Research Article
Residential Energy Consumption Prediction Based on Encoder-Decoder LSTM
@INPROCEEDINGS{10.1007/978-3-031-31733-0_27, author={Junni Su and Lide Zhou and Fengchao Chen and Zejian Qiu}, title={Residential Energy Consumption Prediction Based on Encoder-Decoder LSTM}, proceedings={Smart Grid and Innovative Frontiers in Telecommunications. 7th EAI International Conference, SmartGIFT 2022, Changsha, China, December 10-12, 2022, Proceedings}, proceedings_a={SMARTGIFT}, year={2023}, month={5}, keywords={Load Forecast LSTM encoder-decoder}, doi={10.1007/978-3-031-31733-0_27} }- Junni Su
Lide Zhou
Fengchao Chen
Zejian Qiu
Year: 2023
Residential Energy Consumption Prediction Based on Encoder-Decoder LSTM
SMARTGIFT
Springer
DOI: 10.1007/978-3-031-31733-0_27
Abstract
Accurate forecast of load profile is of great benefit to electricity dispatch and power grid management. Recently, the widespread of smart meters enable the power system to collect fine-grained data from massive users. Also, the development of deep learning techniques allow the load forecasting to have better performance. However, the hyperparameter tuning in neural networks is a laborious but ineluctable part to achieve higher accuracy. Combing with huge information concealed in the fine-grained data, data mining is a significant process to accelerate hyperparameter tuning. In this paper, we first explore the metadata to help filter the data, compare the performances with different input and output by varying granularities, and evaluate predictability on various aggregation levels. Numerical studies suggest that on filtered data, accuracy has a higher correlations with predictability, and granularity of 1 h is the most appropriate.
