A Multivariate Distorted Time Series Forecasting Method

Abstract

The invention discloses a method for predicting multivariate distorted time series. The prediction method includes the following steps: 1. Establishing a training sample set; 2. Constructing a multivariate time series convolutional neural network model; the multivariable convolutional neural network model is at least Including sequentially connected input layer, feature extraction layer, convolution layer module, fully connected layer of each variable, fully connected layer of all variables and output layer; 3 training multivariate time series convolutional neural network model; 4 assembling multivariate time series volume Using the product neural network model, the multivariate time series forecasting system is obtained; 5. Using the multivariate time series forecasting system to forecast the multivariate time series of power consumption. This method is used to predict the power consumption of the power grid system. It overcomes the shortcomings of traditional methods that do not make full use of sequence abstract features and is easily affected by data distortion. It can reduce the impact of distorted data on the accuracy of prediction results and has strong reliability.

Description

technical field [0001] The invention relates to the technical field of multivariate time series forecasting, in particular to a multivariate distorted time series forecasting method. Background technique [0002] With the development of industry and the construction of cities, the power supply capacity of urban power grids has become closely related to the quality of life of residents. Time series analysis is one of the effective ways to plan power grid power supply. Time series is a series of data point values ​​arranged in chronological order. Time series analysis is based on these ordered observation data, using mathematical statistics and other methods to study its statistical laws, so as to predict future data and solve practical problems. [0003] Existing time series forecasting methods do not make full use of all the abstract features of the available sequences, and often use univariate time series forecasting, that is, only use the abstract features of the forecast ...

AI Technical Summary

Problems solved by technology

But in fact, things are interrelated, that is to say, there are correlations and co-occurrences in the time series of different variables in the same application field, and there are limitations of univariate time series characteristics, and the obtained information is not comprehensive and accurate enough. The result is not precise enough

In addition, the inability to overcome the distortion of time series data is another defect of existing methods. The translation or disorder of time segments on the time axis is a challenge for the analysis and prediction of time series.

Images