Thermal Load Forecasting And Control Optimization Of Power Plant Based On Machine Learning

With the development of the goal of building a conservation-oriented society in China,central heating has gradually become the main heating method in northern winter.The central heating system has a complex structure and features such as non-linearity,large lag,large inertia,time variability,and uncertainty.If you want to achieve on-demand heating of the heating network,increase energy utilization,optimize the control strategy of the pipe network,and accurately heat Load forecasting is very necessary.The machine learning method has a strong fitting ability,can fully reflect the nonlinear characteristics of the heat load,not only provide reliable heat load data for the control system,but also can carry out timely and effective heat load adjustment.Based on the characteristics of machine learning,combined with the characteristics of actual heat load data samples(large sample,scattered data distribution,and fuzzy sample function),this paper further explores the heat load prediction method and optimal control of heat exchange stations.The heat load forecast is the premise and basis for the heating system to meet the heat demand of the heat users.First,based on the obtained data,a careful analysis is performed to predict the change in heating load in the next day.Due to the limitation of the currently obtained heat load heating data,this paper uses an advanced algorithm based on machine learning to analyze the characteristics and prediction methods of the heat load data samples.The flow variables of the heating network supply(return)water,the heating network supply(return)jellyfish tube pressure,and the heating network supply(return)jellyfish tube temperature are the input variables,and the heat load is the output variable.The modeling idea of this article is to preprocess a large amount of heat network historical data collected and convert it into data form that can be analyzed by Python.Scikit-Learn and Tensor Flow perform neural network construction and model learning.Although the sample data information obtained during modeling is limited,the selected input variables are sufficient to cover the key influencing factors of the heat load,so the prediction result is valid.The ultimate goal of central heating is to achieve on-demand heating,automatic control system is an important means to achieve this goal.A control scheme to control the water supply temperature of the secondary network by controlling the opening degree of the electric regulation valve on the primary network side is proposed to adapt to the dynamic characteristics of the pipeline network and realize the real-time control of the heating system.The fuzzy PID control model was constructed to control the water supply temperature of the heating network to regulate the heat load,and the SIMULINK software package was used to simulate the model.The simulation results showed that the fuzzy PID control system had better stability and improved heating quality and energy consumption.To achieve the purpose of heating on demand.