Optimized LSTM-based electric power consumption forecasting for dynamic electricity pricing in demand response scheme of smart grid

Accurate Electric Power Consumption (EPC) forecasting is necessary for smart grid systems in order to improve Demand Response (DR) strategies and optimize energy use. The main objective of this work is to systematically tune hyperparameters to enhance the performance of Long Short-Term Memory (LSTM) models for time-series forecasting. Accurate EPC forecasts significantly influence peak demand management and the application of dynamic electricity pricing techniques, which is the motivation behind it. Choosing the best LSTM hyperparameters is part of the proposed method to increase forecasting robustness and accuracy. Day-ahead Dynamic Electricity Pricing (DDEP) is used in a Strategic Demand Response Program (DRP) to illustrate the use of this enhanced forecasting. The system allows customers to efficiently schedule Price-Dependent Loads (PDL) and Electric Vehicle (EV) charging sessions by including the forecasted EPC into DDEP. The analysis is done on metrics for performance, including Root Mean Squared Error (RMSE), Mean Squared Error (MSE), Mean Absolute Error (MAE), and root-mean-squared correlation (R2). Based on the obtained results, the most accurate performance is 0.4454 for RMSE, 0.1984 for MSE, 0.3224 for MAE, and 0.9677 for R2. This strategy improves system dependability, attains a balanced demand-supply equilibrium, and delays the need for transmission and distribution system upgradation. The results highlight the benefits of using hyperparameter-tuned LSTM models for forecasting and incorporating them into dynamic pricing systems for a smart grid.