Cascade enhanced node gas consumption prediction method based on width learning system

Abstract

The invention discloses a cascade enhanced node gas consumption prediction method based on a width learning system. Belongs to the technical field of intelligent city construction. According to the method, data enhancement is carried out on original data by using curve smoothing, dimension expansion is carried out on the data by using phase-space reconstruction (PSR), data features are extracted, gas consumption data are predicted by using a width learning model of cascaded enhancement nodes, and finally, a predicted value is corrected through iteration by using an adaptive weight method, so that the gas consumption is predicted. And the influence of noisy points on prediction is reduced. Simulation results show that in the gas consumption prediction process, the influence of outliers and noise points on the prediction model is fully considered, and the accuracy and robustness of gas consumption prediction are remarkably improved. According to the method, neglect on the influence of noise points in the field of gas consumption prediction is made up, and the stability of gas consumption prediction is improved.

Description

technical field [0001] The invention relates to a method capable of accurately predicting gas consumption, and belongs to the technical field of smart cities. Background technique [0002] Natural gas, as a kind of clean green energy, is widely popularized in the life of urban citizens and used in many industrial fields because of its economic benefits and high safety. With the development of urban gas pipeline networks, gas load forecasting has become an important indicator for gas companies to carry out project planning and pipeline network management. However, affected by factors such as weather and temperature, the load value of gas is difficult to predict. Therefore, how to choose a load forecasting method to improve the forecasting accuracy is the key to gas load forecasting. [0003] Researchers at home and abroad have proposed many research methods on time series forecasting. Traditional time series forecasting models are mainly based on statistical methods, such a...

AI Technical Summary

Problems solved by technology

Therefore, traditional batch learning-based time series forecasting techniques will no longer be suitable for real-time forecasting scenarios that require rapid response as data is added

At the same time, considering the long running time and difficulty of convergence of the traditional deep structure, this paper aims to seek a simple and efficient machine learning technology that can use incremental data for new data on the basis of reducing model training time and saving costs. The way to update the model and improve the real-time prediction accuracy of the system

Images