Towards green energy for sustainable development: Machine learning based MPPT approach for thermoelectric generator

Abstract

The rapidly increasing demand for electrical energy and the depletion of fossil fuels ignited research in renewable energy generation and management. Thermoelectric generation (TEG) systems as a green energy resource are designed to recover waste heat recovery. Concentrated solar tackles the disadvantages of low power generation efficiency. This paper investigates the feasibility of using machine learning (ML) based MPPT technique, to harvest maximum power of a centralized TEG system under various operating conditions. The environment-dependent issues such as the non-constant distribution of heat, loss to the heat transfer coating between sinks and sources, and mechanical faults create non-uniform current generation and impedance mismatch causing power loss. In this article, we have proposed a feed-forward neural network (FNN) trained by a novel flow direction algorithm (FDA) with a tuned PID controller to harvest the energy under non-uniform temperature gradient conditions. The non-uniformity of physical conditions generates multiple local maxima's (LMs) in electrical characteristics of TEG systems. Common gradient-based MPPT techniques are unlikely to track true GMPP. As a solution to this non-linear problem, novel FNN-FDA is tested under non uniformly distributed temperature (NTUD) over heat source surfaces under varying load and temperature conditions. In this study, certain contributions to the field of TEG systems were made by tackling the issues such as GMPP tracking, low efficiency, and oscillations around GMPP. The results are compared to the particle swarm optimization (PSO), Barnacle mating optimization (BMO), and grey wolf optimization (GWO) algorithm. Five experiments under different weather conditions are performed. Results are validated and proved with experiments and MATLAB/SIMULINK. The experimental results demonstrated that FNN-FDA performs significantly better when compared with other control techniques.