقة باهتمام كبير في السنوات الأخيرة. أشار سومن وآخرون (2024) إلى أن النماذج المعتمدة على الذكاء الاصطناعي كانت لها دور حاسم في تقييم الأثر البيئي والاجتماعي والاقتصادي لاعتماد الطاقة المتجددة، مما يساعد صانعي السياسات على اتخاذ قرارات أكثر استنارة [13]. في الوقت نفسه، استكشف شيل وآخرون (2024) القدرات التنبؤية للتعلم الآلي في اعتماد السيارات الكهربائية، موضحين كيف تتفوق الأساليب المدفوعة بالذكاء الاصطناعي على طرق التنبؤ التقليدية [12].
تم التأكيد على أن الانتقال إلى المركبات الكهربائية (EVs) كان محورًا لبحث واسع النطاق نظرًا لأهميته في التخفيف من تغير المناخ، وتعزيز التخطيط الحضري، وتعزيز استدامة الطاقة. اعتمدت الدراسات المبكرة في هذا المجال بشكل أساسي على تقنيات إحصائية تقليدية، مثل الانحدار الخطي وتحليل السلاسل الزمنية، للتنبؤ باتجاهات اعتماد المركبات الكهربائية.
وبالمثل، ركز تشودري وآخرون (2024) على دور الذكاء الاصطناعي في استهلاك الطاقة المنزلية، موضحين أن تقنيات النمذجة الذكية يمكن أن تحسن كفاءة الطاقة واستراتيجيات الاستجابة للطلب [5]. لقد حول دور الذكاء الاصطناعي وتعلم الآلة إدارة الطاقة بشكل كبير استراتيجيات التحليل التنبؤي والتحسين.
يسلط براينجولفسون وآخرون (2017) الضوء على كيفية ثورة الذكاء الاصطناعي في الصناعات من خلال تسهيل اتخاذ القرارات المعتمدة على البيانات وتحسين الأنظمة المعقدة، بما في ذلك شبكات الطاقة [3]. لقد عزز دمج تقنيات التعلم العميق، مثل الشبكات العصبية التلافيفية (CNNs) والشبكات العصبية المتكررة (RNNs)، دقة النمذجة التنبؤية في توقع الطاقة (لكون وآخرون 2015) [10].
بالإضافة إلى ذلك، يؤكد والير وآخرون (2013) على أهمية تحليلات البيانات الضخمة في تحسين سلسلة التوريد وإدارة الطاقة، مما يسمح بتوقع الطلب بدقة أكبر وزيادة الكفاءة التشغيلية []15.
بينما يُعتبر الطاقة الكهرومائية غالبًا مصدرًا نظيفًا للطاقة، إلا أنها تشكل مخاطر على النظم البيئية المائية. إن بناء السدود يعطل تدفق الأنهار الطبيعي ويهدد تجمعات الأسماك. وبالمثل، على الرغم من أن محطات الطاقة الحرارية الأرضية لديها انبعاثات منخفضة، إلا أنها يمكن أن تؤدي إلى هبوط الأرض أو قد تحفز أحداث زلزالية في بعض المناطق.
علاوة على ذلك، قدم حسنين وآخرون (2025) إطار عمل للكشف عن الأعطال مدعوم بالذكاء الاصطناعي للمركبات الجديدة للطاقة (NEVs)، مما يعزز بشكل كبير موثوقية البطارية وأداء المركبة [7]. كما أكدت أبحاث باروا وآخرون (2025) على أهمية الذكاء الاصطناعي في تحسين أنماط استهلاك الطاقة الحضرية، مما يضمن نظام توزيع طاقة أكثر استدامة وكفاءة [2].
تدعم الدراسات التجريبية أن تقنيات وممارسات كفاءة الطاقة أثبتت فعاليتها في مواقف متنوعة. على سبيل المثال، يمكن أن تعمل منظمات الحرارة الذكية التي تستخدم تعلم الآلة على تحسين جداول التدفئة والتبريد في المنازل، مما قد يؤدي إلى تقليل يصل إلى في استهلاك الطاقة المنزلية، كما تشير الأبحاث حول استخدام الطاقة السكنية. في كاليفورنيا، على سبيل المثال، كشفت التحقيقات عن توفير كبير في الطاقة نتيجة لتجديد المنازل بعزل متقدم ونوافذ موفرة للطاقة، خاصة في المناطق التي تشهد ظروف جوية قاسية.

2.2 الفجوات والتحديات

على الرغم من التقدم الأخير في تطبيقات الذكاء الاصطناعي لإدارة الطاقة، لا تزال هناك عدة فجوات. واحدة من أكبر التحديات هي توفر البيانات وجودتها، حيث تحتوي مجموعات بيانات الطاقة غالبًا على قيم مفقودة أو تناقضات يمكن أن تؤثر على أداء النموذج [13].
بالإضافة إلى ذلك، بينما أثبتت نماذج تعلم الآلة فائدتها في توقع معدلات اعتماد المركبات الكهربائية (EV)، لا يزال من الصعب دمج سلوك المستهلك في الوقت الحقيقي والتدخلات السياسية الخارجية بشكل فعال [12]. يؤكد جيروسييه وآخرون (2024) أنه على الرغم من التقدم الملحوظ، فإن الطرق التنبؤية الحالية لاعتماد المركبات الكهربائية (EV) غير كافية بشكل أساسي. هذه الطرق لا تعالج مجموعة من المتغيرات ذات الأبعاد العالية والمتداخلة التي تشكل السوق EV بشكل حاسم.
يمكن أن يؤدي التحيز الخوارزمي في نماذج تعلم الآلة إلى توزيع غير عادل لموارد الطاقة، مما يؤثر على عدالة الطاقة وصياغة السياسات [6]. علاوة على ذلك، يجب معالجة المخاوف الأخلاقية المتعلقة بدور الذكاء الاصطناعي في الاستدامة لضمان نشر مسؤول [9]. واحدة من القيود الرئيسية للنماذج الحالية للذكاء الاصطناعي هي صعوبة تكيفها مع التقلبات في الطلب على الطاقة في الوقت الحقيقي.
يقترح شنايدر وآخرون (2021) أن الأطر الذكية التكيفية، القادرة على التعلم من البيانات في الوقت الحقيقي، يمكن أن تعزز من قوة النموذج وأن الأساليب التقليدية غير مجهزة بشكل جيد لإدارة تعقيد وعشوائية العوامل التي تدفع انتشار المركبات الكهربائية، بما في ذلك تقلبات أسعار الوقود، والدعم الحكومي غير المتسق، والتطورات التكنولوجية السريعة المتعلقة بكل من البطاريات الموعودة والمتاحة تجاريًا [11].
يمكن أن تحلل نماذج السلاسل الزمنية، على سبيل المثال، البيانات التاريخية، لكنها غير قادرة على التقاط التغيرات المفاجئة في السياسة والتكنولوجيا بشكل فعال، مما يترك فجوات كبيرة في الدقة التنبؤية. تعتمد الأساليب التقليدية بشكل كبير على البيانات المجمعة والافتراضات الخطية، مما يحجب التباينات في معدلات الاعتماد الإقليمية ويخفي الاتجاهات الأساسية في سلوك المستهلك وتفضيلاته. هذه الظاهرة بارزة بشكل خاص في الولايات المتحدة، حيث تختلف معدلات اعتماد المركبات الكهربائية بشكل كبير عبر الولايات بسبب اختلافات صارخة في السياسة والمناخ والظروف الاجتماعية والاقتصادية. أشار تشودري وآخرون (2024) إلى أن قابلية التوسع لا تزال قيدًا كبيرًا للعديد من نماذج الذكاء الاصطناعي، التي غالبًا ما يتم تدريبها على بيانات إقليمية وتكافح للتعميم بشكل فعال على المستوى الوطني [5]. كما أشار حسنين وآخرون (2025) إلى أن التنبؤ بالأعطال المعتمد على الذكاء الاصطناعي في المركبات الكهربائية الجديدة (NEVs) معقد بسبب التباينات في تصميم المركبات وتكنولوجيا البطاريات [7].
تظهر ماماتا وآخرون (2024) أنه على الرغم من مزايا NEVs، فإنها غالبًا ما تعاني من أعطال متنوعة تؤثر بشكل كبير على الأداء وتقلل من رضا العملاء. تشمل المشكلات الشائعة تدهور بطارية الجر، وفشل نظام الشحن، وأعطال البرمجيات. لا تؤدي هذه المشكلات إلى تقويض وظيفة المركبة فحسب، بل تؤدي أيضًا إلى زيادة تكاليف الصيانة وتراجع ثقة المستهلك في تكنولوجيا NEV.
تشكل تعقيدات الكشف عن الأعطال وحلها داخل أنظمة NEV، التي تنبع من تكاملها المعقد بين الأجهزة والبرمجيات، تحديًا كبيرًا. يتطلب معالجة هذه القضايا تحسينات في طرق جمع البيانات، وزيادة قابلية تفسير النموذج، وكفاءة حسابية أكبر لضمان حلول طاقة مدفوعة بالذكاء الاصطناعي قوية وموثوقة.

3. المنهجية

3.1 جمع البيانات والمعالجة المسبقة

مصادر البيانات

تستند هذه الدراسة إلى مجموعة متنوعة من مجموعات البيانات، بما في ذلك سجلات استهلاك الطاقة المنزلية والصناعية، وبيانات عدادات الطاقة الذكية في الوقت الحقيقي، والمعلومات المناخية، وسجلات أداء المركبات الكهربائية. بالإضافة إلى ذلك، تستخدم بيانات متاحة للجمهور من إدارة معلومات الطاقة الأمريكية (EIA)، والمختبر الوطني للطاقة المتجددة (NREL)، وتقارير الصناعة التي تفصل اتجاهات استخدام الطاقة. معًا، تخلق هذه المصادر مجموعة بيانات شاملة ستستخدم لتدريب نماذج تعلم الآلة لتوقع أنماط استهلاك الطاقة وتحسين تخصيص الطاقة المتجددة.

المعالجة المسبقة للبيانات

لضمان سلامة البيانات وزيادة دقة النموذج، يتم تنفيذ مجموعة من تقنيات المعالجة المسبقة. يتم معالجة القيم المفقودة من خلال تطبيق طرق الاستيفاء والتقدير. تُستخدم تقنيات قياس الميزات، بما في ذلك قياس Min-Max والتوحيد، لضمان التناسق عبر مجموعات البيانات المتنوعة. يتم إجراء كشف عن القيم الشاذة باستخدام تحليل Z-score لتحديد وإزالة الشذوذات التي قد تؤثر على التوقعات. بالنسبة لبيانات السلاسل الزمنية، يتم استخدام تقنيات التحليل لتقليل الضوضاء والتقاط الاتجاهات طويلة الأجل والتغيرات الموسمية بشكل فعال. بالإضافة إلى ذلك، يتم استخدام هندسة الميزات لاستخراج المعلومات ذات الصلة من مجموعة البيانات.
الشكل 1. توفر هذه التصور نظرة شاملة على خطوات المعالجة المسبقة للبيانات اللازمة لتنظيف وإعداد مجموعة البيانات لنماذج تعلم الآلة.
في الشكل 1، يبرز خريطة الحرارة القيم المفقودة في مجموعة البيانات، وخاصة ضمن سجلات استهلاك الطاقة. يمكن أن تحدث القيم المفقودة بسبب فشل المستشعرات، أو تأخيرات في الإبلاغ، أو تسجيل بيانات غير متسق. إن التعامل مع هذه الفجوات أمر ضروري للحفاظ على دقة النموذج. الرسم البياني الثاني (تقييس الميزات: استهلاك الطاقة قبل وبعد التطبيع) يقارن توزيع قيم استهلاك الطاقة الخام مقابل القيم المعيارية. يضمن التطبيع أن تكون الميزات على مقياس مشابه، مما يمنع نماذج التعلم الآلي من التحيز نحو القيم العددية الكبيرة. الرسم البياني الصندوقي (كشف القيم الشاذة في بيانات استهلاك الطاقة) يحدد القيم المتطرفة في بيانات استهلاك الطاقة. قد تنتج القيم الشاذة عن ارتفاعات غير متوقعة في استخدام الطاقة، أو قراءات عدادات معيبة، أو ظروف جوية قاسية. إن اكتشاف هذه القيم والتعامل معها يمنع التنبؤات المضللة للنموذج. التصور الأخير (مكون الاتجاه لاستهلاك الطاقة على مر الزمن) يستخرج الاتجاه طويل الأجل لاستهلاك الطاقة، مع تصفية التغيرات الموسمية والعشوائية. يمكن أن تشير اتجاهات واضحة صاعدة أو هابطة إلى زيادة في الطلب على الطاقة أو تغييرات في أنماط استخدام الطاقة مع مرور الوقت.

3.2 تطوير النموذج

يتطلب تطوير نماذج تنبؤية لتطبيقات الطاقة المستدامة مزيجًا من تقنيات التعلم الآلي المتقدمة المصممة لتناسب حالات الاستخدام المختلفة. تستخدم هذه الدراسة التعلم العميق، ونماذج الانحدار، والتعلم الجماعي لتعزيز دقة التنبؤ وتحسين توزيع الطاقة. يتم استخدام الشبكات العصبية التكرارية (RNNs) لتوقع استهلاك الطاقة واتجاهات اعتماد المركبات الكهربائية (EV)، مستفيدة من بيانات الاستخدام التاريخية لالتقاط الأنماط الزمنية. يتم نشر الشبكات العصبية التلافيفية (CNNs) والمشفّرات الذاتية (Autoencoders) لاكتشاف الشذوذ في أداء بطاريات المركبات الجديدة (NEV)، مما يضمن الموثوقية والصيانة التنبؤية. تُطبق نماذج الانحدار الخطي لتوقع الطلب على الطاقة في المنازل والصناعات، مع الأخذ في الاعتبار عوامل خارجية مثل الطقس، والأسعار، والمتغيرات الاجتماعية والاقتصادية. بالإضافة إلى ذلك، تُستخدم نماذج الغابة العشوائية (Random Forest) وXGBoost لتجميع استهلاك الطاقة وتوقع الطلب، مما يوفر تنبؤات قوية تأخذ في الاعتبار الاعتماديات غير الخطية في البيانات. تعمل هذه النماذج معًا لتحسين كفاءة الطاقة، وتقليل الفاقد، وتعزيز موثوقية النظام.

3.3 إجراءات تدريب النموذج والتحقق منه

يتطلب تطوير نماذج التنبؤ لتطبيقات الطاقة المستدامة مزيجًا من تقنيات التعلم الآلي المتقدمة المصممة لتناسب حالات الاستخدام المختلفة. تستخدم هذه الدراسة التعلم العميق، ونماذج الانحدار، والتعلم الجماعي لتعزيز دقة التنبؤ وتحسين توزيع الطاقة. يتم استخدام الشبكات العصبية التكرارية (RNNs) للتنبؤ باستهلاك الطاقة وتتبع اتجاهات اعتماد المركبات الكهربائية (EV) من خلال الاستفادة من بيانات الاستخدام التاريخية لالتقاط الأنماط الزمنية. يتم نشر الشبكات العصبية التلافيفية (CNNs) والمشفّرات الذاتية (Autoencoders) لاكتشاف الشذوذ في أداء بطاريات المركبات الجديدة (NEV)، مما يضمن الموثوقية ويمكّن من الصيانة التنبؤية. يتم تطبيق نماذج الانحدار الخطي للتنبؤ بالطلب على الطاقة للأسر والصناعات، مع الأخذ في الاعتبار العوامل الخارجية مثل الطقس، والأسعار، والمتغيرات الاجتماعية والاقتصادية. علاوة على ذلك، يتم استخدام نماذج الغابة العشوائية (Random Forest) وXGBoost لتجميع استهلاك الطاقة وتوقع الطلب، مما يوفر تنبؤات قوية تأخذ في الاعتبار الاعتماديات غير الخطية في البيانات. تعمل هذه النماذج معًا لتحسين كفاءة الطاقة، وتقليل الفاقد، وتعزيز موثوقية النظام. في الشكل 2، يظهر الرسم البياني الدائري تقسيم مجموعة البيانات. تدريب التحقق الاختبار)، بينما تمثل خريطة الحرارة كيفية تنفيذ التحقق المتقاطع بخمسة أضعاف على مجموعة البيانات. يمثل الرسم البياني الخطي استراتيجية التحقق المتقدم المستخدمة في نماذج التنبؤ بالسلاسل الزمنية، مما يضمن عدم استخدام النموذج لمعلومات مستقبلية أثناء التدريب ويساعد في التقاط الاتجاهات الموسمية وطويلة الأجل في الطلب على الكهرباء وإنتاج الطاقة وأنماط الاستهلاك المدفوعة بالمناخ. يقارن هذا الرسم البياني العمودي توزيع الفئات قبل وبعد تطبيق SMOTE (تقنية زيادة العينة الأقلية الاصطناعية) في مجموعات البيانات غير المتوازنة. يتم تطبيق SMOTE لمنع تحيز النموذج وتحسين دقة التصنيف.
الشكل 2. توفر التصورات الأربعة نظرة شاملة على إجراءات التدريب والتحقق المستخدمة في الدراسة.

3.4 مقاييس تقييم الأداء

لتقييم أداء النماذج المطورة بشكل فعال، استخدمنا مجموعة من المقاييس الإحصائية المصممة لكل تطبيق محدد. استخدمنا متوسط الخطأ التربيعي (MSE) وجذر متوسط الخطأ التربيعي (RMSE) لقياس دقة التنبؤ في توقعات الطلب على الطاقة، مع ضمان الحد الأدنى من الانحراف بين القيم المتوقعة والقيم الفعلية. R -مربع ( تم تطبيق ( ) لتقييم القوة التفسيرية لنماذج الانحدار حول استهلاك الطاقة في المنازل والصناعات بشكل نهائي. لاكتشاف الشذوذ في المركبات الكهربائية الجديدة (NEVs)، استخدمنا الدقة والاسترجاع ودرجة F1 لتقييم دقة التصنيف بشكل صارم. علاوة على ذلك، تم إجراء تحليل منحنى خصائص التشغيل المستقبلية (ROC) لتقييم أداء النموذج بشكل قوي في التمييز بين الظروف الطبيعية والظروف المعيبة. في تحليلنا للأثر الاقتصادي، قمنا بتطبيق متوسط الخطأ المطلق (MAE) جنبًا إلى جنب مع اختبارات الدلالة الإحصائية للتحقق بدقة من دقة وموثوقية النماذج التنبؤية. من خلال تطبيق هذه المقاييس التقييمية الصارمة، تؤكد هذه الدراسة بشكل قاطع أن استراتيجيات تحسين الطاقة المدفوعة بالذكاء الاصطناعي ليست دقيقة فحسب، بل قابلة للتنفيذ أيضًا في العالم الحقيقي.

4. النتائج والمناقشة

4.1 أداء النموذج

لتقييم فعالية النماذج المستخدمة في هذه الدراسة، تم استخدام مقاييس أداء متنوعة مثل متوسط الخطأ التربيعي (MSE)، الجذر التربيعي لمتوسط الخطأ التربيعي (RMSE)، معامل التحديد (R-squared) ( تم حساب الدقة، والاسترجاع، ودرجة F1، والمساحة تحت المنحنى (AUC-ROC) لكل مجال. تم تصور مقارنة النماذج المختلفة، بما في ذلك الهياكل العميقة مثل الشبكات العصبية المتكررة (RNNs) والشبكات العصبية التلافيفية (CNNs)، ونماذج الانحدار مثل الانحدار الخطي وXGBoost، بالإضافة إلى طرق التجميع بما في ذلك الغابة العشوائية (Random Forest) وتعزيز التدرج (Gradient Boosting)، لتسليط الضوء على نقاط قوتها وضعفها. في سياق توقع استهلاك الطاقة، برزت الشبكات العصبية المتكررة (RNNs) كأفضل أداء، محققة RMSE قدره 12.5 كيلوواط ساعة و 0.91، مما يدل على قدرة قوية على التقاط الاتجاهات الاستهلاكية المعتمدة على الزمن. تلاها XGBoost عن كثب، مع RMSE قدره 14.2 كيلوواط ساعة و 0.88. بالنسبة لاكتشاف أعطال المركبات الكهربائية، أظهرت النماذج المعتمدة على الشبكات العصبية التلافيفية أداءً متفوقًا، حيث سجلت درجة F1 قدرها 0.92 و AUC-ROC قدرها 0.96، متفوقة بشكل كبير على الانحدار اللوجستي وغابة عشوائية، التي حققت درجات F1 قدرها 0.85 و 0.88 على التوالي. في تحسين الطاقة المتجددة، تميزت نماذج الغابة العشوائية ونماذج تعزيز التدرج بأدنى قيم MSE، حيث حققت الغابة العشوائية MSE قدره 3.1 مقارنة بـ 3.5 لنموذج تعزيز التدرج، مما يبرز فعاليتها في تحسين استخدام الطاقة الشمسية وطاقة الرياح. أخيرًا، في تحليل الأثر الاجتماعي والاقتصادي، بينما قدمت نماذج الانحدار الخطي توقعات قابلة للتفسير مع 0.79 ، تفوق XGBoost عليه بـ 0.85 ، مما يدل على أنه كان أكثر قدرة على التقاط الاعتماديات غير الخطية. الشكل 3 يمثل مدى دقة النماذج في التنبؤ باتجاهات استهلاك الطاقة للأسر والصناعات والمستشفيات. RMSE هو مقياس التقييم المستخدم مع الشبكات العصبية التلافيفية والشبكات العصبية المتكررة التي تعتبر أفضل النماذج أداءً لهذه المهمة. وذلك لأن الشبكات العصبية المتكررة والشبكات العصبية التلافيفية يمكنها التقاط الاعتماديات التسلسلية في بيانات الطاقة الزمنية مقارنةً بالنماذج الأخرى مثل الانحدار الخطي الذي يواجه صعوبة مع الاتجاهات غير الخطية المعقدة للطاقة. تساعد التنبؤات الدقيقة لاستهلاك الطاقة مشغلي الشبكات الكهربائية وصانعي السياسات والشركات في تحسين تخصيص الطاقة وتقليل الفاقد وضمان الاستدامة. الشكل 4 يوضح مدى جودة كل نموذج في تفسير التغيرات في الطلب على الطاقة بناءً على البيانات التاريخية. R -مربع يقيس مقدار التباين في استهلاك الطاقة الذي يشرحه النموذج. كلما كان أعلى يعني أن النموذج يلتقط معظم أنماط الطلب على الطاقة. كانت الشبكات العصبية التلافيفية والشبكات العصبية المتكررة الأفضل أداءً، مما يعني أنها نمذجة اتجاهات استهلاك الطاقة بشكل فعال. كان أداء الانحدار الخطي هو الأسوأ، مما يعني أنه قام بتبسيط أنماط الطاقة المعقدة. يضمن ارتفاع تنبؤات موثوقة، مما يسمح باتخاذ قرارات أفضل في سياسة الطاقة واستراتيجيات التسعير وتخطيط الشبكة المستدام. الشكل 5 يتناول مدى جودة النماذج في اكتشاف الشذوذ والأخطاء في المركبات الكهربائية الجديدة (NEVs) وتهديدات الأمن السيبراني. F1-Score يوازن بين الدقة (الكشف الصحيح عن الشذوذ) والاسترجاع (الكشف عن جميع الشذوذ). يعني ارتفاع درجة F1 أن النموذج يحدد بشكل صحيح أخطاء نظام الطاقة والانتهاكات الأمنية. كان نموذج CNN هو الأفضل أداءً، متفوقًا في اكتشاف الأخطاء في NEVs وتهديدات الأمن. كما أن Random Forest قدم أداءً جيدًا، موازنًا بين الكفاءة والدقة. واجه الانحدار اللوجستي بعض الصعوبات، مما يشير إلى معدل إيجابيات كاذبة أعلى. يساعد الكشف الموثوق عن الأخطاء في منع انقطاع الطاقة، وتحسين أداء المركبات الكهربائية، وتعزيز الأمن السيبراني في الشبكات الكهربائية.
الشكل 3. أداء النموذج في التنبؤ بالطلب على الطاقة باستخدام متوسط الخطأ التربيعي العشوائي (RMSE).
الشكل 4. تصور أداء النموذج في التنبؤ بالطلب على الطاقة باستخدام -مربع ( ).
الشكل 5. تصور أداء النموذج في اكتشاف الشذوذ واكتشاف الأخطاء في المركبات الكهربائية الجديدة (NEVs) وتهديدات الأمن السيبراني.
الشكل 6 يوضح مدى جودة النماذج في التمييز بين سلوك نظام الطاقة الطبيعي والفاسد. AUC-ROC (المساحة تحت المنحنى – خاصية التشغيل المستقبلية) تقيس قدرة النموذج على تصنيف الأخطاء بشكل صحيح. كلما كان AUC-ROC أعلى، كان النموذج أفضل في التمييز بين الحالات الصحية والفاسدة. كان لدى CNN أعلى AUC-ROC، مما يثبت تفوقه في اكتشاف أخطاء NEV واكتشاف الشذوذ في الأمن السيبراني. قدم Random Forest أداءً جيدًا، ولكن أقل قليلاً من CNN. أظهر الانحدار اللوجستي قدرة مخفضة على اكتشاف الشذوذات المعقدة. يضمن AUC-ROC العالي الكشف المبكر عن فشل النظام، مما يمنع انقطاع الطاقة، والانتهاكات الأمنية، وعدم الكفاءة في الشبكات الكهربائية المتجددة.
الشكل 6. تصور أداء النموذج في تصنيف سلوك أنظمة الطاقة الطبيعية والفاسدة.

4.2 المناقشة والعمل المستقبلي

تسلط نتائج هذه الدراسة الضوء على الدور المهم للذكاء الاصطناعي (AI) وتعلم الآلة (ML) في تحسين تطبيقات الطاقة المستدامة. تظهر النتائج أن هياكل التعلم العميق، وخاصة الشبكات العصبية المتكررة (RNNs) والشبكات العصبية التلافيفية (CNNs)، فعالة جدًا في مهام التنبؤ بالسلاسل الزمنية واكتشاف الشذوذ [12]. بالإضافة إلى ذلك، توفر طرق التجميع مثل Random Forest وXGBoost تنبؤات موثوقة لتحسين الطاقة والتحليل الاقتصادي [2]. تلعب تحليلات البيانات الضخمة دورًا حاسمًا في تحسين عمليات اتخاذ القرار من خلال السماح لنماذج الذكاء الاصطناعي بالتكيف ديناميكيًا مع الطلبات المتغيرة للطاقة. بالإضافة إلى ذلك، يمكن أن تعزز نماذج الذكاء الاصطناعي الهجينة التي تجمع بين التعلم العميق وتقنيات التنبؤ الاقتصادي من دقة التنبؤ. يبرز Varian وآخرون (2014) إمكانية دمج تعلم الآلة مع الأساليب الإحصائية التقليدية لتحقيق موثوقية أفضل في التنبؤ، مما يجعلها طريقًا واعدًا للبحث المستقبلي [14].
ومع ذلك، على الرغم من هذه التقدمات، لا تزال هناك عدة تحديات قائمة. أحد القيود الرئيسية هو التعقيد الحسابي، خاصة بالنسبة لنماذج التعلم العميق التي تتطلب قوة معالجة كبيرة ومجموعات بيانات كبيرة. علاوة على ذلك، فإن قابلية تفسير نماذج الذكاء الاصطناعي هي مصدر قلق، خاصة في سياقات صنع السياسات والتنبؤ الاقتصادي، حيث تفضل النماذج القابلة للتفسير [8].
يجب أن تركز الأبحاث المستقبلية على تطوير نماذج ذكاء اصطناعي هجينة تجمع بين التعلم العميق والأساليب الاقتصادية التقليدية لتعزيز كل من الدقة وقابلية التفسير. يمكن أن يؤدي دمج تدفقات البيانات في الوقت الحقيقي من الشبكات الذكية المدعومة بإنترنت الأشياء (IoT) أيضًا إلى تحسين تكيف النموذج وأدائه. بالإضافة إلى ذلك، يجب أن تحقق الدراسات الإضافية في تقنيات التعلم الانتقالي لتمكين النماذج من التعميم عبر سياقات جغرافية واقتصادية متنوعة [1].

5. الخاتمة

تظهر هذه الدراسة بفعالية القدرات الكبيرة للذكاء الاصطناعي (AI) وتعلم الآلة (ML) في تحسين التنبؤ باستهلاك الطاقة، وتعزيز الاستخدام الأكبر للطاقة المتجددة، وتحديد الأخطاء في المركبات الكهربائية الجديدة (NEVs). استخدم الباحثون مجموعات بيانات واسعة شملت استهلاك الطاقة للأسر، واتجاهات اعتماد المركبات الكهربائية، وتحليلات الشبكات الذكية. طبقوا تقنيات الذكاء الاصطناعي المتقدمة، بما في ذلك التعلم العميق، ونماذج الانحدار، و
التعلم التجميعي، لتعزيز دقة التنبؤ بشكل كبير وزيادة كفاءة الطاقة. تؤكد النتائج أن نماذج التعلم العميق، وخاصة الشبكات العصبية التلافيفية (CNNs) والشبكات العصبية المتكررة (RNNs)، تقدم أداءً أفضل من نماذج الانحدار التقليدية في التنبؤ باتجاهات استهلاك الطاقة، محققة دقة أعلى ومعدلات خطأ أقل بشكل ملحوظ. بالإضافة إلى ذلك، تعتبر طرق التجميع مثل Random Forest وXGBoost فعالة للغاية في تحسين توزيع الطاقة المتجددة وتقليل الفاقد في الطاقة. كما تؤكد الدراسة على الدور الأساسي للكشف عن الأخطاء المدعوم بالذكاء الاصطناعي في NEVs، مما يحسن استقرار الشبكة ويدير الطلب على الطاقة بشكل أكثر فعالية. بينما تقدم الدراسة نتائج مثيرة، تعترف بعدة تحديات تحتاج إلى معالجة، بما في ذلك التحيز الخوارزمي، والتعقيد الحسابي، والطلب على دمج البيانات في الوقت الحقيقي. يجب أن تركز الأبحاث المستقبلية على إنشاء نماذج ذكاء اصطناعي هجينة تجمع بين التعلم العميق والأساليب الاقتصادية للتنبؤ لتعزيز كل من الدقة وقابلية التفسير. علاوة على ذلك، فإن دمج الشبكات الذكية المدعومة بإنترنت الأشياء واستخدام تقنيات التعلم الانتقالي أمر حاسم لتكييف النماذج مع مجموعة متنوعة من السياقات الجغرافية والاقتصادية. من خلال معالجة هذه التحديات، تتمتع الحلول المدفوعة بالذكاء الاصطناعي بالقدرة على التأثير بشكل كبير على تطوير نظام طاقة أكثر استدامة وفعالية من حيث التكلفة وصديق للبيئة في الولايات المتحدة.

بيانات المؤلف:

  • الموافقة الأخلاقية: البحث الذي تم إجراؤه لا يتعلق باستخدام البشر أو الحيوانات.
  • تضارب المصالح: يعلن المؤلفون أنهم ليس لديهم مصالح مالية متنافسة معروفة أو علاقات شخصية قد تكون قد أثرت على العمل المبلغ عنه في هذه الورقة.
  • شكر: يعلن المؤلفون أنهم ليس لديهم أحد أو أي شركة ليشكروا.
  • مساهمات المؤلف: يعلن المؤلفون أن لديهم حقوق متساوية في هذه الورقة.
  • معلومات التمويل: يعلن المؤلفون أنه لا يوجد تمويل يجب الإقرار به.
  • بيان توفر البيانات: البيانات التي تدعم نتائج هذه الدراسة متاحة عند الطلب من المؤلف المراسل. البيانات ليست متاحة للجمهور بسبب قيود الخصوصية أو الأخلاق.

References

[1] Ahmed, A., Jakir, T., Mir, M. N. H., Zeeshan, M. A. F., Hossain, A., Hoque Jui, A., & Hasan, M. S. (2025). Predicting Energy Consumption in Hospitals Using Machine Learning: A Data-Driven Approach to Energy Efficiency in the USA. Journal of Computer Science and Technology Studies, 7(1), 199-219.
[2] Barua, A., Karim, F., Islam, M. M., Das, N., Sumon, M. F. I., Rahman, A., & Khan, M. A. (2025). Optimizing Energy Consumption Patterns in Southern California: An AI-Driven Approach to Sustainable Resource Management. Journal of Ecohumanism, 4(1), 2920-2935.
[3] Brynjolfsson, E., & McAfee, A. (2017). The Business of Artificial Intelligence: What It Can – and Cannot Do for Your Organization. Harvard Business Review.
[4] Choi, T. M., Wallace, S. W., & Wang, Y. (2021). Big data analytics in operations management. Production and Operations Management, 30(3), 534-547.
[5] Chowdhury, M. S. R., Islam, M. S., Al Montaser, M. A., Rasel, M. A. B., Barua, A., Chouksey, A., & Chowdhury, B. R. (2024). Predictive Modeling of Household Energy Consumption in the USA: The Role of Machine Learning and Socioeconomic Factors. The American Journal of Engineering and Technology, 6(12), 99-118.
[6] Danks, D., & London, A. J. (2017). Algorithmic bias in autonomous systems. Proceedings of the 26th International Joint Conference on Artificial Intelligence (IJCAI), 4691-4697.
[7] Hossain, M. S., Mohaimin, M. R., Alam, S., Rahman, M. A., Islam, M. R., Anonna, F. R., & Akter, R. (2025). AI-Powered Fault Prediction and Optimization in New Energy Vehicles (NEVs) for the US Market. Journal of Computer Science and Technology Studies, 7(1), 01-16.
[8] Hossain, S., Hasanuzzaman, M., Hossain, M., Amjad, M. H. H., Shovon, M. S. S., Hossain, M. S., & Rahman, M. K. (2025). Forecasting Energy Consumption Trends with Machine Learning Models for Improved Accuracy and Resource Management in the USA. Journal of Business and Management Studies, 7(1), 200217.
[9] Jobin, A., Ienca, M., & Vayena, E. (2019). The global landscape of AI ethics guidelines. Nature Machine Intelligence, 1(9), 389-399.
[10] LeCun, Y., Bengio, Y., & Hinton, G. (2015). Deep learning. Nature, 521(7553), 436-444.
[11] Schneider, T., Lan, S., Stuart, A., & Teixeira, J. (2021). Earth system modeling 2.0: A blueprint for models that learn from observations and targeted high-resolution simulations. Geophysical Research Letters, 48(7), e2020GL091656.
[12] Shil, S. K., Chowdhury, M. S. R., Tannier, N. R., Tarafder, M. T. R., Akter, R., Gurung, N., & Sizan, M. M. H. (2024). Forecasting Electric Vehicle Adoption in the USA Using Machine Learning Models. Journal of Computer Science and Technology Studies, 6(5), 61-74.
[13] Sumon, M. F. I., Osiujjaman, M., Khan, M. A., Rahman, A., Uddin, M. K., Pant, L., & Debnath, P. (2024). Environmental and Socio-Economic Impact Assessment of Renewable Energy Using Machine Learning Models. Journal of Economics, Finance and Accounting Studies, 6(5), 112-122.
[14] Varian, H. R. (2014). Big data: New tricks for econometrics. Journal of Economic Perspectives, 28(2), 3-28.
[15] Waller, M. A., & Fawcett, S. E. (2013). Data science, predictive analytics, and big data: A revolution that will transform supply chain design and management. Journal of Business Logistics, 34(2), 77-84.

Journal: International Journal of Applied Sciences and Radiation Research, Volume: 2, Issue: 1
DOI: https://doi.org/10.22399/ijasrar.19
Publication Date: 2025-03-14

AI and Machine Learning for Sustainable Energy: Predictive Modelling, Optimization and Socioeconomic Impact In The USA

Chidera Victoria Ibeh , Ayodeji Adegbola Harrisburg University of Science and Technology, USA*Corresponding Author Email: victoriaibeh23@gmail.com – ORCID: 0000-0002-5247-785X Independent Researcher, UKEmail: adegbola_ayodeji@yahoo.com – ORCID: 0000-0002-5247-785Y

Article History:

DOI: 10.22399/ijasrar. 19
Received: Jan. 10, 2025
Accepted: Mar. 14, 2025

Keywords:

Artificial Intelligence, Machine Learning, Sustainable Energy, Predictive Modeling, Energy Optimization, Energy Forecasting, Socioeconomic Impact.

Abstract

This research explores how Machine Learning and AI can be used to enhance energy efficiency, forecast energy consumption trends, and optimize energy systems in the USA. This research used datasets comprising household energy usage, electric vehicle adoption trends, and smart grid analytics obtained from public sources, databases, and IoT sensor devices. This study applies advanced machine learning techniques such as deep learning, regression models, and ensemble learning to improve forecasting accuracy aimed at achieving efficient resource allocation. Additionally, this study investigates fault prediction in New Energy Vehicles (NEVs) and its implications for grid stability and energy demand management. The research also examines the socioeconomic impact of AIdriven energy policies and highlights their role in reducing carbon footprints, promoting energy equity, and fostering sustainable economic growth. Recurrent Neural Networks are applied to predict energy consumption trends and electric vehicle(EV) adoption rates by analyzing historical usage data. Convolutional Neural Networks and Autoencoders are used for anomaly detection in NEV battery performance and predictive maintenance. Deep Learning models also use real-time IoT sensor data to enhance the efficiency of energy distribution in smart grids. Linear Regression models are used to predict household and industrial energy demand based on factors such as weather, pricing, and socioeconomic variables. Linear Regression also predicts energy consumption trends in hospitals and factories. Random Forest and XGBoost are used in energy demand forecasting and energy consumption clustering. Performance evaluation metrics such as Mean Squared Error (MSE), Root Mean Squared Error (RMSE), and R-squared ( ) are utilized to assess model accuracy and effectiveness.

1. Introduction

1.1 Background

With advancements in renewable energy technologies, electric vehicles (EVs), and smart grid systems energy consumption in the USA has experienced significant shifts. Since traditional forecasting methods often fail to capture the complex dynamics of energy demand and supply, AI-powered predictive models are used to improve accuracy and decision-making [1-12]. According to Ahmed et al.(2025), machine learning models have been successfully employed to forecast EV adoption trends, assess household energy consumption, and predict hospital energy efficiency, providing insights for sustainable energy policies [1]. AI-driven fault detection in New Energy Vehicles ensures the reliability of these emerging transportation solutions while minimizing energy wastage [7]. The combination of AI and ML in energy research has led to significant breakthroughs in demand-response systems, grid optimization, and renewable resource management [2]. Hossain et al.(2025) articulate that the socioeconomic benefits of these advancements extend beyond efficiency gains, as optimized energy strategies contribute to economic stability, cost savings, and reduced environmental impact [8].

1.2 Importance Of the Research

Sumon et al.(2024) posit that with the increasing demand for sustainable energy solutions, AI and ML have emerged as powerful tools for addressing energy efficiency challenges [13]. Traditional energy
management systems often rely on static models that fail to adapt to real-time fluctuations, leading to inefficiencies and energy waste. This consequently prompts the use of AI and machine learning in energy management systems [12]. Energy providers can also use AI-powered models to dynamically adjust grid loads, optimize renewable resource utilization, and implement real-time fault detection mechanisms [5]. Predictive analytics using AI and machine learning enables accurate forecasting of energy demand, grid stability, and socioeconomic impacts, which in turn fosters data-driven policy formulation to optimize energy consumption patterns in the USA [2]. Furthermore, machine learning techniques enhance energy equity, ensuring fair distribution and reducing energy poverty in underserved communities [1]. Given the pressing need for energy sustainability, climate action, and economic resilience, this research highlights how AI and ML contribute to the long-term viability of intelligent energy management systems.

1.3 Objectives

This research aims to explore how Artificial Intelligence (AI) and Machine Learning (ML) can transform the way we manage and optimize energy consumption in the USA. One key focus is improving energy consumption forecasting, ensuring that households, industries, and hospitals can better predict and plan for their energy needs. Traditional methods often struggle to capture the complexities of energy demand, but AI-driven models can analyze vast amounts of data and uncover hidden patterns to improve accuracy. Another important aspect of this study is renewable energy optimization. With increasing reliance on solar and wind power, it’s crucial to develop models that can balance these energy sources with grid electricity to minimize waste and improve efficiency. Machine learning algorithms can help predict fluctuations in renewable energy generation and ensure a more stable and reliable power supply. Additionally, this research delves into fault detection in New Energy Vehicles (NEVs). As electric vehicles (EVs) become more common, maintaining their reliability and efficiency is essential. AIpowered fault prediction systems can identify potential issues before they become major problems, helping to extend battery life and improve overall performance. Beyond technical optimization, this study also examines the socioeconomic impact of AI-driven energy policies. By analyzing how AI can promote energy equity, reduce carbon footprints, and cut costs for consumers, we aim to provide valuable insights that can shape smarter, more sustainable policies. Ultimately, this research seeks to harness the power of AI and ML to create a more efficient, cost-effective, and environmentally friendly energy future.

2. Literature Review

The application of machine learning and artificial intelligence in the energy sector has gained significant attention in recent years. Sumon et al. (2024) highlighted that AI-based models have been instrumental in assessing the environmental and socio-economic impacts of renewable energy adoption, helping policymakers make more informed decisions [13]. Meanwhile, Shil et al. (2024) explored the predictive capabilities of machine learning for electric vehicle adoption, demonstrating how AI-driven approaches outperform traditional forecasting methods [12].
It was emphasized that the transition to electric vehicles (EVs) has been a focal point of extensive research due to its significance in mitigating climate change, advancing urban planning, and promoting energy sustainability. Early studies in this domain primarily relied on conventional statistical techniques, such as linear regression and time-series analysis, to predict EV adoption trends.
Similarly, Chowdhury et al. (2024) focused on the role of AI in household energy consumption, showing that intelligent modeling techniques can improve energy efficiency and demand-response strategies [5]. The role of AI and machine learning in energy management has significantly transformed predictive analytics and optimization strategies.
Brynjolfsson et al.(2017) highlight how AI revolutionizes industries by facilitating data-driven decisionmaking and optimizing complex systems, including energy grids [3]. The integration of deep learning techniques, such as convolutional neural networks (CNNs) and recurrent neural networks (RNNs), has further enhanced the accuracy of predictive modeling in energy forecasting (LeCun et al. 2015) [10].
Additionally, Waller et al.(2013) emphasize the importance of big data analytics in improving supply chain and energy management, allowing for more precise demand forecasting and greater operational efficiency []15.
While hydropower is often regarded as a clean energy source, it does pose risks to aquatic ecosystems. The construction of dams disrupts the natural river flow and threatens fish populations. Likewise, although geothermal plants have low emissions, they can lead to land subsidence or potentially trigger seismic events in some regions.
Moreover, Hossain et al. (2025) introduced an AI-powered fault detection framework for New Energy Vehicles (NEVs), significantly enhancing battery reliability and vehicle performance [7]. Research by Barua et al. (2025) also underscored the importance of AI in optimizing urban energy consumption patterns, ensuring a more sustainable and efficient energy distribution system [2].
Empirical studies support that energy-efficient technologies and practices have proven effective in various situations. For instance, smart thermostats that utilize machine learning can optimize heating and cooling schedules in homes, potentially leading to a reduction of up to in household energy consumption, as indicated by research on residential energy use. In California, for instance, investigations have revealed significant energy savings resulting from retrofitting homes with advanced insulation and energy-efficient windows, particularly in regions experiencing extreme weather conditions.

2.2 Gaps and Challenges

Despite recent advancements in AI applications for energy management, several gaps remain. One of the most significant challenges is data availability and quality, as energy datasets often contain missing values or inconsistencies that can impact model performance [13].
Additionally, while machine learning models have proven useful for predicting EV adoption rates, realtime consumer behavior and external policy interventions remain difficult to integrate effectively [12]. Gerossier et al. (2024) assert that, despite notable advancements, current predictive methods for electric vehicle (EV) adoption are fundamentally inadequate. These methods fall short in addressing the myriad of high-dimensional and interdependent variables that critically shape the EV market.
Algorithmic bias in machine learning models can lead to an unfair distribution of energy resources, impacting energy equity and policy formulation [6]. Moreover, ethical concerns regarding AI’s role in sustainability must be addressed to ensure responsible deployment [9]. One of the key limitations of current AI models is their struggle to adapt to real-time fluctuations in energy demand.
Schneider et al. (2021) suggest that adaptive AI frameworks, capable of learning from real-time data, could enhance model robustness and traditional approaches are ill-equipped to manage the complexity and nonlinearity of factors driving the diffusion of electric vehicles, including fluctuating fuel prices, inconsistent government subsidies, and rapid technological developments concerning both pledged and commercially available batteries [11].
Time-series models, for instance, may analyze historical data, but they are incapable of effectively capturing abrupt changes in policy and technology, leaving significant gaps in predictive accuracy. Traditional approaches heavily depend on aggregate data and linear assumptions, which obscure variations in regional adoption rates and mask the essential trends in consumer behavior and preferences. This phenomenon is particularly pronounced in the United States, where electric vehicle adoption rates vary significantly across states due to stark differences in policy, climate, and socioeconomic conditions. Chowdhury et al. (2024) highlighted that scalability remains a significant limitation for many AI models, which are often trained on regional data and struggle to generalize effectively at a national level [5]. Hossain et al. (2025) further pointed out that AI-based fault prediction in new electric vehicles (NEVs) is complicated by variations in vehicle design and battery technology [7].
Mamatha et al. (2024) demonstrate that, despite the advantages of NEVs, they are frequently plagued by various faults that significantly impair performance and diminish customer satisfaction. Common issues include traction battery degradation, charging system failures, and software malfunctions. These problems not only undermine vehicle functionality but also lead to increased maintenance costs and a decline in consumer confidence in NEV technology.
The complexity of fault detection and resolution within NEV systems, which stems from their intricate integration of hardware and software, presents a substantial challenge. Addressing these issues requires improvements in data collection methods, enhanced model interpretability, and greater computational efficiency to ensure robust and reliable AI-driven energy solutions.

3. Methodology

3.1 Data Collection and Preprocessing

Data sources

This study draws on a variety of datasets, including records of household and industrial energy consumption, real-time smart meter data, meteorological information, and logs of electric vehicle performance. Additionally, it utilizes publicly available data from the U.S. Energy Information Administration (EIA), the National Renewable Energy Laboratory (NREL), and industry reports detailing energy usage trends. Together, these sources create a comprehensive dataset that will be used to train machine learning models for predicting energy consumption patterns and optimizing the allocation of renewable energy.

Data Preprocessing

To ensure data integrity and enhance model accuracy, a range of preprocessing techniques is implemented. Missing values are addressed through the application of interpolation and imputation methods. Feature scaling techniques, including Min-Max scaling and standardization, are utilized to ensure consistency across diverse datasets. Outlier detection is conducted using Z-score analysis to identify and eliminate anomalies that may skew predictions. For time-series data, decomposition techniques are employed to effectively capture long-term trends and seasonal variations. Additionally, feature engineering is utilized to extract pertinent information from the dataset.
Figure 1. This visualization provides a comprehensive overview of the data preprocessing steps necessary to clean and prepare the dataset for machine learning models.
In Figure 1 the heatmap highlights missing values in the dataset, specifically within Energy Consumption records. Missing values can occur due to sensor failures, reporting delays, or inconsistent data logging. Handling these gaps is essential to maintain model accuracy. The second plot(Feature scaling: Energy Consumption Before and After Normalization) compares the distribution of raw vs. normalized energy consumption values. Normalization ensures that features are on a similar scale, preventing machine learning models from being biased toward large numerical values. The boxplot(Outlier Detection In Energy Consumption Data) identifies extreme values in energy consumption data. Outliers may result from unexpected spikes in energy use, faulty meter readings, or extreme weather conditions. Detecting and handling these values prevents misleading model predictions. The last visualization(Trend Component of Energy Consumption Over Time) extracts the long-term trend of energy consumption, filtering out seasonal and random variations. A clear upward or downward trend can indicate increasing energy demand or shifts in energy usage patterns over time.

3.2 Model Development

The development of predictive models for sustainable energy applications requires a combination of advanced machine-learning techniques tailored to different use cases. This study employs deep learning, regression models, and ensemble learning to enhance forecasting accuracy and optimize energy distribution. Recurrent Neural Networks (RNNs) are utilized for energy consumption prediction and electric vehicle (EV) adoption trends, leveraging historical usage data to capture time-dependent patterns. Convolutional Neural Networks (CNNs) and Autoencoders are deployed for anomaly detection in New Energy Vehicle (NEV) battery performance, ensuring reliability and predictive maintenance. Linear regression models are applied to forecast household and industrial energy demand, incorporating external factors such as weather, pricing, and socioeconomic variables. Additionally, Random Forest and XGBoost models are used for energy consumption clustering and demand forecasting, providing robust predictions that account for nonlinear dependencies in the data. These models work collectively to improve energy efficiency, reduce waste, and enhance system reliability

3.3 Model Training and Validation Procedures

The development of predictive models for sustainable energy applications requires a blend of advanced machine-learning techniques tailored to various use cases. This study employs deep learning, regression models, and ensemble learning to enhance forecasting accuracy and optimize energy distribution. Recurrent Neural Networks (RNNs) are utilized for predicting energy consumption and tracking electric vehicle (EV) adoption trends by leveraging historical usage data to capture time-dependent patterns. Convolutional Neural Networks (CNNs) and Autoencoders are deployed for detecting anomalies in New Energy Vehicle (NEV) battery performance, ensuring reliability and enabling predictive maintenance. Linear regression models are applied to forecast energy demand for households and industries, incorporating external factors such as weather, pricing, and socioeconomic variables. Furthermore, Random Forest and XGBoost models are used for clustering energy consumption and forecasting demand, providing robust predictions that account for nonlinear dependencies in the data. Collectively, these models work together to improve energy efficiency, reduce waste, and enhance system reliability. In Figure 2, the pie chart shows the dataset split ( training, validation, testing), whereas the heatmap represents how 5-fold Cross-validation is performed on the dataset. The Line plot represents the walk-forward validation strategy used in time-series forecasting models, this ensures the model does not use future information while training and helps capture seasonal and long-term trends in electricity demand, energy production, and climate-driven consumption patterns. This bar chart compares the class distribution before and after applying SMOTE (Synthetic Minority Over-sampling Technique) in imbalanced datasets. SMOTE is applied to prevent model bias and improve classification accuracy.
Figure 2. The four visualizations provide a comprehensive overview of the training and validation procedures used in the study.

3.4 Performance Evaluation Metrics

To effectively evaluate the performance of the developed models, we employed a range of statistical metrics tailored to each specific application. We utilized Mean Squared Error (MSE) and Root Mean Squared Error (RMSE) to measure prediction accuracy in energy demand forecasting, ensuring minimal deviation between predicted and actual values. R -squared ( ) was applied to definitively assess the explanatory power of regression models about household and industrial energy consumption. For anomaly detection in New Energy Vehicles (NEVs), we used precision, recall, and F1-score to rigorously evaluate classification accuracy. Furthermore, Receiver Operating Characteristic (ROC) curve analysis was performed to robustly assess model performance in distinguishing between normal and faulty conditions. In our economic impact analysis, we implemented Mean Absolute Error (MAE) alongside statistical significance tests to thoroughly validate the accuracy and reliability of the predictive models. By applying these rigorous evaluation metrics, this study firmly establishes that AI-driven energy optimization strategies are not only precise but also actionable for real-world implementation.

4. Results and Discussion

4.1 Model Performance

To evaluate the effectiveness of the models used in this study, various performance metrics such as Mean Squared Error (MSE), Root Mean Squared Error (RMSE), R-squared ( ), Precision, Recall, F1Score, and Area Under the Curve (AUC-ROC) were calculated for each domain. The comparison of different models, including deep learning architectures like RNNs and CNNs, regression models such as Linear Regression and XGBoost, as well as ensemble methods including Random Forest and Gradient Boosting, was visualized to highlight their strengths and weaknesses. In the context of energy consumption forecasting, RNNs emerged as the top performer, achieving an RMSE of 12.5 kWh and an of 0.91 , which signifies a strong ability to capture time-dependent consumption trends. XGBoost followed closely behind, with an RMSE of 14.2 kWh and an of 0.88 . For electric vehicle fault detection, CNN-based models showcased superior performance, recording an F1-score of 0.92 and an AUC-ROC of 0.96 , significantly outperforming Logistic Regression and Random Forest, which had F1scores of 0.85 and 0.88 , respectively. In renewable energy optimization, Random Forest and Gradient Boosting models excelled with the lowest MSE values, where Random Forest achieved an MSE of 3.1 compared to 3.5 for Gradient Boosting, underscoring their effectiveness in optimizing solar and wind energy utilization. Lastly, in socioeconomic impact analysis, while linear regression models provided interpretable predictions with an of 0.79 , XGBoost outperformed it with an of 0.85 , indicating that it was more capable of capturing nonlinear dependencies. Figure 3 represents how accurate the models are in predicting energy consumption trends for households, industries, and hospitals. RMSE is the evaluation metric used with Convolutional Neural Networks and Recurrent Neural Networks being the best-performing models for this task. This is because RNNs and CNNs can capture sequential dependencies in time-series energy data as compared to other models like Linear Regression which struggles with complex non-linear energy trends. Accurate energy consumption predictions help grid operators, policymakers, and businesses optimize energy allocation, reduce waste, and ensure sustainability. Figure 4 visualizes how well each model explains variations in energy demand based on historical data. R -squared measures how much variance in energy consumption is explained by the model. A higher means the model captures most of the energy demand patterns. CNN and RNN performed best, meaning they effectively modeled energy consumption trends. Linear Regression performed the poorest, meaning it oversimplified complex energy patterns. A high ensures trustworthy predictions, allowing for better energy policy decisions, pricing strategies, and sustainable grid planning. Figure 5 addresses how well the models detect anomalies and faults in New Energy Vehicles (NEVs) and cybersecurity threats. F1-Score Balances precision (correct anomaly detections) and recall (detecting all anomalies). A higher F1 score means the model correctly identifies energy system faults and security breaches. The CNN model performed best, excelling at detecting faults in NEVs and security threats. Random Forest also performed well, balancing efficiency and accuracy. Logistic Regression struggled slightly, indicating a higher false positive rate. Reliable fault detection helps prevent power outages, optimize electric vehicle performance, and strengthen cybersecurity in energy grids.
Figure 3. Model performance for energy demand forecasting with Random Mean Squared Error(RMSE).
Figure 4. Model performance visualization for energy demand forecasting using -squared ( ).
Figure 5. Model performance visualization for anomaly detection and detection of faults in New Energy Vehicles (NEVs) and cybersecurity threats.
Figure 6 visualizes how well the models distinguish between normal and faulty energy system behavior. AUC-ROC (Area Under the Curve – Receiver Operating Characteristic) measures a model’s ability to classify faults correctly. A higher AUC-ROC means the model is better at differentiating between healthy and faulty states. CNN had the highest AUC-ROC, proving its superiority in NEV fault detection and cybersecurity anomaly detection. Random Forest performed well, but slightly lower than CNN. Logistic Regression showed a reduced ability to detect complex anomalies. A high AUC-ROC ensures early detection of system failures, preventing power disruptions, cybersecurity breaches, and inefficiencies in renewable energy grids.
Figure 6. Model performance visualization for classification of normal and faulty energy systems behaviour.

4.2 Discussion and Future Work

The findings of this study highlight the important role of artificial intelligence (AI) and machine learning (ML) in optimizing sustainable energy applications. The results demonstrate that deep learning architectures, particularly recurrent neural networks (RNNs) and convolutional neural networks (CNNs), are very effective for time-series forecasting and anomaly detection tasks [12]. Additionally, ensemble methods such as Random Forest and XGBoost provide reliable predictions for energy optimization and economic analysis [2]. Big data analytics plays a crucial role in refining decisionmaking processes by allowing AI models to adjust dynamically to fluctuating energy demands. Additionally, hybrid AI models that combine deep learning with econometric forecasting techniques can enhance predictive accuracy. Varian et al.(2014) highlight the potential of combining machine learning with traditional statistical methods to achieve better forecasting reliability, making it a promising avenue for future research [14].
However, despite these advancements, several challenges persist. One major limitation is computational complexity, especially for deep learning models that require significant processing power and large datasets. Furthermore, the explainability of AI models is a concern, particularly in the contexts of policymaking and economic forecasting, where interpretable models are preferred [8].
Future research should focus on developing hybrid AI models that combine deep learning with traditional econometric approaches to enhance both accuracy and interpretability. Incorporating realtime data streams from the Internet of Things (IoT)-)-enabled smart grids could also improve model adaptability and performance. Additionally, further studies should investigate transfer learning techniques to enable models to generalize across various geographic and economic contexts [1].

5. Conclusion

This study effectively demonstrates the substantial capabilities of artificial intelligence (AI) and machine learning (ML) in refining energy consumption forecasting, promoting greater use of renewable energy, and identifying faults in New Energy Vehicles (NEVs). Researchers utilized extensive datasets that included household energy consumption, trends in electric vehicle adoption, and smart grid analytics. They applied advanced AI techniques, including deep learning, regression models, and
ensemble learning, to significantly enhance forecasting accuracy and increase energy efficiency. The results confirm that deep learning models, particularly Convolutional Neural Networks (CNNs) and Recurrent Neural Networks (RNNs), deliver better performance than traditional regression models in predicting energy consumption trends, achieving higher accuracy and noticeably lower error rates. Additionally, ensemble methods such as Random Forest and XGBoost are highly effective in optimizing renewable energy distribution and reducing energy waste. The study also emphasizes the essential role of AI-powered fault detection in NEVs, which improves grid stability and manages energy demand more effectively. While the study presents compelling results, it recognizes several challenges that need to be addressed, including algorithmic bias, computational complexity, and the demand for real-time data integration. Future research must focus on creating hybrid AI models that combine deep learning with econometric forecasting approaches to enhance both accuracy and interpretability. Furthermore, integrating IoT-enabled smart grids and utilizing transfer learning techniques is crucial for adapting models to a variety of geographic and economic contexts. By tackling these challenges, AI-driven energy solutions have the potential to significantly impact the development of a more sustainable, costeffective, and environmentally friendly energy system in the USA.

Author Statements:

  • Ethical approval: The conducted research is not related to either human or animal use.
  • Conflict of interest: The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper
  • Acknowledgement: The authors declare that they have nobody or no-company to acknowledge.
  • Author contributions: The authors declare that they have equal right on this paper.
  • Funding information: The authors declare that there is no funding to be acknowledged.
  • Data availability statement: The data that support the findings of this study are available on request from the corresponding author. The data are not publicly available due to privacy or ethical restrictions.

References

[1] Ahmed, A., Jakir, T., Mir, M. N. H., Zeeshan, M. A. F., Hossain, A., Hoque Jui, A., & Hasan, M. S. (2025). Predicting Energy Consumption in Hospitals Using Machine Learning: A Data-Driven Approach to Energy Efficiency in the USA. Journal of Computer Science and Technology Studies, 7(1), 199-219.
[2] Barua, A., Karim, F., Islam, M. M., Das, N., Sumon, M. F. I., Rahman, A., & Khan, M. A. (2025). Optimizing Energy Consumption Patterns in Southern California: An AI-Driven Approach to Sustainable Resource Management. Journal of Ecohumanism, 4(1), 2920-2935.
[3] Brynjolfsson, E., & McAfee, A. (2017). The Business of Artificial Intelligence: What It Can – and Cannot Do for Your Organization. Harvard Business Review.
[4] Choi, T. M., Wallace, S. W., & Wang, Y. (2021). Big data analytics in operations management. Production and Operations Management, 30(3), 534-547.
[5] Chowdhury, M. S. R., Islam, M. S., Al Montaser, M. A., Rasel, M. A. B., Barua, A., Chouksey, A., & Chowdhury, B. R. (2024). Predictive Modeling of Household Energy Consumption in the USA: The Role of Machine Learning and Socioeconomic Factors. The American Journal of Engineering and Technology, 6(12), 99-118.
[6] Danks, D., & London, A. J. (2017). Algorithmic bias in autonomous systems. Proceedings of the 26th International Joint Conference on Artificial Intelligence (IJCAI), 4691-4697.
[7] Hossain, M. S., Mohaimin, M. R., Alam, S., Rahman, M. A., Islam, M. R., Anonna, F. R., & Akter, R. (2025). AI-Powered Fault Prediction and Optimization in New Energy Vehicles (NEVs) for the US Market. Journal of Computer Science and Technology Studies, 7(1), 01-16.
[8] Hossain, S., Hasanuzzaman, M., Hossain, M., Amjad, M. H. H., Shovon, M. S. S., Hossain, M. S., & Rahman, M. K. (2025). Forecasting Energy Consumption Trends with Machine Learning Models for Improved Accuracy and Resource Management in the USA. Journal of Business and Management Studies, 7(1), 200217.
[9] Jobin, A., Ienca, M., & Vayena, E. (2019). The global landscape of AI ethics guidelines. Nature Machine Intelligence, 1(9), 389-399.
[10] LeCun, Y., Bengio, Y., & Hinton, G. (2015). Deep learning. Nature, 521(7553), 436-444.
[11] Schneider, T., Lan, S., Stuart, A., & Teixeira, J. (2021). Earth system modeling 2.0: A blueprint for models that learn from observations and targeted high-resolution simulations. Geophysical Research Letters, 48(7), e2020GL091656.
[12] Shil, S. K., Chowdhury, M. S. R., Tannier, N. R., Tarafder, M. T. R., Akter, R., Gurung, N., & Sizan, M. M. H. (2024). Forecasting Electric Vehicle Adoption in the USA Using Machine Learning Models. Journal of Computer Science and Technology Studies, 6(5), 61-74.
[13] Sumon, M. F. I., Osiujjaman, M., Khan, M. A., Rahman, A., Uddin, M. K., Pant, L., & Debnath, P. (2024). Environmental and Socio-Economic Impact Assessment of Renewable Energy Using Machine Learning Models. Journal of Economics, Finance and Accounting Studies, 6(5), 112-122.
[14] Varian, H. R. (2014). Big data: New tricks for econometrics. Journal of Economic Perspectives, 28(2), 3-28.
[15] Waller, M. A., & Fawcett, S. E. (2013). Data science, predictive analytics, and big data: A revolution that will transform supply chain design and management. Journal of Business Logistics, 34(2), 77-84.