| Mainly to coal-fired thermal power generating units in Chinese power industry has an important role for coal-fired power plant boiler. As the coal-fired station boiler, monitor the furnace flame temperature is very important to the safe operation of boilers and its economy. Flame burning furnace process of dynamic, high-temperature, the parameters of complex conditions such as the characteristics of the traditional contact temperature measurement technology by the poor in addition to the environmental impact of the factors is limited, its own can not reflect the entire temperature distribution.The non-contact temperature measurement technology based on CCD image as its unique advantages to become one of the hot spots research in the field of flame temperature detection. This paper use the furnace flame temperature measurement as the background, describe systematically about related principle and major methods of temperature measurement of radiation thermometry based on CCD image. On the basis of mathematical models based on the CCD image flame temperature measurement of radiation, the introduction of the BP neural network to study the neural network, as a flame temperature of the radiation image data processing means to carry out a simulation experiment. In order to reduce the calculation error, an improved form of neural network input, the simulation results, the two methods of simulation results were compared to select the best network. Do the temperature measurement experiment to a flame image by the best network, and get the simulation map of temperature distribution.In addition, this paper present a forecast method of flame temperature based on neural network. Discuss the input and output vector design, and do simulation experiments, the results show that the temperature of short-term forecasting is feasible. This paper summarizes some of the conclusions of the main tasks and weaknesses, the outlook based on image processing temperature of the flame detection technology development. |
