Parameter extraction of photovoltaic models using an improved teaching-learning-based optimization

Highlights

• A new TLBO (ITLBO) is proposed for parameters estimation of solar cells/modules.

• The ITLBO is based on improved teaching and learning strategies.

• The accuracy and reliability of ITLBO is verified through different PV models.

• The ITLBO performs better than most reported algorithms.

Abstract

Accurate and reliable parameter extraction of photovoltaic (PV) models is urgently desired for the simulation, evaluation, control, and optimization of PV systems. Although many meta-heuristic algorithms have been used to extract the PV parameters, the extracted parameters are usually not very accurate and reliable. To accurately and reliably extract the parameters of different PV models, an improved teaching-learning-based optimization (ITLBO) algorithm is proposed in this paper. The novelty of ITLBO lies primarily in the improved teaching and learning strategies with two improvements: (i) the teacher adopts different teaching strategies according to learner levels in the teacher phase; and (ii) in the learner phase, a new learning strategy is proposed to balance exploration and exploitation. The performance of ITLBO is verified by extracting the parameters of the single diode model, the double diode model, and three PV modules. The experimental results indicate that ITLBO obtains better performance with respect to accuracy and reliability compared to the other algorithms.

Introduction

In recent years, because of environmental pollution, climate change, global warming, and fuel exhaustion, the use of alternative renewable energy sources, such as wind, wave, nuclear, tidal, geothermal, biomass, and so on, has received growing attention [1], [2], [3], [4]. Due to its wide availability and cleanliness, solar energy is considered as one of the most promising renewable energy resources [5]. The main application of solar energy is photovoltaic (PV) power generation [6]. Because solar PV systems are able to directly convert solar energy into electricity, they have been applied worldwide [7], [8]. However, using PV systems to generate electricity is an important challenging due to their dependence on weather and environmental factors, particularly temperature and global irradiance [9]. Therefore, to optimize a PV system, an accurate model based on measured current-voltage data is necessary [6], [10]. There are several models that are used to represent the relationship between current and voltage. The most widely used are the single diode and double diode models [11]. The accuracy of models parameters is critical to the study of solar PV systems. Therefore, it is very important to use an effective method to extract the parameters of PV models.

Recently, various methods have been devoted to parameter extraction of PV models. These can be mainly categorized into three groups:

• Analytical methods: The advantages of these methods are simplicity and rapid computation [12] because they usually solve the problem by analyzing a series of mathematical equations [13]. However, some assumptions need to be made before analyzing, which reduces the accuracy of the solutions [14], [15].

• Deterministic methods: These methods [16], [17] are highly sensitive to the initial guess and are easily trapped in local optimum [18]. Additionally, the deterministic methods have strict requirements on the models, such as differentiability and convexity. However, PV models are often implicit, nonlinear, and multi-modal, leading to poor solutions when employing deterministic methods.

• Meta-heuristic methods: To overcome the shortcomings of the first two methods, meta-heuristic methods inspired by natural phenomenon have been serviced as a promising alternative for parameter extraction of PV models. Because these methods do not have strict requirements and are easily implemented, they have recently drawn more attention.

Up to now, many meta-heuristic methods have been used to extract the parameters of PV models, such as particle swarm optimization [19], [20], simulated annealing algorithm [21], genetic algorithm [22], cuckoo search [13], differential evolution [23], [24], bird mating optimizer [25], artificial bee colony (ABC) [26], harmony search-based algorithms [27], artificial bee swarm optimization [28], improved chaotic whale optimization algorithm [29], ant lion optimizer [30], improved JAYA algorithm (IJAYA) [10], bee pollinator flower pollination algorithm [31], and multiple learning backtracking search algorithm (MLBSA) [32]. Although these meta-heuristic methods have obtained satisfactory results, their accuracy and reliability need to be further improved. In addition, there are many algorithmic parameters that need to be set by the users, which may greatly affect the performance of the algorithms.

The teaching-learning-based optimization algorithm (TLBO) [33] is based on the effect of the influence of a teacher on the output of learners in a class. The TLBO is a simple and efficient optimization algorithm, yet only has one algorithmic parameter (i.e., the population size). Recently, several TLBO variants have been used to extract the parameters of PV models, such as TLBO with learning experience (LETLBO) [34], generalized oppositional TLBO (GOTLBO) [35], self-adaptive TLBO (SATLBO) [36], and TLBO artificial bee colony (TLABC) [18]. However, these TLBO variants also suffer from the drawbacks of insufficient accuracy and low reliability, especially for the double diode model.

Based on these considerations, in this paper, an improved TLBO algorithm, namely ITLBO, is proposed to accurately and reliably extract the parameters of different PV models. In ITLBO, two improvements are proposed to overcome the drawbacks of the original TLBO. First, in the teacher phase, the teacher uses different teaching strategies to teach learners according to the learners levels (fitness values), rather than just adopting a teaching strategy like the original TLBO, which will guide all learners to a promising area. Secondly, in the learner phase, we put forward to a new learning strategy where the learners are divided into two groups (i.e., better and worse learners) according to their levels. Better learners are fully utilized to their own exploitation capacity, while worse learners are used to improve global search ability and enhance the diversity of the population. Thus, exploration and exploitation are well balanced. To validate the performance of ITLBO, the algorithm was used to extract the parameters of different PV models, i.e., the single diode model, the double diode model, and the PV modules. The results demonstrate that our approach is able to exactly and reliably extract the parameters of different PV models as well as provide highly competitive results compared with other methods.

The main contributions of this paper are as follows:

• An improved TLBO algorithm, ITLBO, is proposed. In ITLBO, two improved strategies are implemented in the teacher phase and the learner phase.

• The performance of the ITLBO algorithm has been extensively investigated by applying it to the parameter extraction problems of different PV models.

• By comparing with other state-of-the-art algorithms, the accuracy and reliability of ITLBO are demonstrated. Thus, ITLBO can be an effective alternative to parameter extraction of PV models.

The rest of this paper is structured as follows. Section 2 states different PV models and the objective function. The original TLBO algorithm is briefly described in Section 3. Section 4 presents the proposed ITLBO algorithm in detail. Section 5 analyzes the results. Finally, Section 6 concludes the paper.