| 火电行业年度碳排放高精度预测模型构建方法 |
| Construction Method of High Precision Prediction Model for Annual Carbon Emissions of Thermal Power Industry |
| 投稿时间：2024-08-01 修订日期：2024-11-26 |
| DOI：10.19316/j.issn.1002-6002.2025.04.18 |
| 中文关键词: 碳排放 预测模型 相关性分析 粒子群优化 极限学习机 |
| 英文关键词:carbon emissions prediction model correlation analysis particle swarm optimization extreme learning machine |
| 基金项目:国家电网公司总部科技项目(1400-202457293A-1-1-ZN);国网能源研究院青年英才工程项目 |
|
| 通讯作者:周苏洋* 东南大学电气工程学院, 江苏 南京 210096 |
| 摘要点击次数: 333 |
| 全文下载次数: 1086 |
| 中文摘要: |
| 火电作为现阶段的电力行业碳排放大户,已逐渐成为推动实行碳减排与碳达峰的主要对象。然而,用于构建火电行业碳排放预测模型的相关数据大多呈现高维度小样本特征,导致现有基于神经网络理论的碳排放预测模型的输出精度难以得到有效提升。针对该问题,提出了一种基于粒子群优化-极限学习机的碳排放预测模型构建策略。首先,利用相关系数分析方法预处理碳排放影响因素,实现影响因素合理降维,降低极限学习机的泛化负担;其次,利用极限学习机的强泛化能力构建碳排放预测模型,并引入粒子群优化算法,求解极限学习机的最佳输入层权重和隐藏层偏差,提高碳排放预测模型精度;最后,基于历史数据和预测模型输出数据进行对比。结果表明,所构建预测模型的平均输出相对误差为2.77%,相较于传统极限学习机模型和粒子群神经网络预测模型分别降低了42.29%和32.60%。研究结果可为火电行业实现碳达峰碳中和目标提供理论支撑和技术借鉴。 |
| 英文摘要: |
| As the major carbon emitter in the current power industry,thermal power generation has gradually become the primary focus for promoting the implementation of carbon reduction and carbon peaking initiatives.However,most relevant data used for constructing carbon emission prediction models in this sector exhibit high-dimensional and small-sample characteristics,limiting the effectiveness of existing neural network-based carbon emission prediction models in improving output accuracy.To address this issues,this paper proposes a Particle Swarm Optimization-Extreme Learning Machine (PSO-ELM) strategy for carbon emission prediction model construction.Firstly,the correlation coefficient analysis method is used to preprocess the factors influencing carbon emissions,achieving reasonable dimensionality reduction and alleviating the generalization burden on the extreme learning machine.Secondly,an extreme learning machine with strong generalization capabilities is employed to construct the carbon emission prediction model,and particle swarm optimization is introduced to optimize input-layer weights and hidden-layer biases of the extreme learning machine,thereby enhancing the accuracy of the carbon emission prediction model.Finally,a comparison between historical data and the output data of the prediction model shows that the proposed prediction model has an output relative error of 2.77%,which is 42.29% and 32.60% lower than that of the traditional extreme learning machine model and neural network prediction model.This provides theoretical support and technical reference for planning the achievement of carbon peak and carbon neutrality goals in the thermal power industry. |
| 查看全文 查看/发表评论 下载PDF阅读器 |
|
|
|
