AI based Approach for Solar PV Power Prediction with Varying Dust Accumulation Levels

Energy demand and concerns over greenhouse gases have made the integration of solar PV into the grid imperative. Solar power forecasting models must have a high prediction accuracy to address the intermittent nature of solar irradiation. Solar PV power is significantly affected by the dust deposited on the PV panel surface. The amount of dust deposited on PV panel as one of the input parameters to predict solar PV power and solar irradiation is performed in this paper. Multivariate analysis of three deep learning techniques that is LSTM (Long short-term memory), 1D CNN (Convolution Neural Network) and BilSTM (Bidirectional Long short-term memory) to predict the solar PV power and solar irradiation with varying dust accumulated levels for the 335 watt PV module installed on the rooftop of the lab at Delhi Technological University is presented. An artificial dust scenario is created by continually incrementing the dust level by 1.258 mg/cm2.