Prediction Of Short-term Controllable Output Interval Of Boiler Based On LSTM

At present,thermal power generation is still the main form of power generation in China With the proposal of the "dual carbon" goal and the consumption problem faced by renewable energy power generation,it is urgent to achieve the purpose of energy conservation and emission reduction and optimization of power resource allocation through the flexibility transformation of thermal power units.However,changes in the operating environment of the unit,the quality of coal entering the furnace,and the state of auxiliary machines will affect the output range of thermal power units,which will bring difficulties to the power grid dispatching.In view of the above problems,in order to grasp the power generation load boundary of the thermal power unit in advance,this paper takes the 660 MW ultra-supercritical coal-fired generating unit as the research object,and uses the historical data in the DCS system of the power plant to study the short-term adjustable output prediction of the boiler,including boiler side data processing,the establishment of the LSTM boiler output prediction model and the maximum and minimum output prediction.Firstly,by introducing the overall structure and design parameters of the DC boiler,the correlation analysis of the boiler side multi-variates is carried out,the parameters affecting the output are screened out,the Laida criterion is used to eliminate the possible outliers in the big data of the power plant,and the noise reduction treatment is carried out with the help of the wavelet transformation method,and then the steady-state data is screened by the sliding window method.Secondly,in order to reduce the sample dimension of the model input,PCA dimensionality reduction is further adopted to lay a data foundation for the establishment of a output prediction model.Next,the full working condition data and steady-state data are used as model training samples,and the simulation results show that the modeling accuracy of steady-state data is higher,and the performance of LSTM and SSA-LSTM models is compared to verify the superiority of the SSA optimized LSTM prediction model,and the boiler output prediction model is built based on the steady-state data.Based on this,the high,medium and low load test samples under steady-state conditions are simulated,and the root mean square error is found to be within 1.50%,which confirms the reliability of the model.Finally,considering that coal quality and auxiliary machine performance are the main factors affecting boiler output,the maximum adjustable output value is obtained based on the established model and combined with the boiler side parameter boundary provided by the intelligent early warning system of the power plant according to the database screening according to the coal quality coefficient and the combination of the mill under the current steady-state working conditions.Similarly,the minimum output value is obtained based on the model,and the relative error between the two is 1.20% compared with the low-load stable combustion test data of the power plant.Through the power plant case analysis,the maximum and minimum output values are obtained based on the LSTM prediction model established in this paper,and then the adjustable output range of the boiler is determined,which provides a reliable short-term output reference range for the power grid dispatching center,realizes the safe dispatch of the boiler,and guides the unit to reasonably prepare for grinding and stopping grinding,ensuring the safe and stable operation of the power grid and improving the level of new energy consumption.