COMMITTEE CHAIR: Dr. Sarhan Musa
TITLE: OPTIMIZATION OF SOLAR ENERGY EFFICIENCY USING NEURAL NETWORK CONTROLLERS WITH DIRECT CURRENT CONVERTER
ABSTRACT: Solar systems rely entirely on sunlight to convert energy, so it can be categorized as a totally renewable energy source for as long as humans have lived on Earth. An electrical system that converts photovoltaic energy into electrical energy is called a Photovoltaic (PV) solar system. A PV array is any class of electrical machines that converts photovoltaic energy into electrical energy. The use of a PV solar system has many advantages over other conventional electrical generating methods, such as there is no sound pollution in collecting solar energy via solar panels, rather, it is by any means a silent machine. Green initiatives are higher on the global agenda than ever. Unlike traditional energy sources, when PV solar systems convert energy, they do not emit any gases, especially harmful greenhouse gases. The Photovoltaic (PV) system power generation has a challenge in voltage and power control because of its nonlinearity in structure. The wide use of the algorithm of the Perturbation and Observation (P&O) controller shows a significant decrease in efficiency as the change in Irradiance or Temperature values are greatly spaced out for a small number of time periods. This research work presents controllers to control the power of any PV system with a Maximum Power Point Tracker (MPPT). For some time, research has focused on Artificial Intelligence (AI) to find solutions to maximize the efficiency of the PV system output power. With AI showing to have a great performance and higher efficiency in predicted power outputs, a focus on an Artificially Neural Network (ANN) controller and Recurrent Neural Networks (RNN) controller with the use of a DC-DC Boost converter is presented as having improved the control of the MPPT. Findings also suggest that these AI controllers have a great response in adjusting their MPP with small or great changes in input conditions, as by definition. A PV array output power commonly changes as two basic variables change, Irradiance (G) and Temperature (T).
The traditional P&O controller can be used to control the MPPT of any PV system. For nonlinear control design, the P&O is preferred to be used in a small number of instant changes in G or T values; for many instant changes in G in T values, P&O will not be the best control method to be used as the efficiency of the overall system will decrease. AI controllers such as ANN or RNN controllers respond greatly to nonlinear mathematical models. ANN and RNN are also used in linear controls; but as the change in the values of G and T becomes large, ANN and RNN will have greater higher efficiency in prediction for such applications, thus ANN and RNN MPPT controllers are implemented. We compared these AI controllers ANN and RNN with a DC-DC Boost, Cuk, Singel Ended Primary Inductance Converter (SEPIC), Positive Output Super Lift Luo (P/O SLL), and Ultra Lift Luo (ULL) converters to predict outputs based on varying input samples, simulated using the MATLAB software package and validated by MATLAB implementation. The AI-based converters using RNN with a DC-DC Boost, Cuk, Singel Ended Primary Inductance Converter (SEPIC), Positive Output Super Lift Luo (P/O SLL), and Ultra Lift Luo (ULL) converters was the optimal method based on our results, surpassing the ANN with a DC-DC Boost, Cuk, Singel Ended Primary Inductance Converter (SEPIC), Positive Output Super Lift Luo (P/O SLL), and Ultra Lift Luo (ULL) converters in terms of output power efficiency.
Keywords: Maximum power point tracking, Artificial neural network, Recurrent neural network, DC-DC converters, Photovoltaic system.
Room Location: ENCARB, Room 363.