Techno-economic-environmental optimization of hybrid photovoltaic-thermoelectric generator systems based on data-driven approach

Hybrid photovoltaic-thermoelectric generator (PV-TEG) system combines two types of energy conversion which is an important innovation to advance the development of renewable energy technologies. Hybrid system in practice needs to track the best operating point in real-time with the help of maximum power point tracking (MPPT) technology to ensure the system outputs maximum energy and optimizes performance. To improve the energy conversion efficiency and utilization of hybrid PV-TEG systems, as well as to effectively combat the negative impacts under partial shading conditions (PSC) and non-uniform temperature distribution (NTD), this paper proposes a hybrid system MPPT technique based on the bi-directional long and short-term neural networks (Bi-LSTM). The Bi-LSTM can consider both past and future data information, which will enhance the training process of the network and the accuracy of the model, making the results more comprehensive. It approaches the global maximum power point by searching for the maximum power down in the forward layer and up in the backward layer. The algorithm can effectively prevent from falling into local optimum under PSC and NTD. The case study section tests and explores the feasibility of utilizing Bi-LSTM to obtain the maximum power from a hybrid PV-TEG system under five case study scenarios, namely, start-up test, step change in solar irradiance, stochastic variation, Hong Kong four-season real case, and uncertainty analysis. Comparative analysis with the other five algorithms are also conducted to thoroughly verify the superiority of the proposed technique for hybrid system MPPT applications. The simulation results indicate that the hybrid PV-TEG system with Bi-LSTM achieves the optimal MPPT performance stably and efficiently under different operating conditions. In particular, low irradiate condition in spring, the energy generated by Bi-LSTM exceeds the INC, SAO and P&O &O energy outputs by 80.51%, 61.22%, and 31.04%, respectively.