An Information Granulated Based SVM Approach for Anomaly Detection of Main Transformers in Nuclear Power Plants

Abstract

The main transformer is critical equipment for economically generating electricity in nuclear power plants (NPPs). Dissolved gas analysis (DGA) is an effective means of monitoring the transformer condition, and its parameters can reflect the transformer operating condition. This study introduces a framework for main transformer predictive-based maintenance management. A condition prediction method based on the online support vector machine (SVM) regression model is proposed, with the input data being preprocessed using the information granulation method, and the parameters of the model are optimized using the particle swarm optimization (PSO) algorithm. Using DGA data from the NPP data acquisition system, two experiments are designed to verify the trend tracing and prediction envelope ability of main transformers installed in NPPs with different operating ages of the proposed model. Finally, how to use this framework to benefit the maintenance plan of the main transformer is summarized.

A Novel Wind Power Outlier Detection Method with Support Vector Machine Optimized by Improved Harris Hawk

“…In summary, the IHHO-SVM model demonstrates the best detection accuracy, as well as strong stability and generalization performance. In order to further validate the performance of the IHHO-SVM model in wind power outlier detection, comparative test was conducted by selecting commonly employed machine learning models for outlier detection, including isolation forest (IF) [17], local outlier factor (LOF) [18], SVM, as well as combined models of SVM with widely applied optimization algorithms, such as GWO-SVM [27], PSO-SVM [24], and HHO-SVM, in comparison with the proposed IHHO-SVM model. The test uses the labeled dataset 5, which contains 2000 samples with an outlier content of about 4%, and the results are given in Table 3.…”

“…In the HHO algorithm, the population tends to converge towards the region of the current best solution, which reduces the diversity of the population and can lead to the problem of the algorithm getting trapped in local optima. In order to overcome this problem, we consider the characteristics of Gaussian and Cauchy distributions, set an adaptive weight coefficient by increasing the number of iterations, and use this strategy to perturb the optimal individual that remains unchanged for two successive iterations, as shown in Equations ( 23) and (24). If the perturbed position of the individual improves on the current best position, it will be incorporated in the next iteration.…”

“…When identifying wind power outliers using SVM, careful selection of SVM model appropriate hyper-parameters plays a crucial role in achieving optimal outlier classification performance of outliers [22,23]. In the field of SVM hyper-parameters optimization research, scholars have adopted various intelligent search algorithms, including particle swarm optimization (PSO), grey wolf optimization (GWO), and Harris hawks optimization (HHO) [24][25][26][27][28][29]. In [30], the advantages and disadvantages of the HHO algorithm, application fields, and improved variants are outlined.…”

Dissolved Gas Analysis and Application of Artificial Intelligence Technique for Fault Diagnosis in Power Transformers: A South African Case Study

“…SVM is a vigorous supervised learning technique for developing a dataset classifier. SVM purports to establish a decision boundary among two classes of datasets that facilitate the forecasting of data labels from one or several feature vectors [11][12][13][14][15][16][17]. A decision boundary can be described as the area of a problem space whereupon the output dataset label of a classifier is equivocal.…”

Multivariate-coupling LOCA prediction using zLSTM

“…Methods of this class assume linear dependences among all the system parameters, leading to quadratic increase of the model complexity and over-fitting problem. The second class uses traditional machine learning methods, such as Support Vector Machine (SVM) for break size estimation (Yu et al, 2022;Liu et al, 2021) and Group Method of Data Handling (GMDH) for the break location and break size estimation (Radaideh and Kozlowski, 2020). Artificial Intelligence (AI) methods groups the third class, in which Deep Neural Network (DNN) is deployed to predict the reactor core water level (Koo et al, 2018) and Convolutional Neural Networks (CNN) is widely used for multivariate time series prediction (Kollias and Zafeiriou, 2020) and operation parameter prediction of LOCA (Fukun et al, 2022).…”