Fengqi Si

A new approach for function approximation in boiler combustion optimization based on modified structural AOSVR

In the scheme of boiler combustion optimization, a group of optimal controller settings is found to provide recommendations to balance desired thermal efficiency and lowest emissions limit. Characteristic functions between particular objectives and controlling variables can be approximated based on data sets obtained from field tests. These relationships can change with variations in coal quality, slag/soot deposits and the condition of plant equipment, which can not be sampled on-line. Thus, approximation relationships based on test conditions could have little applicability for on-line optimization of the combustion process. In this paper, a new approach is proposed to adaptively perform function approximation based on a modified accurate on-line support vector regression method. Two modified criteria are proposed for selection of the unwanted trained sample to be removed. A structural matrix is used to process and save the model parameters and training data sets, which can be adaptively regulated by the on-line learning method. The proposed method is illustrated with an example and is also applied to real boiler data successfully. The results reveal their validity in the prediction of NOx emissions and function approximation, which can correctly be adapted to actual variable operating conditions in the boiler.

Integrated real-time optimization of boiler and post-combustion system in coal-based power plants via extremum seeking

AES Cayuga Unit 1 is a 160MW unit, equipped with a low nitrogen oxide (NOx) firing system and an anhydrous ammonia (NH3), TiO2=V2O5=WO3 selective catalytic reduction (SCR) system for NOx emission control. An ammonium bisulfate (ABS) probe was retrofit to the SCR to monitor ABS formation in real-time with the ultimate goal of minimizing air preheater (APH) plugging (ABS concentration) by regulating the APH bypass damper. Recent work on static optimization of coal-based power plants has played a crucial role in improving overall efficiency. However, static optimization falls short in dealing with real-time scenario changes (i.e., cycling unit load, coal quality, firing system maintenance conditions, subsystem failures, plant aging, etc.). Extremum seeking (ES) is proposed in this work to optimally tune boiler operation in order to minimize NOx production in real-time. The effectiveness of the ES adaptive controller in keeping the system at an optimal operation point in presence of input disturbances and system changes is demonstrated through simulations based on identified models of the boiler, SCR and APH systems.

Monitoring method of slurry quality in wet flue gas desulfurization system based on fuzzy C-means clustering

In this paper, a method of slurry quality monitoring and diagnosis in Wet Flue Gas Desulfurization(WFGD) system was proposed based on feature extraction of slurry quality and Fuzzy C-means(FCM) clustering. Focusing on the WFGD system of a 600 MW unit in a certain power plant, a new index for slurry quality monitoring was put forward. And clustering centers could be obtained to be the standard modes for slurry quality identification by adopting FCM to perform clustering analysis, in which the desulfurization efficiency and pH were regarded as feature information. Slurry quality diagnosis could be realized eventually by calculating the membership between the unknown samples and the standard modes of slurry quality. Furthermore, a fuzzy quantitative monitoring index was presented to quantitatively monitor the slurry quality state during its actual operation according to the theory of fuzzy membership. On the basis of diagnostic analysis of the field operating data, it demonstrates that the method raided in this dissertation can monitor the slurry quality state efficiently, providing foundation for operation adjustment.

Field Tests and Optimization Operation Research of a 600 MW Power Plant WFGD

High efficiency and low cost are the two main goals of desulfurization system operation optimization. Some field tests were performed on a wet desulfurization system for a certain 600 MW coal-fired power unit. By changing the factors such as the absorber entrance concentration of SO2, absorber slurry pH value, the number of slurry circulating pumps, the regularity of desulfurization efficiency in different working conditions was studied. The results indicated that the desulfurization efficiency became higher when the entrance concentration of SO2 was lower or the slurry ph value was higher. Running a pump at any load will increase the liquid–gas ratio so as the desulfurization efficiency. On the basis of field tests and the analysis of operation cost, the artificial intelligence methods were used in desulfurization system operation optimization. Firstly, BPNN models of desulfurization efficiency and booster fan current were built; secondly, an optimization model of desulfurization system operation cost was established to obtain the optimal parameters by the BBO algorithm, such as limestone slurry pH value, booster fan opening degree, liquid–gas ratio, etc. The optimal solution and data analysis showed that the proposed optimization control scheme in this chapter was effective to improve desulfurization efficiency and reduce operation cost.

Fault diagnosis of slurry pH data base on autoregressive integrated moving average and least squares support vector machines

A hybrid model that exploits the unique strength of the autoregressive integrated moving average (ARIMA) model and the least squares support vector machine (LSSVM) model was proposed for slurry pH value fault diagonosis in wet flue gas desulfurization (WFGD) system. The hybrid model was validated and evaluated by operating data and compared with individual ARIMA and LSSVM models. The results show that the hybrid prediction model can capture both linear and nonlinear patterns and has a better prediction performance than any single model. On this base, a sensor fault diagnosis system for pH value was designed by using the hybrid model. Firstly, the sensor fault location is determined on the reconstruction residuals, and then data reconstruction is implemented by the hybrid model instead of fault data. The simulation results from a 600 MW unit case study show that the model has high modeling precision and strong generalization. The fault diagnosis based on the hybrid model can diagnose the sensors fault and obtain credible reconstruction data.

Optimization of the Initial Steam Pressure for Supercritical Unit Based on ANN and PSO Algorithms

The initial steam pressure is one of the most important parameters affecting the heat rate of supercritical steam turbine. In this paper an approach for the optimization of the initial steam pressure is proposed. Firstly, the real-time data sets acquiring from process control system are processed as training data sets. Then, the characteristic functions for the relationship of unit load, initial steam pressure and heat rate can be developed by an artificial neural network algorithm. Based on these trained functions, an improved particle swarm optimization algorithm is introduced to calculate the optimal initial steam pressure. The proposed approach is applied in a 600MW unit to optimize the initial steam pressures at different unit loads. The results reveal the good performance of proposed approach and algorithms.