Photovoltaic cell parameter estimation based on improved equilibrium optimizer algorithm

Parameter estimation of photovoltaic cells is essential to establish reliable photovoltaic models, upon which studies on photovoltaic systems can be more effectively undertaken, such as performance evaluation, maximum output power harvest, optimal design, and so on. However, inherent high nonlinearity characteristics and insufficient current–voltage data provided by manufacturers make such problem extremely thorny for conventional optimization techniques. In particular, inadequate measured data might save computational resources, while numerous data is also lost which might significantly decrease simulation accuracy. To solve this problem, this paper aims to employ powerful data-processing tools, for instance, neural networks to enrich datasets of photovoltaic cells based on measured current–voltage data. Hence, a novel improved equilibrium optimizer is proposed in this paper to solve the parameters identification problems of three different photovoltaic cell models, namely, single diode model, double diode model, and three diode model. Compared with original equilibrium optimizer, improved equilibrium optimizer employs a back propagation neural network to predict more output data of photovoltaic cell, thus it can implement a more efficient optimization with a more reasonable fitness function. Besides, different equilibrium candidates of improved equilibrium optimizer are allocated by different selection probabilities according to their fitness values instead of a random selection by equilibrium optimizer, which can achieve a deeper exploitation. Comprehensive case studies and analysis indicate that improved equilibrium optimizer can achieve more desirable optimization performance, for example, it can achieve the minimum root mean square error under all three different diode models compare to equilibrium optimizer and several other advanced algorithms. In general, the proposed improved equilibrium optimizer can obtain a highly competitive performance compared with other state-of-the-state algorithms, which can efficiently improve both optimization precision and reliability for estimating photovoltaic cell parameters.