Hybrid model for microgrid short term load forecasting based on machine learning

Short-term load forecasting (STLF) is crucial for microgrid (MG) operators to optimize energy generation and storage schedules based on anticipated load variations. Accurately predicting peak demand periods enables operators to ensure sufficient power supply while minimizing reliance on expensive backup sources, resulting in cost savings, and improved overall system efficiency. Residential MG power demand is highly dynamic due to external factors like residents' lifestyles, behaviors, and weather responses, leading to significant irregularity and management challenges. To deal with these challenges, we propose a hybrid STLF model that combines Artificial Neural Networks (ANN) and Fuzzy Logic (FL), referred to as the Adaptive Neuro-Fuzzy Inference System (ANFIS). This model was trained and tested using real power consumption data. We evaluated the performance of the ANFIS model by comparing it with another ANN model using three evaluation metrics: Mean Squared Error (MSE), Root Mean Squared Error (RMSE), and Mean Absolute Percentage Error (MAPE). The trained ANFIS model achieved an MAPE of 8.7528% and an RMSE of 0.4752kw, while the ANN model achieved an MAPE of 8.8123% and an RMSE of 0.4816kw. These results confirm the accuracy of the hybrid ANFIS model compared to ANN. The ANFIS model demonstrated its ability to capture complex and nonlinear relationships between various factors affecting load demand, making it suitable for handling the dynamic nature of MG load.