基于ANFIS的蓄电池剩余电量检测系统的研究与实现

发布时间：2018-05-21 13:44

本文选题：蓄电池 + SOC(state　；参考：《电子科技大学》2014年硕士论文

【摘要】：本文得到了来自中国航天科工集团“蓄电池剩余电量监控系统”项目的支持,主要研究和建立了基于神经网络的蓄电池剩余电量的预测模型,系统可应用于对蓄电池剩余电量有精确预测需求的设备中。本文建立了基于自适应神经网络模糊推理系统(ANFIS)的蓄电池剩余电量预测模型。首先介绍了ANFIS的相关理论,指出选用该算法的原因;接着分析了影响蓄电池SOC的主要因素,以此确定作为模型输入的相关蓄电池参数,再对各输入参数的论域进行了模糊集划分和隶属度函数计算;最后得出ANFIS模型结构,确定了网络学习算法。进行相关实验,依照实验数据对模型进行了训练及验证。首先介绍了实验所用设备及制定的实验方案,对实验数据进行初步分析;然后编写了MATLAB仿真程序并制定了具体仿真方案,用不同方法构造初始ANFIS模型,利用实验数据对模型网络进行训练,分析过程中ANFIS的结构和参数的变化,将模型值与实际测得的结果进行对比。仿真结果表明了该模型对几种ANFIS网络的预测都比较准确,采用减法聚类法产生的ANFIS网络最优—经训练后,其节点数相对最少。最后,对网络的各个参数进行了调整并再次用仿真比对预测效果。完成了硬件的设计和软件的开发。硬件设计方面,测量单元采用单片机集成了蓄电池内阻、电压、电流、温度的测量及通信等模块,其中的难点在于蓄电池内阻的测量,文中用交流(正弦波)注入法测量内阻。软件方面,系统由测量单元和显控单元组成,前者主要完成蓄电池数据的采集并传输给显控,后者主要运行ANFIS算法并将结果提供给用户。测量单元同时要求完成蓄电池各参数各测量及低功耗控制等工作;另外,显控单元包括用户UI的设计。最后,本部分还制定了基于MODBUS-RTU的通信协议的指令及应答规则。文章最后部分测试了系统的实际运行效果。测试结果显示:测量单元所测得数据误差在5%以内,ANFIS算法处理显控单元及测量单元数据得到的蓄电池SOC预测值与实测值误差最大为0.0046,该结果满足工程应用需求。然而,系统(尤其是显控单元)存在运行速度较慢且占用显控单元资源较多的问题,导致了显控系统的整体性能降低,鉴于此,提出了两种改进方法,并对前景进行了审慎的预测。
[Abstract]:This paper is supported by the project of "Monitoring and Control system of Battery residual quantity" of China Aerospace Science and Technology Group. The prediction model of battery residual quantity based on neural network is studied and established. The system can be used to accurately predict the demand for battery surplus. In this paper, a prediction model of battery residual quantity based on adaptive neural network fuzzy inference system (ANFIS) is established. This paper first introduces the relevant theory of ANFIS, points out the reason why the algorithm is selected, then analyzes the main factors that affect the storage battery SOC, and then determines the relevant battery parameters as the input of the model. Then the fuzzy set partition and membership function calculation of each input parameter are carried out. Finally, the structure of ANFIS model is obtained, and the network learning algorithm is determined. The model was trained and validated according to the experimental data. This paper first introduces the equipment used in the experiment and the experimental scheme, analyzes the experimental data, then writes the MATLAB simulation program and formulates the specific simulation scheme, and constructs the initial ANFIS model with different methods. The model network is trained with experimental data, and the changes of the structure and parameters of ANFIS are analyzed. The model values are compared with the measured results. The simulation results show that the model is accurate for several kinds of ANFIS networks, and the subtractive clustering method is used to generate the optimal ANFIS networks. Finally, the parameters of the network are adjusted and the simulation results are compared again. Completed the hardware design and software development. In the aspect of hardware design, the single chip microcomputer is used to integrate the internal resistance, voltage, current, temperature and communication module of the battery. The difficulty lies in the measurement of the internal resistance of the battery. The AC (sinusoidal wave) injection method is used to measure the internal resistance in this paper. In software, the system consists of measurement unit and display control unit, the former mainly completes the data acquisition and transmission to display and control, and the latter mainly runs ANFIS algorithm and provides the results to the user. In addition, the display and control unit includes the design of user UI. Finally, this part also formulates the instruction and reply rule of communication protocol based on MODBUS-RTU. In the last part of the paper, the actual running effect of the system is tested. The test results show that the error of the measured data is less than 5%. The maximum error between the predicted value and the measured value of battery SOC is 0.0046, which meets the requirement of engineering application. However, the system (especially the display and control unit) has the problems of slow running speed and occupying more resources of the display and control unit, which leads to the deterioration of the overall performance of the display and control system. In view of this, two improved methods are proposed. The prospect is forecasted prudently.
【学位授予单位】：电子科技大学
【学位级别】：硕士
【学位授予年份】：2014
【分类号】：TM912;TP274

【参考文献】