Improved load demand prediction for cluster microgrids using modified temporal convolutional feed forward network

This research addresses the challenge of accurate load forecasting in cluster microgrids, where distributed energy systems interlink to operate seamlessly. As renewable energy sources become more widespread, ensuring a consistent and reliable power supply in the face of variable weather conditions is a significant challenge for power providers. The variability in energy consumption patterns, influenced by human behavior and environmental conditions, further complicates load prediction. The inherent instability of solar and wind energies adds complexity to forecasting load demand accurately. This paper suggests a solution in addressing some challenges by proposing a Modified Temporal Convolutional Feed Forward Network (MTCFN) for load forecasting in cluster microgrids. The Fire Hawk Optimization algorithm is employed to determine optimal configurations, addressing the intricacies of this complex optimization problem. Data collected from the Microgrid Market Share and Forecast 2024-2032 report, the efficiency of the proposed approach is evaluated through metrics such as Mean Absolute Error (MAE), Root Mean Square Error (RMSE), Mean Absolute Percentage Error (MAPE), Mean Square Error (MSE), and R-squared. The RMSE, MSE, MAE, MAPE, and R-squared values of the MTCFN are 0.4%, 1.5%, 0.6%, 6.8%, and 0.8%, respectively. The optimization algorithm's effectiveness is cross-validated through rigorous testing, training, and validation processes, revealing that the FFNN model based on the Fire Hawk Optimization algorithm yields superior load forecasting results. This research contributes to the advancement of signal, image, and video processing in the context of resilient and accurate energy management in cluster microgrids.