基于遗传算法优化BP神经网络的供暖系统热负荷预测模型

针对供暖系统热负荷短期预测问题,利用改进的遗传算法(genetic algorithm,GA)对BP神经网络(back propagation neural network)的初始权值和网络结构进行优化,并在遗传进化过程中采取保留最佳个体的方法。该方法克服了一般BP网络初始权值的随机性和网络结构训练过程中的所带来的网络震荡,以及一般BP网络容易陷入局部极小等问题。同时结合一般BP神经网络方法进行仿真实验和分析比较,结果表明:该方法具有全局寻优能力,预测精度高,绝对和相对误差较小,收敛速度快,能够有效针对供暖系统热负荷进行短期预测。

Thermal Load Forecasting Model of Heating System Based on Genetic Algorithm Optimization BP Neural Network

For short-term forecasting heating system heat load problem, use of improved genetic algorithm （GA） to initial weights and network structure to optimize BP（ back propagation） neural network, and re- tain the best individual approach taken in the process of evolution. The method overcomes the randomness and the network structure of BP network training process is generally the initial weights of the network caused by the shock, and general BP network is easy to fall into local minima problems. Combined with the general method of BP neural network simulation and analysis and comparison results show that the method has global search capability, high forecast accuracy, absolute and relative error is small, fast con- vergence, can effectively heat load for the heating system for short-term forecasting.