| Modern coal chemical industry is an essential pathway for the strategy of clean and efficient utilization of coal in China,how to realize low-carbon and efficient production,reduce energy consumption and production costs,improve the competitiveness of enterprises has become an urgent problem for coal chemical enterprises.As the core subsystem of energy production and consumption of coal chemical enterprises,steam power system(SPS)needs to provide a large number of steam with different pressure to match the process demand,and its energy consumption accounts for about 30% of the total energy consumption of the whole plant.After long-term on-site investigation of Inner Mongolia Datang Keqi coal-to-gas enterprises,it is found that the steam demand under various pressures in the chemical park is subject to the influences of internal parameters such as feed composition and external environment,the steam demand is uncertain.In order to meet the steam and electricity demand of the whole plant at all times,the power center of the plant needs to adjust operating variables such as boiler load,turbine steam extraction and desuperheating water frequently.The key equipment often deviates from the design operating conditions or operates under partial load,resulting in reduced energy efficiency of equipment and systems and high fuel costs.Aiming at the problem that operation deviating from the optimal state in SPS caused by uncertain demand,this paper provides suggestions and solutions for coal-to-gas enterprises.Based on on-site investigation and data collection,this paper first introduces the steam supply and demand under various pressures in the whole plant,and obtains the relationship between the steam flow of each unit in the chemical park and the SNG output by data regression,equipment modeling and unit process simulation,so as to estimate the steam demand of each unit at different pressure levels under different SNG output.As the key equipment of SPS,this paper establishes the steady-state mechanism models for coal-fired boiler and multistage steam turbine.The internal node status of the boiler is obtained by analyzing the heat transfer mechanism of the coal-fired boiler,and the relationship between the multiple parameters of the boiler and the exhaust gas temperature is established for subsequent equipment operation optimization,the average relative error of the boiler heat transfer model is 2.2%.In view of the lack of steam extraction data at all stages of steam turbines,this paper establishes the multi-stage turbine and auxiliary equipment models,analyzes the relationship between isentropic efficiency and steam inlet flow of steam turbines at all stages.Taking the power generation as the reference,model validation is conducted on extraction condensing and extraction back-pressure steam turbines,the average relative errors are 1.74% and 2.96%,which can meet industrial applications.Short-term forecasting of uncertain steam demand can guide the power center to adjust the steam-to-electricity ratio and key equipment load in time,which is conducive to the scheduling and management of steam consumption in the whole plant.In order to improve the prediction accuracy,the K-means++ method is used to divide the annual load samples of steam demand at different pressures in the chemical park,which can effectively identify the steam yield changes caused by equipment start and stop.Then,an improved Markov Chain Monte Carlo(MCMC)prediction model is proposed,and the empirical cumulative distribution function(ECDF)and variable parameter robust kernel density estimation(VRKDE)methods are used to replace the traditional scene division and state decoding methods of Markov process,and the scene generation and reduction are carried out by Monte Carlo sampling and backward reduction to determine the optimal prediction duration and number of scenarios.The improved method can extract more sample distribution information and predict the steam demand at all levels of coalto-gas enterprises based on the division of working conditions.Compared with the traditional MCMC method,the improved model has the smallest mean squared error for steam demand prediction results at all levels,which improves the accuracy of steam demand forecasting.This paper subsequently optimizes the operation based on the boiler heat transfer model to reduce the exit gas temperature(EGT)and improve the boiler efficiency.The mechanism model is simplified and approximated based on the data-driven and surrogate model,which solves the problem that the traditional model needed to initialize the heat absorption of the heating surface and iteration for many times.The average relative error of the simplified model and the surrogate model is 2.8% and 2.2%.With the goal of minimizing EGT,the operation variables of the three subsystems of air flue gas,water feed and coal supply are optimized based on the efficient global optimization algorithm(EGO).By adjusting operating parameters such as feed water temperature and furnace oxygen content,the EGT can be reduced by 31.3～51.0 ℃,and the average fuel cost can be saved by 7.95 million yuan per year.The feasibility of the optimization results is proved by calculating the flue gas acid dew point.This method can efficiently optimize the operation parameters by using the surrogate model and the optimization method based on the surrogate model,analyzes the status of key nodes in the furnace and flue combined with heat transfer mechanism.The uncertainty optimization methods currently used for SPS mainly include robust optimization and stochastic programming.Aiming at the problem of conservative for robust optimization and sample priori for stochastic programming,this paper constructs an uncertain set based on VRKDE method,combines clustering scene division,establishes a multi-scenario data-driven adaptive robust optimization(DDARO)model for SPS,and the affine decision rule is used to establish a robust counterpart model for model solution.The uncertain set is used as the input of the two-stage stochastic programming method at the same time,and the data-driven interval two-stage stochastic programming method is constructed.Taking the steam power system of the coal-to-gas enterprise as an example,the effectiveness of the method is verified.The results of both optimization methods increase the steam demand of each scenario,improve the efficiency of key equipment,and eliminate irrationality in power generation.The annual operating costs of the SPS are reduced by 6.4 million yuan and 8.3 million yuan for the two optimization methods,and the comprehensive thermal efficiency increased by 5.4% and 6.2%.Adjusting the parameter confidence and budget degrees used in constructing uncertain sets can control the conservative level of optimization results.Compared with deterministic optimization and traditional robust optimization methods,both methods presented in this paper can achieve the trade-off between economic and conservatism. |
PDF Full Text Request