A method and device for predicting urban natural gas load in heating season

Abstract

The present application discloses a method and device for predicting urban natural gas load in heating season. The method obtains original data of urban natural gas load and performs feature extraction to obtain a feature vector, and inputs the feature vector into a preset cyclic neural network model to obtain a first predicted value , determine multiple temperature vector groups according to the daily historical temperature feature vector in the feature vector, determine the second predicted value according to the multiple temperature vector groups, and integrate the first predicted value and the second predicted value to obtain the prediction of urban natural gas load gas consumption value. Because the prediction is based on the actual natural gas load gas consumption data, the prediction process is combined with the actual business application scenario, and the first predicted value is obtained by predicting the preset cyclic neural network model that is sensitive to temperature changes. Determining the second predicted value can adapt to the change of temperature characteristics, achieve the purpose of accurately predicting the urban natural gas load value, and can provide effective guidance to the real production environment.

Description

technical field [0001] The present application relates to the technical field of natural gas load prediction, and in particular, to a method and device for predicting urban natural gas load in heating season. Background technique [0002] In the natural gas load forecasting of northern cities, the load forecasting during the heating season is particularly important, mainly because the surge in gas consumption caused by urban heating in winter has brought great challenges to the load forecasting. [0003] At present, there are many types of methods for urban natural gas load forecasting technology. Load forecasting methods include not only traditional forecasting methods such as multiple linear regression analysis, time series method, least squares method, and gray model, but also artificial intelligence-based support vector machines and neural networks. and other machine learning algorithms. However, these natural gas load forecasting methods only stay at the academic level...

AI Technical Summary

Problems solved by technology

However, these natural gas load forecasting methods are only analyzed at the academic level, and have not penetrated into the actual business scenario application. They cannot accurately predict the urban natural gas load value, resulting in the inability to effectively guide the real production environment.

Images