Research on speculation method for compressor map of PG9351FA gas turbine

• Published in: Proceedings of the Institution of Mechanical Engineers. Part A, Journal of power and energy Journal of power and energy, 2024-09, Vol.238 (6), p.954-968
• Main Authors: Hao, Xuedi, Zhang, Zeyuan, Chi, Jinling, Sun, Lei, Zhang, Jiajin
• Format: Article
• Language: English
• Subjects: Back propagation networks; Data points; Error correction; Gas turbines; Mean square values; Methods; Neural networks; Parameter estimation
• ISSN: 0957-6509; 2041-2967
• DOI: 10.1177/09576509241251909

A compressor map is usually represented by a limited number of feature points to speculate the entire operating range. Also, accurate compressor map models can be obtained quickly by using the appropriate methods. In this paper, 9351FA gas turbine is used as the research object, and a set of targeted compressor map speculation scheme is proposed. At 15 data points, high-precision compressor maps are obtained based on BP neural network, and this method is suitable for a large number of data points. At 6 data points, compressor maps are obtained based on the parameter estimation method, and this method is suitable for a small number of data points. The mean square deviation of the compressor map obtained by the neural network is about 0.002, while the minimum mean square deviation of the results of the parameter estimation method is 0.026 and the maximum mean square deviation is 0.088. Since the corrected speed line of 106.4 is almost vertical, the maximum error mean squared deviation and the maximum standard deviation occur on this line. Both methods are suitable for different sample sizes, and the speculated compressor maps are more reliable. The combination of the two methods can provide a set of reference methods for compressor map speculation.