|
|
|
| Identification method of abnormal state of hydropower unit based on PCA-GA-BP neural network |
| LUO Zhengliang1, PAN Hong1*, ZHAO Lei2, TANG Wei3, ZHENG Yuan3 |
| 1. School of Energy and Electrical Engineering, Hohai University, Nanjing, Jiangsu 210037, China; 2. Chongqing Shipping Construction Development Co. Ltd., Chongqing 400000, China; 3. School of Water Conservancy and Hydropower, Hohai University, Nanjing, Jiangsu 210037, China |
|
|
|
|
Abstract In order to improve the efficiency and accuracy of the unit abnormal operation identification, a unit abnormal state detection model based on BP neural network optimized by principal component analysis and genetic algorithm was proposed. Taking the real-time sensor data recorded by the unit as a sample, the principal component analysis method was used to reduce the dimension of multidimensional data, and the comprehensive variables were obtained after processing. On this basis, the BP neural network is built and the genetic algorithm was used to optimize the random initial weights and thresholds of the neural network, so as to complete the simulation training of diffe-rent operating states detection models of the units. Finally, taking the real-time monitoring data of each component sensor in the normal and abnormal operating conditions of a power station unit in a continuous period of time as a sample analysis. The proposed PCA-GA-BP algorithm was compared with other optimization algorithms and traditional algorithms, and the feasibility of this method was verified by simulation training experiments with different sample ratios. The simulation results show that: compared with the traditional BP neural network, the model has a relatively shorter average state detection time of 84% and the average detection accuracy is relatively higher by 2.5%. It can achieve abnormal operation of the unit within 0.7-1.0 s on the basis of an average accuracy rate close to 99%. Accurate identification and early warning of abnormal operation of the unit shall be achieved.
|
|
Received: 03 July 2020
|
|
|
|
|
|
|
