改进灰色神经网络的作动系统功率预测

为提高飞机上作动系统的功率预测精度,建立了改进的多变量灰色神经网络预测模型。考虑了对系统功率需求有较大影响的相关因素,采用主成分分析法提取综合变量作为输入,在提升准确性的基础上有效减少了输入维数;在利用递增方式对初始值进行选择的过程中,引入粒子群优化算法快速求解最优初始值和背景值,模型预测的平均误差由13.35%降为7.53%;考虑到序列波动对预测精度的影响,采用BP神经网络对预测值进行误差修正,进一步将模型的平均预测误差降为4.07%。仿真实验表明,含主成分分析的改进灰色神经网络对飞机作动系统的功率有较高的预测精度,有利于飞机的电能调度。

Power Prediction of Actuator System Based on Improved Grey Neural Network

The accurate power prediction of actuator plays an important role in electric energy management. In order to improve the prediction accuracy in actuator system, an improved multivariable grey neural network model is established. Taken this aspect of the relative factors of producing a great impact on the needs of power in system into account, the PCA is used to extract the comprehensive variables as an input of the prediction model. By so doing, this effectively reduces the dimension on the basis of the accuracy of prediction. In the process of selecting the initial values and background values by utilizing the incremental method, the PSO is used to calculate the optimal background value to improve the prediction accuracy to a large degree. In addition, taken this aspect of the influence matter on sequence fluctuation into account, a BP neural network model is established to correct the error in the prediction value. The results show that the model is valid and advantageous to the energy dispatching of aircraft.