An Integrated LSTM and RNN Forecasting with Genetic Algorithm Scheduling and Reinforcement Learning Control for Smart Grid Demand Prediction, Peak Shaving, and Distribution Loss Minimization

• Author(s): Ogba Uchenna; Eseosa Omorogiuwa; Sunny Orike
• Paper ID: 1713202
• Page: 2138-2142
• Published Date: 29-12-2025
• Published In: Iconic Research And Engineering Journals
• Publisher: IRE Journals
• e-ISSN: 2456-8880
• Volume/Issue: Volume 9 Issue 6 December-2025

Abstract

Modern electricity networks face increasing complexity due to renewable variability, electrification, irregular consumer behavior, and rising peak loads, which collectively intensify the need for accurate short term forecasting and intelligent load management. This paper presents a hybrid framework that couples deep learning based energy demand prediction with evolutionary day ahead load redistribution. Using more than 6,800 hourly consumption records from 2023, the study applies temporal alignment, anomaly treatment, interpolation, feature engineering, and sliding window transformation to construct supervised learning sequences for forecasting. The Long Short Term Memory model achieves the best predictive performance with MAE of 16.19 kWh, RMSE of 20.36 kWh, and MAPE of 86.58%, outperforming the Recurrent Neural Network with MAE of 16.64 kWh and RMSE of 20.85 kWh, while the RNN’s slightly lower MAPE of 82% is linked to sensitivity under very low night time loads. For load management, daily profiles are clustered via K Means, and 20% of shiftable load is redistributed within a 2 hour window using a Genetic Algorithm. The optimization runs for 120 generations with a population of 60 candidate schedules and produces a noticeable reduction in peak demand and operational cost while maintaining total daily energy consumption. Overall, the framework demonstrates stable convergence, lower performance variance, and practical peak shaving capability for smart grid decision support, with clear extension paths to multi horizon forecasting and real time pricing integration. (smart grid, energy forecasting, LSTM, RNN, genetic algorithm)

Citations

IRE Journals:
Ogba Uchenna, Eseosa Omorogiuwa, Sunny Orike "An Integrated LSTM and RNN Forecasting with Genetic Algorithm Scheduling and Reinforcement Learning Control for Smart Grid Demand Prediction, Peak Shaving, and Distribution Loss Minimization" Iconic Research And Engineering Journals Volume 9 Issue 6 2025 Page 2138-2142 https://doi.org/10.64388/IREV9I6-1713202

IEEE:
Ogba Uchenna, Eseosa Omorogiuwa, Sunny Orike "An Integrated LSTM and RNN Forecasting with Genetic Algorithm Scheduling and Reinforcement Learning Control for Smart Grid Demand Prediction, Peak Shaving, and Distribution Loss Minimization" Iconic Research And Engineering Journals, 9(6) https://doi.org/10.64388/IREV9I6-1713202