CN-121997253-A - Range extender temperature prediction method, range extender temperature prediction device, computer equipment and storage medium

CN121997253ACN 121997253 ACN121997253 ACN 121997253ACN-121997253-A

Abstract

The application relates to a range extender temperature prediction method, a range extender temperature prediction device, computer equipment and a storage medium, which comprise the steps of collecting time sequence data of a plurality of data sources of a range extender and determining contribution degree weights of the data sources to the range extender temperature prediction; processing time sequence data of each data source based on a preset processing method to obtain processed time sequence data of each data source, respectively taking the processed time sequence data of each data source as time sequence data of each spread data source, determining contribution degree weight of each spread data source, constructing input features of a temperature prediction model, determining position coding weight of feature values of each dimension feature in the input features, carrying out self-adaptive fusion processing on the input features according to the position coding weight of the feature values of each dimension feature to obtain fusion features, and inputting the fusion features into the temperature prediction model to obtain the prediction temperature of a range extender output by the temperature prediction model. The method can improve the accuracy and reliability of range extender temperature prediction.

Inventors

TAN SHUAI
ZHU HONGXIA
LUO JUN
ZHOU QINHAN
YOU FENG

Assignees

赛力斯汽车有限公司

Dates

Publication Date: 20260508
Application Date: 20260109

Claims (10)

1. A range extender temperature prediction method, the method comprising: Collecting time sequence data of a plurality of data sources of the range extender, and determining contribution degree weights of the data sources to temperature prediction of the range extender; Processing the time sequence data of each data source based on a preset processing method to obtain processed time sequence data of each data source, and taking the processed time sequence data of each data source as the time sequence data of the spread data source respectively; Determining the contribution degree weight of each spread data source based on the preset processing method and the contribution degree weight of each data source; Constructing an input feature of a temperature prediction model based on time sequence data of each data source, contribution degree weight of each data source, time sequence data of each dimension expansion data source and contribution degree weight of each dimension expansion data source, wherein the input feature comprises a plurality of dimension features, and each dimension feature corresponds to each data source or each dimension expansion data source; Determining the position coding weight of the characteristic value of each dimension characteristic in the input characteristics, and carrying out self-adaptive fusion processing on the input characteristics according to the position coding weight of the characteristic value of each dimension characteristic to obtain fusion characteristics; And inputting the fusion characteristics into the temperature prediction model to obtain the predicted temperature of the range extender output by the temperature prediction model.
2. The method of claim 1, wherein determining a contribution weight of each data source to range extender temperature prediction comprises: Collecting sample time sequence data of each data source and sample temperature data of a range extender corresponding to the sample time sequence data of each data source; and obtaining contribution degree weights of all the data sources according to sample time sequence data of all the data sources, sample temperature data of the corresponding range extender and the temperature prediction model and by adopting saprolitic additive interpretation.
3. The method according to claim 1, wherein the preset processing method includes a set mathematical transformation method and/or a feature interaction processing method, and the processing the time series data of each data source based on the preset processing method includes: processing time sequence data of a first target data source based on the mathematical transformation method, wherein the number of the first target data source is one or more; And/or processing time sequence data of a second target data source and a third target data source based on the characteristic interaction processing method, wherein the second target data source or the third target data source is any one of the plurality of data sources; The determining the contribution degree weight of each spread data source based on the preset processing method and the contribution degree weight of each data source comprises the following steps: If the mathematical transformation method is adopted to process the time sequence data of the first target data source, determining a contribution degree weight processing method based on the mathematical transformation method, and processing the contribution degree weight of the first target data source according to the contribution degree weight processing method to obtain the contribution degree weight of the spread data source corresponding to the first target data source; and if the characteristic interaction processing method is adopted to process the time sequence data of the second target data source and the third target data source, determining the contribution degree weight of the processed spread data source based on the contribution degree weight of the second target data source and the contribution degree weight of the third target data source.
4. The method of claim 1, wherein the constructing the input features of the temperature prediction model based on the time series data of each data source and the contribution weights of each data source, the time series data of each spread data source, and the contribution weights of each spread data source comprises: multiplying the time sequence data of each data source with the contribution degree weight of each data source to obtain the characteristic value of the dimension characteristic corresponding to each data source; Multiplying the time sequence data of each spread data source with the contribution degree weight of each spread data source to obtain the characteristic value of the dimension characteristic corresponding to each spread data source; And constructing the input features of the temperature prediction model based on the feature values of the dimension features corresponding to the data sources and the feature values of the dimension features corresponding to the data sources.
5. The method of claim 1, wherein prior to determining the position-coding weights for the feature values for each of the dimensional features, the method further comprises: Dividing the input features according to time windows to obtain sub-input features of each time window, wherein each sub-input feature comprises feature values corresponding to a plurality of continuous time points in each time window; the determining the position coding weight of the feature value of each dimension feature in the input features comprises the following steps: And aiming at any sub-input feature, carrying out position coding calculation on the feature value of each dimension feature based on the parity of the index of each dimension feature, the feature value of each dimension feature and the position index of the feature value of each dimension feature to obtain the position coding weight of the feature value of each dimension feature.
6. The method according to claim 5, wherein the adaptively fusing the input features according to the position coding weights of the feature values of the dimensional features to obtain fused features includes: calculating the similarity between any two dimension features according to the position coding weights of the any two dimension features; Constructing a similarity matrix of the input features according to the similarity between any two dimension features in the input features; taking the similarity matrix as a characteristic position adjacent matrix of the graph convolutional neural network; And carrying out graph convolution processing on the characteristic position adjacent matrix and the input characteristic through the graph convolution neural network to obtain the fusion characteristic.
7. The method of claim 6, wherein prior to the inputting the fusion feature into the temperature prediction model, the method further comprises: Determining contribution degree weights of all dimension features in sub-fusion features to range extender temperature prediction according to the sub-fusion features of any time window; Constructing a contribution matrix according to contribution weights of each dimension feature in the sub-fusion features to the range extender temperature prediction; Calculating the similarity between any two dimension features in the sub-fusion features, and constructing a feature similarity matrix according to the similarity between any two dimension features in the sub-fusion features; Collaborative filtering is carried out according to the contribution degree matrix and the feature similarity matrix, and a plurality of target dimension features with dimension feature correlation degree larger than a set threshold value are screened out; And constructing the fusion feature based on the feature value of each target dimension feature.
8. The method of claim 7, wherein after constructing the fusion feature based on feature values of each target dimension feature, the method further comprises: determining the position coding weight of each time window according to the count value of each dimension characteristic in the fusion characteristic aiming at the fusion characteristic of each time window; performing global time sequence information fusion processing on fusion features of each time window according to the position coding weight of each time window to obtain target fusion features; inputting the fusion characteristic into the temperature prediction model to obtain the predicted temperature of the range extender output by the temperature prediction model, wherein the method comprises the following steps: and inputting the target fusion characteristic into the temperature prediction model to obtain the predicted temperature of the range extender output by the temperature prediction model.
9. A computer device comprising a memory, a processor and a computer program stored on the memory and executable on the processor, characterized in that the processor implements the steps of the method according to any one of claims 1 to 8 when the computer program is executed by the processor.
10. A computer readable storage medium, on which a computer program is stored, characterized in that the computer program, when being executed by a processor, implements the steps of the method according to any one of claims 1 to 8.

Description

Range extender temperature prediction method, range extender temperature prediction device, computer equipment and storage medium Technical Field The present application relates to the field of vehicle control technologies, and in particular, to a range extender temperature prediction method, a range extender temperature prediction device, a computer device, and a storage medium. Background Along with the improvement of the endurance requirements of the electric automobile, the range extender is used as a key auxiliary power system, and the battery is charged through the power generation of the internal combustion engine so as to prolong the endurance mileage. However, the range extender generates a large amount of heat during operation. If the temperature of the range extender is too high, equipment damage, accelerated aging of a battery and even safety accidents can be possibly caused by abnormality, so that the real-time temperature prediction of the range extender is important to ensure the safety and stability of the electric automobile. In the prior art, the temperature of the range extender is predicted by a traditional physical modeling method, and the internal structure, the heat conduction characteristics and other factors of the range extender are usually required to be modeled in detail, and a great deal of engineering experience and assumption are relied on. Because the internal structure of the range extender is complex and the working conditions are variable, the traditional physical modeling method is difficult to cope with the dynamic changes, and the prediction accuracy is limited. With the rapid development of the deep learning technology, the range extender temperature prediction method based on the deep learning technology is attracting attention. The method can carry out deep learning modeling according to the history data of the range extender, and overcomes the limitation of a physical model to a certain extent. However, the high quality sample data available for model training is scarce, especially in extreme conditions. The problem of few samples severely limits the generalization capability and prediction accuracy of the deep learning model, and is difficult to meet the temperature prediction requirements of high accuracy and high reliability in practical application. Disclosure of Invention In view of the foregoing, it is desirable to provide a range extender temperature prediction method, apparatus, computer device, and storage medium, which can improve accuracy and reliability of range extender temperature prediction. A range extender temperature prediction method comprises the steps of collecting time sequence data of a plurality of data sources of a range extender, determining contribution degree weights of the data sources to range extender temperature prediction, processing the time sequence data of the data sources based on a preset processing method to obtain processed time sequence data of the data sources, taking the processed time sequence data of the data sources as time sequence data of expansion data sources respectively, determining the contribution degree weights of the expansion data sources based on the preset processing method and the contribution degree weights of the data sources, constructing input features of a temperature prediction model based on the time sequence data of the data sources and the contribution degree weights of the data sources, the time sequence data of the expansion data sources and the contribution degree weights of the expansion data sources, wherein the input features comprise a plurality of dimension features, the dimension features correspond to the data sources or the expansion data sources, determining position coding weights of feature values of the dimension features in the input features, carrying out self-adaptive fusion processing on the input features according to the position coding weights of the feature values of the dimension features to obtain fusion features, and inputting the fusion features into the temperature prediction model to obtain the temperature prediction model of the temperature extender. In one embodiment, determining the contribution degree weight of each data source to the range extender temperature prediction comprises collecting sample time sequence data of each data source and sample temperature data of the range extender corresponding to the sample time sequence data of each data source, and obtaining the contribution degree weight of each data source by adopting saprolidine additively interpretation according to the sample time sequence data of each data source, the sample temperature data of the corresponding range extender and a temperature prediction model. In one embodiment, the preset processing method comprises a set mathematical transformation method and/or a characteristic interaction processing method, the time sequence data of each data source are processed based on the preset processing m