Deep Learning LSTM-based MPPT Control of 100kW Dual Stage Grid Tied Solar PV System

Abstract

Abstract The simultaneous rise in energy demand brought on by urbanization, industrialization, population growth, and the significant increase in greenhouse gas emissions from conventional energy sources pushes the energy market to divert towards sustainable energy. Among renewables, Solar photovoltaic (PV) technology has been identified as an abundant, clean, environmentally friendly, noiseless, and economically sustainable energy source to fulfill the future energy demand. However, the output power of a solar PV panel is unpredictable due to temperature (T) and irradiance (G) fluctuations, as well as the relatively low efficiency of solar cells (15 to 25%) limits its applications in grid-connected mode. To work for the PV panel at its maximum power, this paper presents the deep learning associated with Long Short Term Memory (LSTM) network-based Maximum Power Point Tracking (MPPT) controller for a 100 kW grid-connected PV array. The performance of the proposed LSTM-based MPPT is contrasted with that of the Feed Forward Neural Network (FFNN) and the traditional Perturb and Optimization (P&O) MPPT controller using the Simulink MATLAB environment. Over one million datasets, the LSTM and FFNN are trained for two inputs (T, G) and a single output (Vmp). The Mean Square Error (MSE), Root Mean Square Error (RMSE), Mean Average Error (MAE), and Prediction error between the actual power and the extracted power by the respective MPPT are used as performance indices in the comparison of LSTM and FFNN. The trained models are exported to Simulink, where an MPPT comparison is accomplished among the LSTM, FFNN, and P&O controllers. LSTM-based MPPT controller extracted more power in kilo watt (99.14) from the PV panel than FFNN (96.75) and P&O (95.11) controllers. The LSTM comprised of least RMSE value (0.20) than FFNN (2.62), and P&O (4.22) respectively. Hence, the proposed LSTM MPPT controller proceeded to establish the control of active power between the PV array and grid, Direct Current (DC) bus voltage control, and grid-tied inverter control

Keywords

Identifiers

DOI

10.21203/rs.3.rs-2539136/v1

Journal

Research Square

Year

2023