Application of an Improved Ridgelet Process Neural Network for Predicting the Temperature Rise of Rotor Structure Optimization

Abstract

To solve the negative-sequence temperature-rise problem of large equipment under asymmetric operating conditions, this paper optimizes the structure of the main components and adopts an improved process neural network to conduct online analysis and calculate the operating data, achieving the accurate prediction of the equipment heating status. Firstly, taking a 300 MW generator that urgently needs equipment improvement as the research object, the typical asymmetric accident characteristics that have occurred in recent years and the main influencing factors of negative-sequence heating of the rotor are analyzed. The influence of the rotor damping structure and shaft length on the temperature-rise change is explored. Secondly, a tent map is introduced to enhance the distribution uniformity of the population in the search space to enhance the global convergence of niche genetic algorithms. Numerical experiments and field experiments show that the improved algorithm, which is applied to optimize the parameters of the ridgelet process neural network, has good temperature-rise prediction performance. Finally, the influence of the rotor length and number of pole damping bars on the negative-sequence heating problem under different negative-sequence component ratios is examined, which provides useful references for the structural optimization and asymmetric operation state prediction of large equipment.