Neural Network-Based Research On Soft Measurement Technology Of Coal Nature

All the measurement of coal ultimate analysis and industry analysis have great significance to the combustion efficiency. Coal ultimate analysis and industry analysis separately are evaluating the quality of coal from the standpoint and also closely linked. At present, most of the internal power plants are difficult to make the ultimate analysis measurement, but easily to determine the content of the proximate analysis. As the development of the soft-measurement technology, many scholars began to pay close attention to soft-measurement technology of the ultimate analysis. Artificial neural network is a new developed science imitating the human reason to analysis. It has good ability to solve nonlinearly relationship. It has become a popular research direction to research on the sof-measurement technology of coal ultimate analysis. This paper studies up on the basis of our predecessors’ research and poses the model of sof-measurement technology of coal ultimate analysis based on the neural network. The neural network model is easily trapped in local minimization, however, genetic algorithm has the ability of global optimization. So we make optimization of the initial parameters between the layers in the neural network based on the genetic algorithm, and train the net using the real proximate analysis data and the ultimate analysis data from the power plant. We use the test sample data to check the generalization ability of the model. The results show that the neural network could make the forcasting of the coal ultimate analysis data from the proximate analysis data, and the forcasting error will be more accurate when initial parameters of the network are optimized by the genetic algorithm. The model has some certain application meaning and provide powerful theoretical basis for the future research.