Node Load Forecasting Method with Distributed Rooftop Photovoltaic Considering Meteorological Conditions

As new energy sources become integrated into the power grid, the proportion of distributed generation (DG) access buses is on the rise. The inherent instability of distributed generation seriously affects the load forecast of nodes, while the conventional load forecasting methodology is unable to meet the requisite accuracy standards. To address the challenge of node load forecasting for diverse user categories, including distributed rooftop photovoltaics, an enhanced wavelet neural network approach to node load prediction was developed. This method builds upon the Elman neural network and represents a significant advancement in the field. This method exhibits robust learning capabilities, a straightforward structure, rapid convergence, and high accuracy. Initially, the K-means is employed to categorize the load to three distinct groups: residential, commercial and industrial consumers. Subsequently, a short-term PV prediction pattern that incorporates weather category index is established according to the photovoltaic ratios of different users. This method is trained with the improved wavelet and Elman neural network, thereby enhancing the model's predictive ability. The net prediction load is obtained by combining the photovoltaic prediction with the load curve characteristics of each user. In conclusion, due to illustrate the efficacy of the model, an example is presented as an actual area with distributed roof photovoltaics.