TL;DR
This paper introduces an enhanced Backpropagation (BP) neural network model, termed ISSA-BP, for predicting carbon dioxide (CO2) emissions in civil aviation. By integrating an Improved Sparrow Search Algorithm (ISSA) with tent mapping and adaptive t-distribution mutation, the model significantly improves prediction accuracy and robustness compared to existing methods, offering a crucial tool for developing effective emission reduction strategies.
What Problem Does This Paper Address?
The aviation industry is a rapidly growing source of greenhouse gas emissions, contributing significantly to climate change. Accurately forecasting these CO2 emissions is vital for developing effective energy-saving and emission-reduction policies. Traditional forecasting methods often struggle with the complex, nonlinear relationships inherent in aviation data. While neural networks, particularly BP networks, show promise, they can suffer from slow convergence and getting trapped in local optima during the optimization of their parameters (weights and biases). This paper aims to bridge this gap by proposing a novel approach that enhances the optimization process for BP networks, leading to more accurate and reliable CO2 emission predictions for civil aviation.
Methods
The core of this research lies in the development of an Improved Sparrow Search Algorithm (ISSA) to optimize the hyperparameters of a Backpropagation (BP) neural network.
Here's a breakdown of the key techniques:
- Backpropagation (BP) Neural Network: A standard, widely used multilayer feedforward neural network for complex prediction tasks. It learns by propagating errors backward through the network to adjust weights and biases. The paper uses a 4-13-1 architecture (4 input nodes, 13 hidden nodes, 1 output node).
- Sparrow Search Algorithm (SSA): A swarm intelligence optimization algorithm inspired by sparrow foraging and antipredation behaviors. It categorizes individuals into "finders," "followers," and "alerters" to search for optimal solutions.
- Improved Sparrow Search Algorithm (ISSA): This is the novel contribution. The authors enhance the traditional SSA by:
- Tent Mapping for Population Initialization: This method generates a more uniform initial population distribution across the search space, preventing premature convergence to local optima.
- Adaptive t-distribution Mutation: This strategy introduces disturbances during the search process. In the early stages, it behaves like a Cauchy distribution, promoting global exploration. As the algorithm progresses, it shifts towards a Gaussian distribution, facilitating local refinement and faster convergence. This dual approach aims to improve global search capabilities and escape local optima.
- ISSA-BP Model: The ISSA is used to optimize the weights and biases of the BP neural network, creating a more robust and accurate prediction model for civil aviation CO2 emissions.
Datasets
The study utilizes historical data for China's civil aviation industry from 1985 to 2019. The key indicators used as input features are:
- Transportation revenue
- Total transport turnover
- Total transportation volume
- Energy intensity
This data was collected from the Statistical Data of the Civil Aviation of China and the China Civil Aviation Compendium of Statistics. The dataset was split into training (80%) and testing (20%) sets.
Key Insights and Ideas
- The synergy of Tent Mapping and t-distribution Mutation is key: The paper argues that combining these two techniques in the Sparrow Search Algorithm significantly enhances its ability to avoid local optima and explore the search space more effectively, leading to better optimization of the BP network.
- ISSA addresses SSA's limitations: The traditional SSA can suffer from diminishing population diversity and convergence to local optima. The ISSA directly tackles these issues, offering improved global search and convergence speed.
- BP Networks benefit greatly from advanced optimization: The study demonstrates that while BP networks are powerful, their performance is highly dependent on the quality of their parameter optimization. The ISSA provides a superior optimization method compared to standard SSA and other swarm intelligence algorithms.
- Energy intensity is a critical factor: Regression analysis highlights that while transportation revenue, turnover, and volume drive emissions, energy intensity has a significant negative impact, suggesting that technological advancements and efficiency improvements are crucial for emission reduction.
Key Results
The ISSA-BP model demonstrated superior performance across multiple evaluation metrics:
- Prediction Accuracy: Achieved an R² value of 0.9996 on the test set, indicating excellent fitting.
- Error Metrics: Reported the lowest MAE (0.1262), RMSE (0.1823), MSE (0.0332), and MAPE (0.6985%) compared to other models.
- Comparison with other algorithms: Outperformed standard BP, SSA-BP, GWO-BP, PSO-BP, WOA-BP, and ABC-BP models in terms of accuracy and error control.
- Computational Efficiency: While not the fastest in total convergence time, the ISSA-BP achieved a significant improvement in optimization speed (23.8% faster than SSA-BP) and consumed less optimization time than many other algorithms.
- Robustness: Sensitivity analysis showed stable performance across reasonable parameter variations, and cross-validation confirmed consistent prediction precision.
Limitations and Future Work
- Indicator Dimensions: The study primarily uses macro-level indicators. Future research could incorporate micro-level variables (e.g., specific route characteristics, aircraft types) for more nuanced predictions.
- ISSA Improvement: While ISSA is an advancement, further improvements to its global search capabilities and convergence speed are possible. The authors suggest exploring diverse heuristic strategies or varied search mechanisms.
- External Factors: The paper focuses on internal aviation factors. Future work could integrate external measures like forest carbon sinks and global carbon market mechanisms for a more holistic approach to achieving net-zero emissions.
Summary updated