Application Research On State Trend Prediction And System Integration Of Pumped Storage Units

The new national energy strategy points out that China needs to further increase the construction of pumped storage power stations.Pumped storage power station is an indispensable adjustment method in China's power system.It plays an important role in ensuring the safe and stable operation of power grids,such as peak shaving,frequency modulation and phase modulation,and accident reserve.At present,the pumped storage unit is developing in the direction of “high head,large inertia”.As the core equipment for energy efficiency conversion of power station,it has the characteristics of complicated operating conditions and frequent conversion,which leads to increasingly prominent security problems.In order to ensure the safe and stable operation of the equipment and other equipment,and improve the predictive maintenance capability of the power station,it is necessary to combine advanced information technology to improve the technical level of the unit equipment monitoring and realize real-time mastery of the equipment operating status.At the same time,it is necessary to carry out unit operation status trend analysis.Establish accurate and reasonable forecasting models,capture signs of equipment failure development,and provide decision analysis for power plant operation and maintenance managers.Based on the actual needs of the above projects,the main research contents and results of this paper are as follows:(1)In order to improve the real-time,reliability and cross-platform performance of the unit condition monitoring platform,and to fully understand the main content framework and data acquisition approach of the unit condition monitoring,a reliable and stable real-time data push and storage scheme of the B/S structure condition monitoring platform is designed according to the characteristics of the unit vibration,swing,pressure,flow and other monitoring data.The full-duplex communication protocol WebSocket is chosen as the communication protocol for real-time monitoring of the Web.A real-time data push module with multi-protocol combination and a database storage architecture with two-tier timeliness mechanism are designed.Redis,an efficient read-write memory database,is combined with Sql Server,a relational database,to efficiently realize data storage and provide data support for subsequent trend analysis.(2)In order to improve the ability of pre-decision analysis of power stations and timely capture relevant fault signs,the characteristics of nonlinear state quantities in the state monitoring of pumped storage units are complex and difficult to predict and analyze.A trend prediction model based on signal decomposition VMD and CNN-LSTM hybrid neural network is proposed.The input and output data sets are constructed by signal decomposition.The Adam optimization algorithm is used for model training.The model is applied to a unit trend analysis experiment.And compared with a single CNN and a single LSTM,the experimental results show that the prediction model has higher prediction accuracy and performs well in multi-step prediction results.(3)Focusing on the above-mentioned main research contents,combined with the current mainstream JavaWeb technology,designed and developed advanced application software systems with monitoring of historical data query,status assessment,trend prediction,file management and other major services to improve the operation and maintenance level of the power station.It is practical to monitor the health status of equipment.