Research On Key Technologies Of Stream Computing Based On Power Data Transmission Link

With the development of emerging information technologies such as cloud computing,big data and the Internet of Things,State Grid Corporation of China has conducted a lot of work on the intelligent construction of power grid.Especially,a large number of measurement components are deployed at power grid terminals to ensure the stable operation of the power grid system through real-time monitoring of equipment data,resulting in a massive amount of multi-source heterogeneous data.These data are real-time data generated by system operation,which hides important information related to equipment failures and system stability.It is necessary to perform real-time calculation and timely feedback to the control system,hoping to obtain its due value within the effective time of the data.Currently,big data processing mainly includes batch processing and stream processing.Batch processing has been widely studied,while stream processing is gradually becoming a focus of attention in the industry due to its advantages of real-time feedback.With the increasing demand for stream computing applications in the market,building a low-latency,high-throughput and continuously reliable stream computing system is one of the current research focuses.Therefore,it is of great significance to further resesrch the stream computing technology in power systems.To solve the problem that the stream data in the current power grid system cannot achieve quasi-real-time calculation,a real-time stream processing system based on the power data transmission link was designed in this work to achieve real-time analysis of data collection,data calculation,data storage and data application service displays,meeting the needs of fault alarms and power consumption data analysis in the power data transmission link process.The data acquisition system monitored the changes of the database in real time by Flink-CDC and transmitted the data to the Kafka message queue.After preprocessing,the data was transmitted to the real-time stream processing framework for real-time calculation and hierarchical storage.Finally,the data was displayed in real time and the alarm information and the operating monitoring status of equipment were output through the data application service module.Combining the GRU neural network model,a real-time wind power coordination prediction algorithm based on RF-BO-GRU was proposed to analyze and extract the hidden relationship between historical wind power and wind measurement data,and achieve real-time prediction model of wind turbine output power.Firstly,the offline wind power prediction model was trained,and then the model was encapsulated in the data application service layer of the stream data processing framework,so as to achieve collaborative real-time prediction of wind power,enhance the operation and maintenance capabilities of wind power plants and optimize power system dispatch control.Flink-CEP complex event extraction model was integrated into the data application service layer,and CEP rules library was used to analyze and extract data calculation results to achieve real-time diagnosis and alarm of power equipment failures.