The Research Of Heating Boiler's Multi-mode Optimal Modeling Based On Machine Learning

Boiler is a complex large-scale energy conversion equipment in heating industry.It is always lacks reasonable guidance in operation,and has low energy conversion efficiency.It is instructive to establish an effective boiler operation model for improve efficiency.However,boiler heating system is a typical complex industrial process,which has multi-input and output,strong non-linear and coupling,no self-balance and large lag and so on.It's difficult to use the traditional mechanism analysis to get accurate mathematical model in the optimization process.Meanwhile,the boiler load changes with the demand of the outside world.Under the different load,the operating mode and status of the boiler is changed.It shows obvious characteristics of multiple operating conditions.How to mine the information of process variables under different operating conditions in high efficiency.To ensure that the boiler operating in different conditions of high efficiency is the key for the optimal operation of the boiler guideWith the widespread application of information technology in the boiler field,a large amount of on-site data is stored.The machine learning algorithm based on data-driven is gradually applied to the boiler field,for example,data classification,state identification and optimization modeling.In this paper,an method that is based on machine learning algorithm for multi-conditions of heating boilers,was proposed of optimization modeling.It completed the multi-boiler modeling and optimization of the multi-conditions guidance.The main contents of this paper include the following points.1.Division and identification of boiler operating conditions.By analyzing boiler historical data and using FCM algorithm to get the optimal cluster number C and cluster center V of boiler data.The boiler operation process is divided into different working conditions;Using the results of FCM clustering,establish a model to identify boiler operating conditions by PCA-RBF neural network method,the final realization of the boiler working conditions and identification.2.To achieve boiler optimization model establishment and optimization of operational guidance.After selecting the clustering efficient data under different conditions to establish the nuclear principal component model,constitute the model library.The PCA-RBF condition recognition model is used to identify the current operating conditions of the boiler,and the optimal model of the current operating conditions is selected to monitor and diagnose the online samples.When the boiler operating status does not match with the main element model,the boiler is judged to be in a low-efficiency operating state.By using the contribution graph of T2 and SPE statistics,find the main operating parameters that lead to inefficiencies.Combined with the real-time operating instructions of the boiler process,to ensure that the boiler operating in the current state of efficient conditions.The method is applied to a 40t/h hot water boiler in a university for verification.The results show that the system can effectively identify the operating status of the boiler,give timely and accurate operation guidance information for boile,and significantly improve the operation level of the boiler.