基于尖峰自组织模糊神经网络的需水量预测  

Prediction of water demand based on spiking self-organizing fuzzy neural network

作　　者：乔俊飞 张力 李文静 QIAO Jun-fei;ZHANG Li;LI Wen-ring(Faculty of Information Technology,Beijing University of Technology,Beijing 100124,China;Beijing Key Laboratory of Computational Intelligence and Intelligent System,Beijing 100124,China)

机构地区：[1]北京工业大学信息学部,北京100124 [2]计算智能与智能系统北京市重点实验室,北京100124

出　　处：《控制与决策》2018年第12期2197-2202,共6页Control and Decision

基　　金：国家自然科学基金项目(61533002;61603009);北京市科技专项课题领军人才项目(Z15110000013151010);北京工业大学日新人才计划项目(2017-RX(1)-04)

摘　　要：短期需水量预测是城市给水管网安全稳定运行的前提和保证.针对日需水量预测提出一种基于尖峰机制的自组织模糊神经网络(SSOFNN)模型.针对影响变量复杂多变的特点,采用主成分分析对原始数据进行降维处理,获取线性无关的主成分变量作为预测模型输入数据. SSOFNN模型根据尖峰强度和误差指标在训练过程中对隐含层神经元进行增长修剪,结合改进Leveberg-Marquardt算法简化参数更新过程中的计算过程,大大减少了计算量,能够获得紧凑的网络结构,且跟踪精度高,运行时间短,预测效果好.Short-term prediction of water demand provides basic guarantee for water supply system operation and management. In this study, an effective model for daily water demand forecasting is proposed. Firstly, principle component analysis(PCA) is utilized to simplify the complexity and reduce the correlation between influence variables, and the score values of selected principle components(PCs) turn into the irrelevant input data of fuzzy neural network(FNN), which models the prediction of water demand. Moreover, an improved Levenberg-Marquardt(ILM) algorithm is employed to optimize the parameters of FNN simultaneously, the problems of heavy computing burden and limited memory space can be solved. Most of all, a growing-pruning mechanism based on spiking integrate-and-fire(IF) model is applied to FNN in order to realize structural self-organization. Finally, contrast experiments are implemented to demonstrate that the spiking self-organizing fuzzy neural network(SSOFNN) has better prediction performance and capability to handle practical issues.

关 键 词：模糊神经网络 自组织 尖峰机制 需水量预测 给水管网 

分 类 号：TP183[自动化与计算机技术—控制理论与控制工程]