Research On Actuation Time Forecasting Of Circuit Breakers Based On Big Data Technology

With the improvement of smart grid construction level,the UHV circuit breaker is required to be intelligent and has higher stability.The reliability of the UHV circuit breaker is closely related to the safety of the smart grid.Selective phase switching technology is an important direction of circuit breaker intelligence,which can improve the economic and reliability of power system operation by reducing the inrush current and overvoltage amplitude in the transient process.However,the actuation time of the UHV circuit breaker is decentralized and difficult to predict and control,which make the phase selection switching technology hard to achieve.Therefore,increasing the stability of the actuation time of the UHV circuit breaker has been given a lot of attention.This study is devoted to predicting the actuation time of UHV circuit breakers,analyzing the factors affecting the actuation time based on big data technology and studying the influence law of influencing factors and actuation time.Firstly,the spring hydraulic mechanism of UHV circuit breaker and the application status of big data technology in power grid are analyzed.Then,the technical parameters,structural parameters and working principle of the spring hydraulic mechanism are explained.Based on the working principle of the mechanism,the influence of structural parameters and environmental factors on the actuation time of the circuit breaker is analyzed,which provides a theoretical basis for processing experimental data.An offline distributed processing platform based on Hadoop-based circuit breaker action time prediction is proposed and built.The experimental environment is set up after analyzing the experimental data.The hardware environment is selected to create three virtual machines and the software environment installs HBase and ZooKeeper software on the virtual machine respectively to form a fully distributed processing platform with three nodes.In addition,the stand-alone mode is set to verify the high efficiency of a fully distributed platform by contrasting the loss time of processing data between single-machine mode and fully distributed mode.Structural preprocessing of a large number of 1100 kV small GIS circuit breaker experimental data is made.Multivariate linear analysis and BP neural network algorithm are implemented by eclipse programming on the platform.The mathematical model of predicting action time of the circuit breaker is gained by controlling factors such as voltage pressure and oil pressure,which is verified and the error is analyzed.The results show that theerror range of the multivariate linear analysis method is less than ±1ms in the prediction of the opening time and less than ±1.5ms in the prediction of the closing time.The BP neural network has higher prediction accuracy and the prediction error ranges of the closing and closing time are less than ±1ms.Comparing the prediction errors of the two algorithms,the BP neural network prediction error range satisfies the phase selection requirements of UHV circuit breakers.Combined with the processing platform for circuit breaker action time prediction,LabVIEW is used to design and implement the visualization platform.According to the functions to be realized,the visualization platform is divided into four modules: serial communication,data reading,opening time prediction and closing time prediction.Programming and making interface design for each module,run the visualization platform.The results show that the platform is simple,intuitive and can effectively analyze and display the prediction results.