Prediction of Carbon Emission of the Transportation Sector in Jiangsu Province-Regression Prediction Model Based on GA-SVM

Zhenggang Huo, Xiaoting Zha, Meng‐Yao Lu, Tianqi Ma, Zhichao LuSustainability 2023 · DOI

TL;DR

This paper addresses the critical need for accurate carbon emission predictions in the transportation sector to meet China's ambitious carbon neutrality goals. The authors propose a novel GA-SVM (Genetic Algorithm-Support Vector Machine) model that significantly improves prediction accuracy compared to other optimization methods, offering a more reliable tool for policymakers in Jiangsu Province.

What Problem Does This Paper Address?

Achieving "carbon peak" and "carbon neutrality" targets is a global imperative, and accurately forecasting carbon emissions is crucial for effective policy-making. The transportation sector is a significant contributor to these emissions. However, predicting these emissions is challenging due to the complex interplay of various influencing factors and the often limited availability of historical data, which can hinder the performance of traditional machine learning models. This paper specifically tackles the problem of predicting carbon emissions and identifying the carbon peak time for the transportation sector in Jiangsu Province, a region with substantial economic activity and a large transportation footprint.

Methods

The core of this research lies in developing a hybrid prediction model: the Genetic Algorithm-Support Vector Machine (GA-SVM).

  • Support Vector Machine (SVM): SVM is chosen for its effectiveness in handling small datasets and its ability to model complex, non-linear relationships, which are common in environmental and economic data. The paper details the mathematical underpinnings of SVM regression, including the use of an insensitive loss function and the introduction of relaxation variables to prevent overfitting.
  • Genetic Algorithm (GA): To overcome the limitations of SVM, particularly its sensitivity to parameter tuning (penalty factor 'c' and kernel function parameter 'g'), a Genetic Algorithm is employed. GA is a metaheuristic inspired by natural selection and evolution. It iteratively refines a population of potential solutions (chromosomes representing model parameters) through selection, crossover, and mutation to find the optimal set of parameters that minimize prediction error.
  • Hybridization (GA-SVM): The GA is used to optimize the 'c' and 'g' parameters of the SVM model. The objective function for the GA is the error between the predicted and actual values of the test set. This approach aims to find a global optimum for the SVM parameters, avoiding local optima that might plague other optimization methods.
  • Dimensionality Reduction (PLS): To address potential multicollinearity issues among the eight selected influencing factors, Partial Least Squares (PLS) is used to extract four principal components. This reduces the dimensionality of the data while retaining most of the explanatory power, leading to a more stable and robust model.

The paper also compares the GA-SVM model against two other optimization algorithms for SVM: Particle Swarm Optimization (PSO-SVM) and Whale Optimization Algorithm (WOA-SVM).

Datasets

The study utilizes historical data for the transportation sector in Jiangsu Province from 2002 to 2020. The data includes:

  • Carbon Emission Data: Derived from the Jiangsu energy balance table within the China Energy Statistics Yearbook.
  • Influencing Factors:
    • Population size
    • GDP per capita
    • Civil vehicle ownership
    • Passenger turnover
    • Freight turnover
    • Urbanization rate
    • Industry structure
    • Carbon emission intensity

This data was logarithmized, normalized, and processed using PLS. The data from 2002–2016 was used for training the GA-SVM model, and the data from 2016–2020 was used for testing.

Key Insights and Ideas

  • The Power of Hybridization: Combining GA with SVM proves to be a potent strategy for carbon emission prediction. GA's ability to explore a wide parameter space effectively mitigates SVM's susceptibility to suboptimal parameter choices.
  • Multicollinearity is a Real Problem: The study highlights that many factors influencing carbon emissions are highly correlated. Simply discarding variables can lead to instability, making dimensionality reduction techniques like PLS essential for building reliable models.
  • GA-SVM Outperforms Competitors: The GA-SVM model consistently demonstrates superior prediction accuracy (lower MAPE, higher R²) compared to PSO-SVM and WOA-SVM, suggesting its robustness and effectiveness for this specific prediction task.
  • Scenario-Based Forecasting: The paper's exploration of three distinct scenarios (standard, low-carbon, high-carbon) provides a nuanced outlook on future carbon emission trends and peak times, offering valuable insights for policy planning.

Key Results

  • The GA-SVM model achieved a Mean Absolute Percentage Error (MAPE) of 0.0297% and a correlation coefficient (R²) of 0.9082 on the test set.
  • This performance was significantly better than PSO-SVM (MAPE: 0.3208%, R²: 0.8450) and WOA-SVM (MAPE: 0.6435%, R²: 0.1203).
  • The relative error of the GA-SVM model's predictions was consistently small, not exceeding 2%, indicating high discrimination accuracy and model stability.
  • The study projected that under a low-carbon scenario, Jiangsu Province's transportation sector could achieve its carbon peak one year earlier than the standard scenario.

Limitations and Future Work

  • Long-Term Prediction Error: The authors acknowledge that while GA-SVM is effective for short- to medium-term predictions, the error tends to increase with longer prediction horizons. This suggests a need for more extensive training and testing datasets for robust long-term forecasting.
  • Data Granularity: The study relies on aggregated provincial data. Future work could explore more granular data (e.g., city-level) for finer-grained predictions and policy recommendations.
  • Additional Influencing Factors: While eight factors were considered, other variables (e.g., technological advancements in vehicle efficiency, specific policy interventions) could potentially be incorporated to further refine the models.

Summary updated

Abstract

To meet the twin carbon goals of “carbon peak” and “carbon neutrality”, it is crucial to make scientific predictions about carbon emissions in the transportation sector. The following eight factors were chosen as effect indicators: population size, GDP per capita, civil vehicle ownership, passenger and freight turnover, urbanization rate, industry structure, and carbon emission intensity. Based on the pertinent data from 2002 to 2020, a support vector machine model, improved by a genetic algorithm (GA-SVM), was created to predict the carbon peak time under three distinct scenarios. The penalty factor c and kernel function parameter g of the support vector machine model were each optimized using a genetic algorithm, a particle swarm algorithm, and a whale optimization algorithm. The results indicate that the genetic algorithm vector machine prediction model outperforms the particle swarm algorithm vector machine model and the whale optimization vector machine. As a result, the model integrating the support vector machine and genetic algorithm can more precisely predict carbon emissions and the peak time for carbon in Jiangsu province.
BibTeX
@inproceedings{huo2023prediction,
  author = {Zhenggang Huo and Xiaoting Zha and Meng‐Yao Lu and Tianqi Ma and Zhichao Lu},
  title = {{Prediction of Carbon Emission of the Transportation Sector in Jiangsu Province-Regression Prediction Model Based on GA-SVM}},
  booktitle = {Sustainability},
  year = {2023},
  doi = {10.3390/su15043631},
}
Figure