论文部分内容阅读
在热态试验数据的基础上,分别应用BP(神经网络)和SVM(支持向量机)回归算法建立了燃煤机锅炉NOx排放特性模型,并验证了模型的准确性。结果表明,BP网络模型对检验样本的最大预测误差、最小预测误差和均方差分别为4.263%、0.556%和2.2133%,支持向量机模型对检验样本的最大预测误差、最小预测误差和均方差分别为2.121%、0.091%和0.4549%。两种智能技术都能对锅炉在不同工况下的NOx排放做出较为准确的预报,但支持向量机在泛化能力、收敛速度、最优性等方面明显优于神经网络。
Based on the thermal test data, a BP (neural network) and SVM (support vector machine) regression algorithm were respectively used to establish a model of NOx emission characteristics of coal-fired boiler and verify the accuracy of the model. The results show that the maximum prediction error, the minimum prediction error and the mean square error of the BP network model for the test samples are 4.263%, 0.556% and 2.2133% respectively. The maximum prediction error, the minimum prediction error and the mean square error of the support vector machine model 2.121%, 0.091% and 0.4549%. Both of them can predict the NOx emissions of boiler under different working conditions accurately, but SVM is superior to neural network in terms of generalization ability, convergence speed and optimality.