Probabilistic individual load forecasting using pinball loss guided LSTM
By Dahua Gan, Mingyang Sun, Ning Zhang, Zongxiang Lu, Chongqing Kang. Applied Energy, 2019
402 citations recorded by the source corpus.
Abstract The installation of smart meters enables the collection of massive fine-grained electricity consumption data and makes individual consumer level load forecasting possible. Compared to aggregated loads, load forecasting for individual consumers is prone to non-stationary and stochastic features. In this paper, a probabilistic load forecasting method for individual consumers is proposed to handle the variability and uncertainty of future load profiles. Specifically, a deep neural network, long short-term memory (LSTM), is used to model both the long-term and short-term dependencies within the load profiles. Pinball loss, instead of the mean square error (MSE), is used to guide the training of the parameters. In this way, traditional LSTM-based point forecasting is extended to probabilistic forecasting in the form of quantiles. Numerical experiments are conducted on an open dataset from Ireland. Forecasting for both residential and commercial consumers is tested. Results show that the proposed method has superior performance over traditional methods.
View Paper: https://www.semanticscholar.org/paper/01aa584c03d5df9666882b0e17c5d0a00819a2e2
How researchers see this paper
The researchers whose trusted networks rank it highest, strongest advocate first. Every researcher gets a different order.
- Ying Sun
- F. Haghighat
- Benjamin C. M. Fung
- Danica Vukadinović Greetham
- Siddharth Arora
- Stephen A. Haben
- Marcus Voss
- Georgios Giasemidis
Citations in the corpus
- A review of the-state-of-the-art in data-driven approaches for building energy prediction (2020, background): “1 Historical data (previous 240 hours’ load or previous 70 days’ load) LSTM Residential Electric power demand Building 1hour 1day 24hour 7day 24 7 MRE, MAE, RMSE, SMAPE [151] 22000 2000 - 200 Time index, historical data LSTM Residential Electric power demand Building 30min 1hour 2hour 4hour 2 4 8 Pi”
- Review of low voltage load forecasting: Methods, applications, and recommendations (2021, methodology): “Individual Residential Very Statistical and time series Approaches [80] Short-term Machine Learning and other Artificial Intelligence Approaches [61,63,128,144,156,195,216,217] Comparing Methods [39,48,106,166] Probabilistic Forecasting [222,222,229] Hybrid, Combination and Ensemble Approaches [7, 5”