Research On Prediction Method Of Steam Volume Of Thermal Power Generation Based On Machine Learning

Thermal power generation is the main way of power generation in China and the pillar of the national economy.Therefore,reducing energy consumption and improving the combustion efficiency of boilers are the main problems we are facing.The technological process of thermal power generation is that steam generated by heating raw materials causes the turbine to rotate,thus generating electricity.Combustion efficiency is the main factor affecting power generation efficiency in the above process.There are many factors affecting the combustion efficiency,including the variable parameters of the equipment,such as the raw materials that produce steam,the air supply of combustion,and the operating conditions of the equipment,such as the control of boiler temperature,pressure and flow rate.At present,for boiler combustion optimization in domestic electric power enterprises,the manual adjustment of fuel and air distribution is usually carried out by relying on the theoretical basis and operation experience of single or multiple variable combinations,which is difficult to achieve the ideal operation effect and leads to the waste of energy.Because the combustion efficiency of boilers is a process involving multi-variables,non-linearity and high complexity,it is difficult to find the most reasonable process parameters according to some theoretical basis and experience,which results in a lot of time in the process of verification and optimization of process parameters.With the development of artificial intelligence technology,we can use machine learning method to analyze and study the historical combustion data of boilers,so as to solve the problem of combustion efficiency of boilers.In this paper,the data set of industrial steam forecasting provided by a A power plant is taken as the research object,and a steam forecasting model is established by using machine learning method,using multiple linear regression,(Support Vector Regression,SVR),tree model(Extreme Gradient Boosting,XGBoost)and(A Highly Efficient Gradient Boosting Decision Tree,Light GBM)and improved model fusion method.The model is used to simulate and validate the operation process of the boiler,which reduces the engineering change of theresearch and development process.Change the quantity;improve the overall process level based on the parameters of the prediction model;realize the evaluation of energy consumption through the prediction results.Starting from steam volume prediction and improving combustion efficiency,this paper explains in detail the theory of machine learning,model selection,model optimization,model evaluation and model fusion,and uses these methods to model steam volume prediction.In the specific experiment,this paper uses Jupter noterbooks as the experimental environment and Python as the research tool.Firstly,the noise and abnormal values in the data are processed,the samples are normalized and the features are standardized,and then the features are transformed and constructed.Finally,multiple linear regression,SVR,XGBoost,Light GBM and improved model fusion are used to optimize and predict,and mean square error(MSE)is used as the index of model performance verification.Based on the model fusion algorithm,we propose a method of adding meta-model,and the effectiveness of the algorithm is verified by experiments.