图共 5个 表共 5
• 图  1  BP神经网络训练误差

Figure 1.  Training error of BP neural network

• 图  2  压缩机第四段压比

Figure 2.  Pressure ratio of the fourth section of the compressor

• 图  3  压缩机第四段多变效率

Figure 3.  Polytropic efficiency of the fourth section of the compressor

• 图  4  裂解气四级压缩系统

Figure 4.  Cracking gas four-stage compression system

• 图  5  压缩机功耗、第四段多变效率与出口温度的关系

Figure 5.  Relationship among power consumption, the fourth polytropic efficiency and the outlet temperature of compressor

•  Item Iterations times Standard deviation BP algorithm 198 0.164 4 LR-GA-BP algorithm 20 0.011 5

表 1  迭代次数与标准差

Table 1.  Iterations times and standard deviation

•  Sample Expected value BP output value LR-GA-BP output value 0,0 0 0.164 3 0.015 2 0,1 1 0.947 7 0.999 6 1,0 1 0.985 9 0.999 2 1,1 0 0.225 8 0.013 2

表 2  两种算法学习结果比较

Table 2.  Comparison of learning results between two algorithms

•  Component Predicted value/% Actual value/% Relative error/% H2 7.061 6.973 1.26 CH4 32.462 32.503 -0.13 C2H4 40.559 40.476 0.21 C2H6 5.611 5.706 -1.66 C3H6 12.256 12.320 -0.52 C3H8 1.057 1.066 -0.84

表 4  压缩机第四段出口气体主要组分预测值与实测值比较

Table 4.  Comparison between predicted values and actual values of main components in cracking gas from the fourth-stage outlet of the compressor

•  Compressor Outlet temperature/℃ Relative error/% Outlet pressure/MPa Relative error/% Predicted value Actual value Predicted value Actual value C100-1ST 82.5 82.5 0 0.271 0.269 0.74 C100-2ND 82.9 83.4 −0.60 0.493 0.494 −0.20 C100-3RD 84.3 84.2 0.12 0.947 0.951 −0.42 C100-4TH 101.1 101.3 −0.20 2.415 2.411 0.17

表 5  压缩机各段出口温度/压力预测值与实测值比较

Table 5.  Comparison between the predicted temperatures /pressures and actual values of each stage outlet of compressor