Research And Application Of Initial Steam Pressure Optimization Method For Steam Turbine In Coal-Fired Power Plant

Along with the increase of electric power plant capacity and the change of powerconsumption, the peak-valley difference of electrical network increases day by day. Thesteam turbine units have to take part in changing the load and even to startup-halt foradjusting the peak output. Steam turbine is in the state of long-term variable loadoperation deviating from its design conditions, so thermal economy decreases greatly. Theinitial steam pressure of turbine must be optimized to ensure the steam turbine maintainthe best runing state under the variable working condition. The optimal initial pressure istaken as the setting value of main steam pressure, which can effectively reduce the heatrate of steam turbine and save the generating cost. This paper establishes the dynamicprediction models of main steam flow and heat rate and gives the optimal sliding pressureoperation curve according to the actual running state of turbine based on the improvingGravitation Search Algorithm (GSA) and Least Squares Support Vector Regression(LSSVR) algorithm. The main research results of this paper can be summarized asfollows:Firstly, this paper proposes an Improved Gravitation Search Algorithm (IGSA). Itsignificantly improves the optimization precision and the operation stability of standardGSA by introducing opposite learning strategy, elite strategy and boundary mutationstrategy. This paper verifies that the improved algorithm has better optimizationperformance than standard GSA through13benchmark functions. Also, this paperproposes an Adaptive Least Squares Support Vector Regression (ALSSVR) algorithm.This algorithm updates the model parameters by a recursion formula. It adopts a "two-stepstrategy" which avoids the complex calculation of the inverse matrix, and simultaneouslydeletes the corresponding support vector of minimum Lagrange multiplier absolute value,which is considered to influence the prediction precision for different samples. Thenumerical simulation experiment verifies that the ALSSVR applies to the online modeling.Secondly, a reverse modeling method is applyed to forecast the main steam flow. Thepredicted performance of LSSVR based on rough sets is better than that those of LSSVR, SVR and BP without attribute reduction. At the same time, as a new reverse modelingalgorithm, ALSSVR can be applied to the online modelling of main steam flow. Thepredicted value of main steam flow is taken as one of the input parameters of heat ratemodelling. Then, a hybrid modeling method based on IGSA-LSSVR is applied to forecastthe heat rate of steam turbine, in which IGSA is used to optimize the super parameters ofleast squares support vector machine, and effectively improve the prediction precision ofheat rate modeling. In addition, this paper still discusses the influence of differentcalculation fitness methods on the prediction precision.Finally, this paper establishs the online prediction model for steam turbine heat rateby ALSSVR algorithm. IGSA searches the optimal operation initial pressure within thefeasible operation initial pressure range according to the rule of minimizing the heat rateunder the variable working conditions of steam turbine. Therefore, the real-time predictionvalue of heat rate is taken as the IGSA’s fitness function. The real-time optimal operationinitial pressure curve can better guide the optimizing operation of steam turbine andeffectively save energy and reduce consumption.