| Chinese per capita resources are relatively scarce.Saving resources,optimizing production scheduling,market security,and upgrading technology are all companies must do.Thermal power generation projects play an important role in the adjustment of Chinese energy structure.The basic principles of thermal power generation are:thermal power plants use coal as fuel,coal is burned in the boiler,and the water in the boiler is heated to generate steam.The steam of a certain temperature and pressure enters the steam turbine through the main steam valve and the regulating steam valve.Specifically,the internal energy of the steam is used to push the fan blades,and the generator rotates,which converts its heat energy into the mechanical energy that drives the rotation of the steam turbine rotor.The shaft drives the generator to generate electricity.In this series of energy conversion processes,the combustion efficiency of the boiler is the core of the power generation efficiency.Hot water evaporates in the fuel-added boiler to generate high-temperature and high-pressure steam.In the field of thermal power generation,the thermal load refers to the hot water.The internal energy of the high-temperature and high-pressure water vapor released per unit time in the fuel-added boiler.There are many factors that affect the steam heat load of the boiler.From the point of view of the adjustable parameters of the production of the boiler,there are combustion feed,primary and secondary air,water supply,etc from the working conditions of the boiler itself,there are furnace temperature and soot blowing Performance,air preheater performance,etc.Since coal and water are both scarce resources,for the steam production status of power plant boilers,short-and medium-term steam heat load forecasting has extremely important research significance for saving resources,optimizing production scheduling,market safety,and upgrading technology.Machine learning technology is not only the core of artificial intelligence,but also the foundation of computer intelligence.Since the 21st century,machine learning technology has gradually become the mainstream development direction of computer science and artificial intelligence and other information industries.Aiming at the problem of steam heat load prediction in the boiler room,this paper introduces three types of model representatives in the field of machine learning.The first one is the representative KNN model of "lazy learner",and the second one is the traditional linear model and variants such as Ridge,Lasso and Elesticnet elastic network regression,the third is an integrated learning model,including a random forest model based on bagging,Adaboost,GBDT and XGBoost based on boosting,a multi-regression fusion model based on stacking,and the thermal load data of A thermal power plant Modeling prediction and error comparison analysis.The results show that,for the problem of heat load prediction,the integrated learning model,especially the multiple regression model based on the stacking idea and the fusion of eight basic models,has achieved very good prediction accuracy,reflecting the innovation of practice. |
