Structural optimization method for biomass boiler economizers

An optimization design method for structural parameters of biomass boiler economizers, comprising:

• S1, obtaining historical data of different heat exchangers in different biomass boiler units operating under different working conditions, and establishing a sample database, each sample in the sample database corresponding to one heat exchanger, an input of each sample being a structural parameter of the corresponding heat exchanger, working condition operating data of the corresponding heat exchanger under a load, and operating data of a corresponding biomass boiler unit where the corresponding heat exchanger is located, and a sample label being values of a plurality of target parameters to be optimized;
• S2, with the sample database as training data, training a heat exchanger residual self-attention convolution model by minimizing a weighted loss of all the target parameters to be optimized, so as to obtain a heat exchanger target parameter prediction model, the heat exchanger residual self-attention convolution model comprising a convolutional neural network and a residual attention network, a vector input to the convolutional neural network, and a dimensionality-reduced feature vector obtained sequentially through a convolution layer, a linear rectifying layer and a pooling layer, and in the residual attention network, a long-distance dependence relationship established and an attention weight generated for the dimensionality-reduced feature vector through multiple layers of self-attention modules, and finally predicted values of a plurality of target parameters output based on the attention weight;
• S3, performing dimensionality reduction screening on the structural parameters according to the attention weight finally obtained in S2, representing a contribution rate of each structural parameter to the target parameters to be optimized with the attention weight corresponding to the structural parameter, and selecting a screened structural parameter with the contribution rate higher than a threshold value as a structural parameter to be optimized, for any biomass boiler economizer to be optimized, with weighting results of all the target parameters to be optimized as a multi-target optimization function, performing multi-target optimization on the structural parameters to be optimized in the economizers within respective value ranges by adopting an iterative optimization algorithm, in an optimization process, outputting predicted values of all the target parameters to be optimized in each round of iteration according to a group of structural parameters to be optimized of the economizers by utilizing the heat exchanger target parameter prediction model, and then calculating a multi-target optimization function value according to the predicted values of all the target parameters to be optimized, and after optimization is completed, obtaining optimal values of the structural parameters to be optimized in the economizers.