Research On Short-term Load Prediction Technology Of Thermal Power Units Under Abnormal Working Conditions

As the electricity load forms become diversitificated and renewable energy taking wind power as the representative is large-scale grid-connected,the grid has to deal with double radom disturbances from the source side of the load,so that it is necessary to strengthen the monitoring and the over all shedulling of electricty and power load changes in medium and small time scale.Several units such as China Electic Power Research Institute,Henan and Guangdong Electric Power Research Institute have carried out researches on primary frequency control capacity of thermal power unit load and the next research focuses on the short-term thermal power unit load forecasting under abnormal conditions within a time scale of one to ten minutes.Due to “Two Regulation”assessment of the power grid and deep peak regulation,thermal power unit runs close to the limit and power reliability reduces significantly.It is easy to appear a condition that the short-time generating units can't follow the change of AGC load instructions,influcing the safe and stable operation of power grid.Short-term load forecasting for thermal power unit aims to base on unit operation state,environmental factors and the complexity of AGC instruction to predict the future possible changes of the actual power load that the power grid can be told in advance in order to realize optimal dispath.The combination of mechanism modeling and numerical prediction algorithm is used to predict the thermal power generation load of the thermal power unit under abnormal conditions.The load object model is established,and the mechanism model of the whole power generation load control system is established based on this.The part with regularity changes is used as the state equation,and the mechanism model is used to predict the real-time input variables and irregularities.The part of the change is used as noise,according to its statistical law,using wavelet transform to process,and establishing a time series model for numerical prediction.Finally,the adaptive weighted data fusion algorithm is used to fuse the prediction results of the two to obtain the final prediction result.For a 600 MW subcritical unit,the load and pressure object model and the coordinated control system model are established to analyze the high-limit or low-limit,lock-up,hold,forced-up or force-down,pressure pull-back and other factors in the control system under abnormal conditions.The final experiment selected a 350 MW supercritical unit as the research object,and multi-scale decomposition of the load command by wavelet transform to establish the ARIMA model,predicting the change law based on the power generation load of t N moments and performing weighted summation to obtain numerical prediction.the result of.The mechanistic method is used to predict the mechanism of the unit,and finally the data fusion method is used to fuse the final results.According to the actual operating parameters of the unit and the current load command,the power generation load is predicted by a step of 5 minutes,and after 1 hour,it is compared with the actual power generation load of the unit.The results show that the relative error of the prediction of the mechanism and data fusion on the 5-minute scale is less than 5.2%.Compared with the results of the numerical prediction method and the mechanism prediction method,the prediction accuracy of the grid scheduling can be satisfied.The research results of the subject can be used as the basis for the harmonious scheduling of the two-way information interaction mode in the source network.