Research On Optimal Strategy Of Time Series Data Forecast Based On Correlation Analysis

Time series data is data collected at various time points in a fixed time period.Through forecasting time series data sets,it is of great significance for studying the historical trajectory of things and for describing the future development trend and dynamic planning of things.Provide a basis for better control and accurate decision-making.The time series data has a long dependence,Therefore,a reasonable and effective prediction network model selection will have a great impact on the prediction results.In addition,time series data is easily interfered by sequence noise during the acquisition process,and it is necessary to process sequence noise and compensate for prediction errors during the prediction process.For this reason,this article mainly focuses on constructing time series data prediction network model,filtering optimization of sequence noise interference and designing prediction error optimization strategy.The specific research content is as follows:(1)Use the optimized Long Short-Term Memory(LSTM)network to perform multi-step prediction of power generation time series data.First,use the K nearest neighbor sliding window to fill in missing data values to ensure data integrity;Then,use the rescaled range method to analyze the correlation of time series variables and the gray correlation method to calculate the correlation between each influencing variable and power generation;Finally,use the optimized LSTM network to predict the power generation sequence data and compare it with several common prediction networks.The analysis and prediction simulation results show that the optimized LSTM network model has a better effect on time series data prediction.(2)Time series denoising filtering method based on linear and nonlinear fusion.Aiming at the problem of being easily interfered by sequence noise in the process of obtaining relevant time series data,by fusing linear and nonlinear Kalman filters,Convert the original nonlinear system filtering problem into a multi-model filtering optimization estimation problem,calculate the weight of the fusion filter,The fused state estimation value is iterated to optimize the sequence noise,and compared with the common EKF and UKF to obtain a better filtering optimization effect.(3)Design a forecast error optimization strategy for time series data.First,construct a general prediction error model and analyze the correlation structure function between prediction errors;Then,use commitment to express the difference between the backsight level control strategy and the average fixed level control strategy,and design a new online algorithm to optimize the strategy commitment level control(CHC)to optimize the analysis of the prediction error;Finally,several common noises are expressed in the form of related structure functions,and the relationship between the best commitment level and the competition difference obtained through the simulation graph is analyzed to analyze the performance of the CHC strategy to optimize the prediction error.