| Coal-fired power generation still accounts for the largest proportion in China's energy structure.The electric coal inventory of coal-fired power station is the guarantee of safe operation,but too much inventory will increase the operating cost of the power station.Because there is a period of lead time between the establishment and implementation of the coal feeding plan of the power station and the ordering and delivery of the electric coal,it is of great practical significance for the safe and stable operation of the coal-fired power station to realize the shortterm prediction of the electric coal stock and issue the stock early warning in advance.Based on the related research results at home and abroad,this paper summarizes the research status,existing problems and development direction of the short-term prediction method of coal power inventory.Due to the non-linear,periodic and non-stationary characteristics of the time series of electric coal inventory,the multi day prediction accuracy of a single linear prediction model is limited.In this paper,combining the advantages of ARIMAX model and LSTM model,the short-term prediction method of electric coal inventory based on combination prediction is studied.The main research results are as follows:(1)A short-term ARIMAX thermal coal inventory forecasting algorithm is proposed based on Prophet time series decomposition,and establishes a multicharacter ARIMAX forecasting model that takes influencing factors such as coal input and coal consumption as input.The Prophet time series decomposition method is used to mine the non-stationary information in the coal inventory time series.The variable growth rate logistic regression model is used to predict the trend of the electrical coal inventory and the discrete Fourier decomposition method is used to extract the periodicity,which effectively avoids the information loss caused by the direct differential stabilization of the ARIMAX model.Simulation results show that the proposed algorithm can improve prediction accuracy and information extraction accuracy compared with traditional ARIMA and ARIMAX algorithms.(2)A short-term forecast algorithm for thermal coal inventory based on ARIMAX-LSTM combined forecasting model is proposed.Aiming at the problem that the linear model of ARIMAX cannot predict the non-linear fluctuations in the time series of thermal coal inventory,a Vanilla LSTM model was established to directly output the multi-day prediction results.The total error function of the multi-step prediction was used as the model loss function to avoid the accumulation of errors caused by iterative prediction;In view of the fact that a single model cannot accurately reflect the complex characteristics of the time series of thermal coal inventory,the Stacking idea is used to fuse the ARIMAX model based on Prophet time series decomposition and the Vanilla LSTM model.The improved ARIMAX model is used to predict the linear component of the thermal coal inventory.The Vanilla LSTM model extracts the non-linear information on coal inventory.Simulation research shows that the proposed combined forecasting algorithm can integrate the advantages of a single model and improve the accuracy of short-term forecast of thermal coal inventory.(3)A distributed implementation framework of the short-term early warning system for power coal inventory is proposed,which includes data acquisition layer,distributed storage computing layer,data optimization processing layer and application layer.In this framework,based on big data integrated platform(TDH),a distributed version of the short-term prediction algorithm for power coal inventory is developed.According to the results of the short-term prediction algorithm,the safe available days of power coal in power plants are calculated,then the early-warning level of power plant inventory is determined to forewarn the safe operation risk in advance and to guide the the formulation and implementation of power plant coal import plan. |
PDF Full Text Request