Scientists have developed a machine learning model using K-means and long short-term memory techniques that aims to overcome ‘fault detection and classification’ in the operation and maintenance of large-scale solar power plants .
Scientists from Malaysia and Thailand have developed a new machine learning model to predict the maintenance needs of large solar power plants. According to a recently published scientific paper, paper The model touting this effort utilizes two machine learning techniques: K-means and long short-term memory (LSTM), and has a root mean square error (RMSE) of 0.7766. The tool aims to overcome “fault detection and classification” commonly found in traditional operating systems, the paper claims.
“Traditional operation and maintenance (O&M) of photovoltaic systems does not utilize machine learning for fault detection and classification,” the paper states. “This makes it difficult for plant operators, especially large-scale solar power (LSS) PV plants, who typically rely on manual approaches to screen large amounts of electrical data and inspect large numbers of string panels. It creates challenges for the operators who manage it, resulting in higher O&M costs.”
K-Means is a data segmentation algorithm that groups similar data points into clusters. Researchers use this algorithm to cluster string module currents with environmental factors such as global solar radiation and module temperature. A center point or mean point is then created that represents the typical behavior of the cluster.
Next, an LSTM method trained on historical data is started. This technique aims to detect anomalies in the expected current of string modules and alert the operator to the need for maintenance.
“LSTMs can process sequential data of variable length input by using a gating mechanism to decide at each timestep what information to keep and what information to discard, so it can handle past trends in the input sequence. and patterns can be used to make predictions,” the paper states. . “LSTM accomplishes this by using a special type of memory cell that can store information for long periods of time, and gates that control the flow of information into and out of the cell.”
The training data and method accuracy analysis are based on information obtained from a large-scale solar power plant located in central Malaysia. Turnkeys and subinverters were used as test cases to monitor 420 string modules and a total of 8,400 PV modules. Compared to the collected data, the root mean square error (RMSE) of the model is 0.7766.
The relative errors are then compared to numbers set by benchmark models based on artificial neural networks (ANNs). “LSTMs and ANNs are often compared because they both belong to neural networks and are commonly used in various natural language processing, computer vision, and speech recognition tasks,” the paper states. They found that LSTM was more accurate, with an average LSTM relative error of 4.316% and ANN relative error of 4.363%.
This algorithm is based on the research “Anomaly Detection Using K-Means and Long Short-Term Memory for Predictive Maintenance of Large Scale Solar Power (LSS) Solar Power Plants” was recently published. energy report. The research group was formed by scientists from Universiti Teknologi Malaysia and Chiang Mai Ratchabhat University.
This content is copyrighted and may not be reused. If you would like to collaborate with us and reuse some of our content, please contact us at [email protected].