Machine Learning-Based Solar Irradiance Prediction for Energy Management

The deployment of a distributed generator (DG) and Volt-VAr control (VVC) devices in distribution systems offer energy management and voltage regulation. VVC devices include optimal coordination among on-load tap changers (OLTC), capacitor banks (CBs), and voltage regulators (VRs). The proposed method presents a two-stage energy management model (EMM) for the distribution system. In the first stage, a multi-layer perceptron (MLP) based machine learning technique (ML) is applied to predict solar irradiance accurately. In contrast, the second stage includes optimal coordination among DG and VVC devices to minimize substation energy demand while meeting system operational constraints. The proposed method is verified on the modified IEEE 69-node system. The performance of the MLP technique is compared with the seasonal auto-regressive integrated moving average (SARIMA) time-series model. The obtained results show the superior performance of MLP over SARIMA. In addition, the outcome of the simulated results reveals that the combined method of DG and VVC devices produces significant energy savings and mitigates voltage problems.