CN-121997289-A - Multi-dimensional carbon emission prediction method, system, equipment and medium

CN121997289ACN 121997289 ACN121997289 ACN 121997289ACN-121997289-A

Abstract

The invention relates to the technical field of carbon emission prediction, in particular to a multi-dimensional carbon emission prediction method, a multi-dimensional carbon emission prediction system, multi-dimensional carbon emission prediction equipment and a multi-dimensional carbon emission prediction medium, which comprise the steps of training a plurality of prediction models in parallel based on a core feature subset, dynamically triggering model retraining or switching based on a model performance detection result, and generating an initial prediction result; and carrying out multi-index comprehensive evaluation on the initial prediction result and the optimized prediction model, and outputting a target prediction result through a weighted integration or rule selection mechanism. The method has the beneficial effects that the difficult problems of data and model fracture in the carbon emission prediction of the gas-electricity enterprises are solved, and the high-precision and high-reliability prediction of the complex nonlinear emission behavior is realized by means of a dynamic self-adaption and interpretability decision mechanism.

Inventors

FANG BING
WANG SHANLI
ZHANG JIAYI
WU JIALE
Xiao Huanxiu

Assignees

海南电网有限责任公司

Dates

Publication Date: 20260508
Application Date: 20251202

Claims (10)

1. A multi-dimensional carbon emission prediction method is characterized by comprising the steps of, The method comprises the steps of obtaining multi-source heterogeneous data in the gas-electricity production process, and generating a standardized data set after pretreatment; Based on the data set, constructing time sequence statistical features through a sliding window with multiple time scales, and screening out a core feature subset based on feature importance evaluation; training a plurality of prediction models in parallel based on the core feature subset, and dynamically triggering model retraining or switching based on a model performance detection result to generate an initial prediction result; Based on the error index feedback of the initial prediction result, automatically optimizing the super-parameter configuration of each model, and generating a tuned and optimized prediction model; and carrying out multi-index comprehensive evaluation on the initial prediction result and the optimized prediction model, and outputting a target prediction result through a weighted integration or rule selection mechanism.
2. The method for predicting multidimensional carbon emissions according to claim 1, wherein said method is deployed through a micro-service architecture, supporting operation on a local server, edge gateway or cloud platform; the system provides a visual operation interface and a data interface, and based on user feedback information, adopts an incremental learning method to correct the prediction model in real time.
3. The method for predicting multidimensional carbon emissions according to claim 1, wherein said preprocessing comprises spatiotemporal alignment of the multi-frequency acquisition data by a sliding window mechanism; for high frequency data, calculating statistics within a window to generate reference frequency data points; for low frequency data, the reference frequency is extended by interpolation.
4. A multi-dimensional carbon emission prediction method as defined in claim 1 or 3, wherein said screening out the core feature subset based on the feature importance assessment comprises, Carrying out importance scoring on each feature in the feature set by adopting at least one method of correlation coefficient analysis, grade correlation analysis or mutual information calculation; and determining a core feature subset of the final input model training from the feature set based on the cross-validation stability assessment and the adaptive threshold.
5. The method of multi-dimensional carbon emission prediction as recited in claim 4, wherein said model performance detection logic comprises, When the continuous prediction error exceeds a threshold value calculated based on the historical error mean value and the standard deviation, automatically triggering a model retraining or switching mechanism; A feature analysis report is generated after each model training, including a SHAP value graph and an impact factor ranking graph.
6. The method for predicting multidimensional carbon emission as recited in claim 5, wherein said automatically optimizing said super-parametric configuration of each model employs a Bayesian optimization and grid search hybrid strategy; the optimized super parameters comprise training period, learning rate and regularization strength, and are optimized through a two-stage parameter adjusting method.
7. A multi-dimensional carbon emission prediction method as defined in claim 3, wherein said reference window length is 1 hour.
8. A multi-dimensional carbon emission prediction system employing the method of any one of claims 1-7, comprising: The acquisition module is used for acquiring multi-source heterogeneous data in the gas-electricity production process, and generating a standardized data set after pretreatment; the screening module is used for constructing time sequence statistical characteristics through a sliding window with multiple time scales based on the data set and screening out a core characteristic subset based on characteristic importance evaluation; The first generation module is used for training a plurality of prediction models in parallel based on the core feature subset, and dynamically triggering model retraining or switching based on a model performance detection result to generate an initial prediction result; the second generation module is used for automatically optimizing the super-parameter configuration of each model based on the error index feedback of the initial prediction result and generating a prediction model after the optimization; And the output module is used for carrying out multi-index comprehensive evaluation on the initial prediction result and the optimized prediction model, and outputting a target prediction result through a weighted integration or rule selection mechanism.
9. A computer device comprising a memory and a processor, the memory storing a computer program, characterized in that the processor implements the steps of the method of any of claims 1 to 7 when the computer program is executed.
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 of any of claims 1 to 7.

Description

Multi-dimensional carbon emission prediction method, system, equipment and medium Technical Field The invention relates to the technical field of carbon emission prediction, in particular to a multi-dimensional carbon emission prediction method, a multi-dimensional carbon emission prediction system, multi-dimensional carbon emission prediction equipment and a multi-dimensional carbon emission prediction medium. Background The energy industry is used as the key field of carbon emission, and is facing increasingly strict emission reduction requirements. The gas-electricity enterprises are taken as important components of the thermal power industry, carbon emission is influenced by a plurality of factors such as equipment structure, energy efficiency, external environment and the like, and an emission rule has obvious nonlinearity and complexity; However, the current mainstream carbon emission monitoring method still takes static accounting as the main part, generally depends on annual emission factors or fuel consumption lists, cannot meet the requirements of real-time, accurate and intelligent prediction, is difficult to effectively fuse multi-source heterogeneous data, lacks a unified data preprocessing and feature extraction mechanism, is mostly based on a single modeling mode, lacks a multi-model integration strategy for high space-time resolution, and has poor stability and interpretability of a prediction result. Disclosure of Invention In order to solve the technical problems, the invention provides the following technical scheme: in a first aspect, the invention provides a multi-dimensional carbon emission prediction method, which comprises the steps of obtaining multi-source heterogeneous data in a gas-electricity production process, and generating a standardized data set after pretreatment; based on the data set, constructing time sequence statistical features through a sliding window with multiple time scales, and screening out a core feature subset based on feature importance evaluation; Training a plurality of prediction models in parallel based on the core feature subset, and dynamically triggering model retraining or switching based on the model performance detection result to generate an initial prediction result; Based on error index feedback of an initial prediction result, automatically optimizing super-parameter configuration of each model, and generating a tuned and optimized prediction model; And carrying out multi-index comprehensive evaluation on the initial prediction result and the optimized prediction model, and outputting a target prediction result through a weighted integration or rule selection mechanism. As a preferable scheme of the multi-dimensional carbon emission prediction method, the method is deployed through a micro-service architecture and is supported to run on a local server, an edge gateway or a cloud platform; the system provides a visual operation interface and a data interface, and based on user feedback information, adopts an incremental learning method to correct the prediction model in real time. As a preferable scheme of the multi-dimensional carbon emission prediction method, the preprocessing comprises the steps of carrying out space-time alignment on multi-frequency acquisition data through a sliding window mechanism; for high frequency data, calculating statistics within a window to generate reference frequency data points; for low frequency data, the interpolation method is used for expanding to the reference frequency. As a preferred embodiment of the multi-dimensional carbon emission prediction method of the present invention, wherein the core feature subset is selected based on feature importance assessment, comprising, Carrying out importance scoring on each feature in the feature set by adopting at least one method of correlation coefficient analysis, grade correlation analysis or mutual information calculation; And determining a core feature subset of the final input model training from the feature set based on the cross-validation stability assessment and the adaptive threshold. As a preferred embodiment of the multi-dimensional carbon emission prediction method of the present invention, wherein the model performance detection logic comprises, When the continuous prediction error exceeds a threshold value calculated based on the historical error mean value and the standard deviation, automatically triggering a model retraining or switching mechanism; A feature analysis report is generated after each model training, including a SHAP value graph and an impact factor ranking graph. As a preferable scheme of the multi-dimensional carbon emission prediction method, the super-parameter configuration of each model is automatically optimized by adopting a Bayesian optimization and grid search mixed strategy; the optimized super parameters comprise training period, learning rate and regularization strength, and are optimized through a two-stage parameter ad