基于深度学习与热辐射成像耦合的炉内温度场在线测量

Research of On-line Measurement of Temperature Field in Furnaces Based on Deep Learning Coupled Thermal Radiative Imaging

利用热辐射成像模型计算训练数据,基于多层感知器(MLP)神经网络重建炉膛温度场,以某300MW锅炉为例进行在线测量实验,并通过机组调峰分析了燃料量和风量对温度的影响。结果表明:MLP神经网络对不同温度场重建结果的最大相对误差小于2%,添加噪声后最大相对误差也小于4%,其具有较好的泛化能力、预测能力和抗噪能力;炉内温度的升高趋势明显快于负荷,且风量对机组负荷的影响更大;所提出的耦合深度学习与热辐射成像的炉内温度场在线测量系统在提升煤电机组的灵活调峰能力方面具有较好的应用潜力。

Thermal radiative imaging model was used to calculate the training data, and the temperature field in furnace was reconstructed based on multi-layer perceptron(MLP) neural network. Taking a 300 MW boiler as an example, the online measurement experiment was carried out and the influence of fuel and air volume on temperature was analyzed by boiler load adjustment. Results show that the maximum relative error of the reconstruction results of the MLP neural network for different temperature fields is less than 2%, and the maximum relative error after adding noise is also less than 4%, which has good generalization, prediction, and anti-noise ability. The increase trend of furnace temperature is significantly faster than that of load, and the influence of air volume on unit load is greater. The proposed online measurement system of furnace temperature field coupled with deep learning and thermal radiative imaging has good application potential in improving the flexible peak regulation ability of coal-fired power units.