Daily gas load combination prediction method based on generalized dynamic fuzzy neural network

Abstract

The invention provides a daily gas load combination prediction method based on a generalized dynamic fuzzy neural network, comprising the following steps: (1) acquiring historical urban gas record data which are used as historical time series data; (2) judging and processing abnormal data in the historical time series data; (3) carrying out differential processing on historical load time series, i.e. sample data, and predicting by adopting a generalized regressive neural network; (4) taking one-time accumulation generation data input into the historical load time series as the input of the network by using a gray neural network, outputting the one-time accumulation generation data corresponding to the predicted daily gas load, training the network, and finally carrying out one-time degressive inverse generation processing on the output value; and (5) taking predicted values obtained in the steps (3) and (4) as the input of the generalized dynamic fuzzy neural network, and grouping thedata. The daily gas load combination prediction method is adopted according to the characteristics of randomness, instability, periodicity and the like of the daily gas load, thereby having higher prediction precision.

Description

technical field [0001] The present invention relates to a combination forecasting method of daily gas load, specifically a new method for forecasting daily gas load by using multiple neural network models and finally performing combined forecasting of gas by generalized dynamic fuzzy neural network. Background technique [0002] The research on gas forecasting relies on the characteristics of natural gas such as periodicity, trend and randomness. There are also many factors that affect gas load, such as temperature, climatic conditions, holidays, residents' living standards, living habits, energy policies, and social and economic development levels, etc. These influences make it more challenging to predict accurately. [0003] Gas forecasting has developed to a certain extent in my country, but because natural gas has not been used for a long time as a new energy in my country, the data and experience available for forecasting are still insufficient. [0004] At present, tr...

AI Technical Summary

Problems solved by technology

[0005] However, these methods also have certain disadvantages and limitations

For example, the time series method has high requirements for data, cannot conveniently consider weather conditions and other related factors that have an important impact on load, only focuses on data fitting, and lacks regularity processing.

Gray prediction is more suitable for the prediction of load with exponential growth law, when the load growth rate is slow, that is, the corresponding exponential function x=be -at When |a| is small, the prediction accuracy is high, and when |a| is large, the prediction accuracy becomes worse

Images