基于HPSO-BP神经网络融合的隧道照明功率预测

针对当前高速公路隧道照明存在的过度照明问题,采用一种混合粒子群算法和BP神经网络相结合的联合优化算法实时预测所需LED功率。该算法优化了神经网络的初始权值和阈值,克服了种群易陷入局部极小的缺点,同时也加快了收敛速度,将优化好的神经网络用于预测LED功率减小了误差。具体措施是将由传感器实时采集的洞外亮度、车流量及其车速、能见度作为照明控制系统的输入量,经过计算所得的隧道照明的调光值作为照明控制系统的输出量,搭建一个4个输入1个输出的控制模型,用混合粒子群优化神经网络对监控数据进行训练分析,拟合输入输出的关系,最终实现实时预测调控的目的。以赣州市尖峰岭隧道的真实数据进行设计,Matlab仿真结果表明,基于HPSO-BP神经网络算法比传统BP神经网络算法在预测精度和收敛速度性能上表现得更加优秀,可以达到实时预测调控的要求,减少了照明能耗。

A joint optimization algorithm combining hybrid particle swarm optimization algorithm and BP neural network(HPSO-BP)is used to predict the required LED power in real time to deal with the excessiev illumination existing in the current highway tunnel lighting,by which the initial weight value and threshold value of neural network are optimized,the shortcoming that the population is easy to fall into the local minimum is overcome,and the convergence speed is accelerated. The optimized neural network is used to predict the LED power to reduce the error. The specific measures are that the luminance,traffic flow,speed and visibility outside the tunnel collected by sensors in real time are taken as the input of the lighting control system,the calculated dimming value of tunnel lighting is taken as the output of the lighting control system,a control model with four inputs and one output is constructed,the HPSO-BP is used to train and analyse the monitoring data,the fitting relationship between the input and output is produced,and the purposes of real-time prediction and control is realized. A simulation experiment was designed based on the real data of Ganzhou Jianfengling tunnel. The Matlab simulation results show that HPSO-BP neural network algorithm has better performance in prediction accuracy and convergence speed than traditional BP neural network algorithm,which can meet the requirements of real-time prediction and regulation,and reduce lighting energy consumption.