In the quest to harness the sun’s energy more efficiently, researchers have developed a novel method to diagnose faults in photovoltaic (PV) arrays, potentially saving the solar power industry millions in maintenance and downtime costs. The study, led by YANG Chen from the College of Engineering Science and Technology at Shanghai Ocean University, introduces an advanced fault diagnosis technique that combines an improved grey wolf optimization algorithm with an extreme learning machine (ELM). The research was published in the journal *Power Technology*.
Photovoltaic arrays, which convert sunlight into electricity, often operate under challenging outdoor conditions. These conditions can lead to various faults, ranging from minor inefficiencies to complete system failures. Accurately identifying and diagnosing these faults is crucial for maintaining the reliability and efficiency of solar power generation.
The method proposed by YANG Chen and his team addresses the limitations of traditional fault diagnosis techniques. By analyzing nine fault simulation output characteristics, the researchers established a five-dimensional fault feature vector. This vector includes key metrics such as short-circuit current, open-circuit voltage, maximum power point current, maximum power point voltage, and fill factor.
One of the standout features of this research is the improvement made to the grey wolf optimization algorithm. “We incorporated Circle mapping and nonlinear convergence factors to enhance the algorithm’s performance,” explained YANG Chen. “This improvement addresses the issues of uneven distribution of initial positions and the imbalance between global search and local exploitation, which are common challenges in traditional grey wolf algorithms.”
The improved grey wolf optimization algorithm is then used to optimize the input layer weights and hidden layer node biases of the extreme learning machine. This optimization process significantly enhances the performance of the ELM, making it more accurate and efficient in diagnosing faults.
To validate their method, the researchers developed simulation models and experimental platforms to collect fault data. Using K-fold cross-validation, they divided the data and input it into the IGWO-ELM model. The results were impressive, with the model achieving a classification accuracy of 98.32% for simulation data and 95.48% for experimental data.
The implications of this research for the energy sector are substantial. Accurate and rapid fault diagnosis can lead to more efficient maintenance schedules, reduced downtime, and ultimately, lower costs for solar power generation. “Our method offers high accuracy, requires fewer iterations, and achieves fast convergence speed,” said YANG Chen. “This makes it a valuable tool for effectively judging the working state of photovoltaic arrays.”
As the world continues to shift towards renewable energy sources, advancements in solar power technology are more important than ever. The research by YANG Chen and his team represents a significant step forward in this field, offering a powerful new tool for maintaining the reliability and efficiency of photovoltaic arrays. With the growing demand for clean energy, such innovations are crucial for ensuring the sustainability and economic viability of solar power generation.