A short-term PV resource assessment method with parallel DenseNet201 and BiLSTM under multiple data features

Recently, renewable energy has been widely deployed. Among all renewable energy sources, photovoltaic power is widely deployed in the grid. Because of the instability of PV power generation, PV grid connection affects electrical grid stability and reliability. An accurate assessment of PV resources is therefore essential to the robustness and reliability of the power system. To increase the precision of PV resource assessment, this study proposes a short-term PV resource assessment method with parallel DenseNet201 and BiLSTM (PDBL). This study predicts the power generation for 24 hours on day 15 with irradiation intensity, full-field power, and temperature for each hour of 14 days at the PV farm. In this study, irradiation intensity data, full-field power data, and field temperature data are fed into BiLSTM, BiLSTM, and DenseNet201 to predict PV farm power respectively; results of three networks are optimized with a deep fully connected layer network to obtain prediction results with higher accuracy. Experiments on the same data set show that the root mean square error (RMSE) of the PDBL short-term PV resource assessment approach is 7.5278, which is 15.8143 smaller than the RMSE of BiLSTM.