基于SVD-SOA-GRU-LSTM算法的天然气负荷预测

Natural Gas Load Forecasting Based on SVD-SOA-GRU-LSTM Algorithm

摘要: 本文提出一种由海鸥优化算法(SOA)优化门控循环单元(GRU)并结合奇异值分解(SVD)和长短期记忆神经网络(LSTM)的天然气负荷预测算法。采用SOA对GRU的模型参数进行优化，得到最佳值，利用SVD对原始数据进行降维和特征提取以提高数据质量，最后使用LSTM进行残差预测，结合GRU的预测结果作为最终预测结果，进一步提高了模型的预测精度，使用均方根误差(RMSE)和平均绝对百分比误差(MAPE)作为评价标准，模型的预测结果分别为4766和5.61%。

Abstract: This paper proposes a natural gas load forecasting algorithm that combines the Seagull Optimization Algorithm (SOA) to optimize the Gated Recurrent Unit (GRU), and incorporates Singular Value Decomposition (SVD) and Long Short-Term Memory (LSTM). The SOA is used to optimize the model parameters of GRU and obtain the optimal values. SVD is employed to reduce dimensionality and extract features from the original data, enhancing data quality. Finally, LSTM is utilized for residual prediction, combining the prediction results of GRU as the final forecast. This approach further improves the prediction accuracy of the model. The evaluation criteria are Root Mean Square Error (RMSE) and Mean Absolute Percentage Error (MAPE). The model achieves an RMSE of 4766 and a MAPE of 5.61% in the prediction results.