Reconstruction of Key Parameters of Marine Supercharged Boiler based on PLS-SVM

摘要

A PLS-SVM method is used in order to improve the performance matching of Marine Supercharged Boiler and Turbocharged set. Value of key parameter of processing control is reconstructed through out of historical operating data. The traditional way of parameter reconstruction is theoretical calculation, which cannot solve the problem caused by performance degradation. Experiment shows that the PLA-SVM method could lead to a more accuracy and efficient results. Marine supercharged boiler system consists of two mainly parts, supercharged boiler and turbocharged set. Gas ejected by the boiler was used to drive a gas turbine. A coaxial compressor powered by the turbine provided pressured and heated air to the boiler's combustion process. There is no doubt that a good performance match between boiler and turbocharged set is very important for the system's stable, efficient and safe operating [1]. Control of the match is implemented by monitoring and regulating of several important operating parameters. Due to the wicked operating environment of marine equipments, a property degradation is unavoidable. It is necessary to reconstruct the key parameter because the design ones no longer represent the real condition of the system. The traditional method of parameter setting can be concluded in the following two parts. Firstly, carry out best running test. Let equipment run in a proper load and the value of the key parameters is recorded. This method is hard to achieve in marine vessels when they are on business. Secondly, theoretical calculate. The effect mainly depends on the accuracy of the calculating model. Besides, a theoretical value is hardly to reach in practice [2]. Excess air coefficient is an important economic indicator of boiler [3], its value correlated to the fuel quality, combustion style, combustion equipment and boiler load. In theoretical analysis, a nominal air mass flow was derived from the chemical formula of the fuel, and then multiplied a series of coefficient including excess air coefficient. Actually, due to the performance degradation of the equipment, such as ash accumulation in air pipeline, the theoretical excess air coefficient is larger than the real one. In the boiler control system, due to the limited technical means, we do not measure the air mass flow directly. A pressure drop of gas from gas in boiler furnace to gas at the end of the gas depurator is considered to be a measure of the air mass flow into boiler. According to the current load of the boiler, the gas pressure drop was maintained at a predetermined value. Here we defined the pressure drop as P_d(See Figure 1). This paper is focus on the modeling of the pressure drop with PLS-SVM method. It is based on system historical operating data and domain knowledge [4]. By gathering parameters' data in typical load level, we tried to express the pressure drop as a function of some controllable parameters.