基于Conv1d-LSTM模型的能源分配预测

能源分配问题往往与其所在区域环境有关,能源分配的预测可以通过当地环境因素数据来推测之后对该区域的能源分配数值,最大程度上分配好能源. LSTM网络预测短期效果良好,但预测较长时期的数据会导致误差积累,速度慢且准确性差; Informer是近期新提出的能源预测算法模型,速度快但在该任务上预测能力不够.本文提出Conv1d-LSTM模型,预测结果优于上述两个模型,具有更低的平均绝对误差和均方根误差.

Energy Distribution Prediction Based on Conv1d-LSTM Model

Energy distribution is often related to the local environment. Regarding energy distribution prediction, data on local environmental factors can be availed to predict the value of energy to be distributed to the region, thereby maximizing the extent of proper energy distribution. The long short-term memory (LSTM) network, despite its favorable short-term prediction effect, is weakened by error accumulation, a slow speed, and poor accuracy when it is used for long-term data prediction. As a new algorithmic energy prediction model recently proposed, Informer is fast but not sufficiently capable of prediction in this task. This study proposes a Conv1d-LSTM model that achieves better prediction results than those of the above two models with a smaller mean absolute error (MAE) and root mean square error (RMSE).