State trend prediction of hydropower units under different working conditions based on parameter adaptive support vector regression machine modeling

• Published in: JOURNAL OF POWER ELECTRONICS Vol. 23; no. 9; pp. 1422 - 1435
• Main Authors: Zhao, Guo, Li, Shulin, Zuo, Wanqing, Song, Haoran, Zhu, Heping, Hu, Wenjie
• Format: Journal Article
• Language: English
• Published: Singapore Springer Nature Singapore 01.09.2023; 전력전자학회
• Subjects: Electrical Machines and Networks; Engineering; Original Article; Power Electronics; 전기공학; Aquila optimizer; Hydropower unit; Parameter adaption; Trend prediction; Neural network; SVR
• ISSN: 1598-2092; 2093-4718
• DOI: 10.1007/s43236-023-00631-1

To address the problem where the different operating conditions of hydropower units have a large influence on the parameters of the trend prediction model of the operating condition indicators, a support vector regression machine prediction model based on parameter adaptation is proposed in this paper. First, the Aquila optimizer (AO) is improved, and a sine chaotic map is introduced to influence the population initialization process. An improved adaptive weight factor is used to balance the local search and global search capabilities. Second, according to the power and the head, the operating conditions of the unit are refined into several typical sets of operating conditions. On this basis, an SVR model is established using the improved AO search algorithm proposed in this paper, and the prediction parameters under each of the operating condition are optimized to establish the data of the operating conditions and optimal parameters. Then a neural network is used to fit the working condition and the optimal prediction parameters. In addition, the nonlinear function mapping of the complex relationship between the two is constructed. Finally, the constructed mapping relationship is added to the traditional SVR, and an adaptive SVR prediction model suitable for changes in the working conditions of hydropower units is realized. Simulation results show that when compared to the traditional SVR prediction model, the adaptive SVR prediction model designed in this paper can automatically adjust the prediction parameters according to changes in the working conditions and achieve the goal of maintaining optimal prediction performance under different working conditions. In addition, it has the ability to accurately predict the development trend of the unit operating state index within a certain time scale.