Intelligent Load Forecasting with Optimized Renewable Energy Management for Non-Linear Load

Due to the increase in energy costs, climatic changes and electrical infrastructure losses, the usage of more renewable sources is expanding. Because of this fact, hybrid energy sources play an important role in the energy management system. This paper introduces a double-deep dueling Q-learning algorithm (D3QLA) for forecasting the load side’s power generation. The weighted mean of the vector optimization algorithm (WMOV) is introduced to enhance the tradeoff between the generated and consumed energy. The proposed deep reinforcement algorithm schedules the power of non-linear loads with minimum errors. The simulation is executed via the MATLAB/Simulink load forecasting improvement tool with optimized renewable sources. The simulation results are recognized with different cases to check the robustness of the proposed method. For the industrial load, the MAE of the proposed is 1.620, the MAPE is 0.316, the SDE is 0.023, and the RMSE is 0.095. Moreover, the MAPE for the residential load of the proposed method is 0.200, MAE is 7.637, RMSE is 0.097, and SDE is 0.201. The proposed method reduces the mean absolute error (MAE) of 0.007% for industrial load and 0.004% for residential load when equated with the Double Deep Q-Learning (DDQL) algorithm. Therefore, the proposed WMOV-optimized D3QLA adequately predicts the energy on the load side with minimal power matrix errors.