A gas flow data prediction method based on non-equal-length granularity characteristics

Abstract

The invention belongs to the technical field of energy data prediction, and provides a gas flow data prediction method based on non-equal-length granularity characteristics. The method comprises the steps of firstly collecting the data based on a gas system, conducting the unequal-length fuzzy granulation on the data based on discrete change poles, showing the granulation form to comprise representative values and support upper and lower boundaries, wherein the particles represent vectors of the number of original data; secondly performing multi-dimensional density clustering based on weight change on the granulated data, and replacing granularity data in a time sequence with cluster division; thirdly designing a data prediction method based on the artificial neural network, performing model prediction by applying a time sequence result divided by a cluster, obtaining a corresponding weight value and a threshold value matrix, checking the accuracy of the weight value and the thresholdvalue matrix, and preparing for the following particle size reduction; and finally providing a particle size reduction method based on one-dimensional interpolation.

Description

technical field [0001] The invention belongs to the field of energy data prediction, specifically relates to granularity analysis, density clustering, model prediction and granularity reduction, and is a method for predicting gas flow data based on non-equal length granularity features. The invention uses the system to collect gas flow data, and designs a gas flow data prediction method based on granular data, variable weight multidimensional density clustering and artificial neural network prediction model. This method extracts and summarizes the characteristics and changes of the data by non-equal-length granulation of the gas flow data, and performs feature summary through the multi-dimensional density clustering method based on variable weights, replaces the granularity features with clustering features, and uses artificial neural The relevant knowledge of the network is used to model and predict the replaced data. Improve the prediction accuracy of data by adjusting rela...

AI Technical Summary

Problems solved by technology

[0004] However, these methods all have certain deficiencies: First, the gas data input information is complex and large-scale, which may be mixed with some redundant points, which is not conducive to the training of the prediction model; second, the model can achieve a higher level for short-term prediction. Accuracy, as the prediction time increases, the prediction error is larger; the present invention improves on the basis of the neural network, first, replaces multiple data points with granularity, which is conducive to reducing the error caused by data floating, reducing the training set and The amount of data in the test set, the obtained granular features can also describe the data more accurately

Images