Long-term and short-term memory network prediction method based on attention mechanism and logistic regression

Abstract

The invention discloses a long-term and short-term memory network prediction method based on an attention mechanism and logistic regression. The method comprises the steps of obtaining the ethylene production data, performing normalization processing on the ethylene production data; arranging the ethylene production data after normalization; and training the arranged ethylene production data by using a long short-term memory network model, carrying out weighted calculation on results of a plurality of historical time points according to an attention mechanism to obtain the data dependency relationships of different time points, and fitting the ethylene production data by using nonlinear transformation based on logistic regression to obtain the final predicted output. According to the technical scheme provided by the invention, the time sequence characteristics of the ethylene production data can be improved, so that the performance of the long-term and short-term memory network model is improved, the accurate prediction of the ethylene production process is realized, the ethylene production capacity is improved, and the purposes of energy conservation and emission reduction are realized.

Description

technical field [0001] The invention relates to the technical field of chemical production, in particular to a long-short-term memory network prediction method based on attention mechanism and logistic regression. Background technique [0002] The ethylene production industry has become the leading industry in the domestic petrochemical industry, and its output output has become one of the main symbols for judging a country's industrial development level. According to statistics, about 3.2 million tons of petroleum hydrocarbons are needed to produce 1,000,000 tons of ethylene, of which 18% (about 576,000 tons) is consumed by processing energy. Therefore, there is a large room for energy efficiency improvement in the domestic ethylene industry. The energy consumption cost of ethylene accounts for more than 50% of the operating cost of ethylene plant, so the establishment of the production prediction model of ethylene plant has a good guiding significance for reducing the ener...

AI Technical Summary

Problems solved by technology

However, complex chemical production data has the characteristics of high dimensionality, uncertainty, and noise, which makes it difficult to directly control the consumption of raw materials and cannot guarantee the product quality of complex chemical production

Images