基于改进烟花算法优化BP网络的混凝投药预测  

Prediction of coagulant dosage in waterworks based on improved FWA-BP neural network

作　　者：李赞 张长胜[1] 田海湧 毛辉 王卓 马泽楠 LI Zan;ZHANG Chang-sheng;TIAN Hai-yong;MAO Hui;WANG Zhuo;MA Ze-nan(Faculty of Information Engineering and Automation,Kunming University of Science and Technology,Kunming 650500,China;Yunnan Shuye Technology Co.,Ltd,Kunming 650221,China;Kunming Branch of North China Municipal Engineering Design and Research Institute Co.,Ltd,Kunming 650051,China)

机构地区：[1]昆明理工大学信息工程与自动化学院,云南昆明650500 [2]云南树业科技有限公司,云南昆明650221 [3]中国市政工程华北设计研究总院有限公司昆明分公司,云南昆明650051

出　　处：《陕西理工大学学报（自然科学版）》2021年第6期24-31,共8页Journal of Shaanxi University of Technology:Natural Science Edition

基　　金：国家自然科学基金资助项目(51665025,61963022)。

摘　　要：针对自来水厂混凝投药量控制滞后、非线性、多干扰等问题,提出一种改进烟花算法(IFWA)优化BP神经网络权值阈值的自来水厂混凝投药量预测模型。通过改进爆炸算子和精英选择策略,提高烟花算法(FWA)的搜索能力和寻优精度,利用布谷鸟搜索算法最优解作为FWA算法的初始解,提高搜索效率,并结合高斯变异操作,跳出局部最优。同时,通过3个函数测试,证明了IFWA算法的有效性。最后以云南某自来水厂的数据训练和测试模型,仿真结果表明,基于IFWA算法优化BP神经网络预测模型的均方根误差约为0.0975,平均绝对误差约为0.0762,在预测精度和收敛速度上优于FWA、CS等算法,具有可行性。Aiming at the problems of large lag,non-linearity,and multi-interference in the coagulation dosage control of a water plant,an improved fireworks algorithm(IFWA)was proposed to optimize the BP neural network weight and threshold values of the water plant coagulation dosage prediction model.By improving the explosion operator and the elite selection strategy,the search ability and optimization accuracy of the fireworks algorithm(FWA)were improved,and the optimal solution of the cuckoo search(CS)algorithm was used as the initial solution of the FWA algorithm to improve the search efficiency,combined with Gaussian mutation operation,jumped out of the local optimum.At the same time,the effectiveness of the IFWA algorithm was proved through 3 function tests.Finally,the model was trained and tested with data from a water plant in Yunnan.The simulation results showed that the RMSE of the BP neural network prediction model optimized based on the IFWA algorithm is about 0.0975 and the MAE is about 0.0762,which is better than FWA,CS,etc.in terms of prediction accuracy and convergence speed.It proves the algorithm is feasible.

关 键 词：投药量预测 BP神经网络 权值阈值优化 改进爆炸算子 精英选择策略 布谷鸟搜索算法 

分 类 号：TU991.22[建筑科学—市政工程] TP183[自动化与计算机技术—控制理论与控制工程]