基于RBF、GRNN和GA-BP神经网络的电动车空调能耗预测模型

Prediction Model of Air Conditioning Energy Consumption of Electric Vehicle Based on RBF, GRNN and GA-BP Neural Network

摘要: 暖通空调系统的能耗值对缓解新能源汽车“里程焦虑”和观测其性能具有重要意义。将环境温度、空气湿度、风速和车速作为输入参数，耗电量作为输出参数，建立径向基函数(Radial Basis Function, RBF)、广义回归神经网络(General Regression Neural Network, GRNN)和经遗传算法优化的反向传播(Genetic Algorithm-Backpropagation, GA-BP)神经网络模型。利用Amesim搭建某款电动车空调系统仿真得到不同工况下的能耗数据对三种模型进行训练和预测，将预测值和实际值比较，以验证其预测性能。结果表明：三种模型均能较好地预测不同工况下空调能耗，采用RBF、GRNN和GA-BP神经网络训练数据，测试数据线性回归系数R²分别为0.93641、0.95521和0.99517；预测结果相对误差分别为5%、3%和2%；均方误差分别为98.29 W/h、90.42 W/h和27.24 W/h。相比之下，GA-BP神经网络模型能更准确地预测空调能耗，可用于驾驶员缓解因空调带来的“里程焦虑”和实时观测空调性能。

Abstract: The energy consumption value of HVAC system is of great significance to alleviate the “range anxiety” of new energy vehicles and to observe their performance. The ambient temperature, air humidity, wind speed and speed are taken as input parameters, and the power consumption is taken as output parameters, RBF (Radial Basis Function), GRNN (General Regression Neural Network,) and GA-BP (Genetic Algorithm-Back propagation) neural network models. Amesim is used to build an electric vehicle air conditioning system to simulate the energy consumption data under different working conditions, and the three models are trained and predicted, and the predicted value is compared with the actual value to verify the predicted performance. The results show that the three models can well predict air conditioning energy consumption under different working conditions. RBF, GRNN and GA-BP neural network training data are used, and the linear regression coefficients R² of the test data are 0.93641, 0.95521 and 0.99517, respectively. The relative errors of the predicted results were 5%, 3% and 2% respectively. The mean square errors were 98.29 W/h, 90.42 W/h and 27.24 W/h, respectively. In contrast, GA-BP neural network model can predict the energy consumption of air conditioning more accurately, which can be used to alleviate the “range anxiety” caused by air conditioning and observe the performance of air conditioning in real time.