A New Combined PV Output Power Forecasting Model Based on Optimized LSTM Network

One of the most challenging problems in designing and implementing effective management strategies and demand responses in renewable-rich grids is the uncertainty associated with the power output (PO) of solar photovoltaic (PV) systems. The exact and trustworthy prediction of PV power can provide substantial decision support for planning and operating power systems. This paper develops an intelligent PV output power (PV-OP) forecasting model. The proposed PV-OP forecasting model consists of an Improved Complete Ensemble Empirical Mode Decomposition with Adaptive Noise (ICEEMDAN) signal decomposition model to decompose the original PV output power signal to different frequencies, LSTM neural network as a main forecaster engine, and multi-objective NSGA-II optimization algorithm as a hyperparameters optimizer. The developed PV power forecasting model was validated using the PV power datasets of an off-grid village located in Chalokwa, Zambia. The obtained results confirmed the performance and accuracy of the proposed PV output power forecasting model.