| 基于改进麻雀搜索算法优化BP神经网络的农业碳排放预测 |
| 摘要点击 2554 全文点击 212 投稿时间：2024-01-28 修订日期：2024-03-12 |
| 查看HTML全文
查看全文 查看/发表评论 下载PDF阅读器 |
| 中文关键词 农业碳排放 预测模型 BP神经网络 麻雀搜索算法 k折交叉验证 Tent混沌映射 高斯变异 |
| 英文关键词 agricultural carbon emissions predictive model BP neural network sparrow search algorithm k-fold cross validation Tent chaotic mapping Gaussian variation |
|
| 中文摘要 |
| 农业碳排放预测能够为农业领域实现碳达峰与碳中和目标的规划提供理论依据和数据支持. 为了优化现有的农业碳排放预测方法,以上海农场为例开展农业碳排放预测研究. 采用排放因子法核算了上海农场2011~2021年碳排放总量,基于碳排放核算结果构建了以种植业GDP、养殖业GDP和渔业GDP为指标的BP神经网络预测模型,并采用改进麻雀搜索算法对模型进行优化,最终利用优化后的BP神经网络模型对上海农场未来碳排放进行预测. 结果表明,改进麻雀搜索算法优化BP神经网络（TGSSA-BPNN）对于上海农场碳排放的预测结果准确率为96.14%,均方根误差为1.21万t·a-1,可决系数R2值为0.995 2,模型预测结果的准确度高,拟合效果良好；与优化前模型相比,TGSSA-BPNN模型多次运行结果的准确率稳定在95%左右,均方根误差稳定在20 000 t·a-1以下,R2值稳定在0.99以上,改进麻雀搜索算法提高了神经网络预测结果的准确性与稳定性,优化效果明显. 对于上海农场未来碳排放的预测结果表明,养殖业对于碳排放总量的影响起到了主导地位,控制养殖业规模可有效控制碳排放总量. |
| 英文摘要 |
| The accurate forecasting of agricultural carbon emissions is essential for formulating strategies to achieve carbon peak and neutrality objectives within the agricultural sector. However, existing methodologies for predicting agricultural carbon emissions have notable limitations. To address these shortcomings, Shanghai farm was considered as a case study to conduct research utilizing a neural network approach. Agricultural carbon emissions from the Shanghai farm from 2011 to 2021 were computed using the emission-factor method. Subsequently, a Back Propagation （BP） neural network model was developed to predict carbon emissions, employing the GDP of the planting, animal husbandry, and fishery sectors as input variables. The model was further improved through the application of an optimized sparrow search algorithm, which was then employed to forecast the future carbon emissions of the farm. The results show that the BP neural network improved via the optimized sparrow search algorithm demonstrated a prediction accuracy of 96.14%, a root mean square error （RMSE） of 12 100 t·a-1 and a correlation coefficient （R2） of 0.995 2. These metrics underscored the superior performance of the enhanced model. Compared with the multiple running results of pre-improved models, the neural network improved by the optimized sparrow search algorithm enhanced both the accuracy and stability of carbon emission prediction significantly, with the prediction accuracy consistently approaching approximately 95%, the root mean square error remaining below 20 000 t·a-1, and the correlation coefficient exceeding 0.99. Predictive analysis of future carbon emissions from the Shanghai farm indicated a predominant contribution from the animal husbandry sector to the total carbon emissions, suggesting that effective management of the scale of animal husbandry operations could significantly mitigate overall carbon emissions. |
|
|
|
