Energy forecasting of the building integrated photovoltaic system based on deep learning dragonfly-firefly algorithm

Building integrated photovoltaic (BIPV) technology is an emerging technology that harnesses solar energy. The BIPV system enhances the energy consumer to energy production in modern buildings. The performance of the BIPV system varies depending on geographical location, seasonal conditions, and environmental parameters. The performance prediction of the building-integrated photovoltaic system plays a vital role in the energy forecast and consumption pattern. In this work, the performance of the BIPV system output power is predicted based on solar radiation, ambient temperature, and wind speed in hot and humid climatic conditions. The study proposes a new neural network method, long short-term memory (LSTM), with feature selection using the dragonfly (DF) and firefly algorithms(FF). The hybrid deep learning algorithm (LSTM-DF-FF) predicts the performance. The performance metrics are used to evaluate the performance of the model. When conditions are ideal, LSTM–FF–predictions DFs have a relative error of just 3.5 %. Network forecasts are less dependable and accurate on days with clouds and rain, with relative errors of 7.8 and 10.1 %, respectively. The LSTM–FF–DF model had the highest correlation in all conditions, with 0.997 in the absence of clouds and 0.991 under overcast. •Novel Hybrid Dragonfly -Firefly optimisation algorithm is implemented.•The LSTM–FF–DF model had the highest correlation in all conditions.•This method improves the accuracy of power projections for PV plants.