Journal Article10.1109/access.2024.3425590
A Deep Learning Framework for Net Load Forecasting With Unsupervised Behind-The-Meter Disaggregated Data
Chaichan Thepprom,Natawut Nupairoj,Peerapon Vateekul +2 more
3
TL;DR: The net load forecasting on the disaggregated series outperforms the net load series directly due to the accuracy of the unsupervised disaggregation of the BTM data, proving superior to the semi-supervised technique.
read more
Abstract: Recently, distributed photovoltaic (PV) generation has increased significantly, leading to a high penetration of behind-the-meter (BTM) solar generation systems. In this work, we aim to improve net load forecasting by disaggregating BTM components to provide better representation. For the disaggregation process, we propose an unsupervised contrastive-based optimization method for estimating BTM PV generation from the net load at the aggregated level. Our proposed method uses a deep neural network to leverage the strong correlation between solar irradiance and PV generation. This means that our proposed method is independent of the availability of BTM data and the assumption of a physical model. Furthermore, to obtain the best forecasted trends on the disaggregated series (pure load and PV generation), various recent forecasting models have been compared i.e. DeepAR, Temporal Fusion Transformer (TFT), and Time-series Dense Encoder (TiDE). The experiment is conducted on two real-world electricity prosumption datasets collected from New York and Texas. Results show that the net load forecasting on the disaggregated series outperforms the net load series directly. Such an improvement is due to the accuracy of our unsupervised disaggregation of the BTM data, proving superior to the semi-supervised technique.
read more
Chat with Paper
AI Agents for this Paper
Find similar papers on Google Scholar, PubMed and Arxiv
Write a critical review of this paper
Analyze citations of this paper to find unaddressed research gaps
Citations
An unsupervised non-intrusive load monitoring method for HVAC systems of office buildings based on MSTL
Wenjie Gang,Ying Zhang,Zhengkai Tu +2 more
Installed PV Capacity Detection on LV Substations: Comparison of Data-Driven and Model-Based Methods
Alexandre Gouveia,Md Umar Hashmi,Reinhilde D’hulst,Dirk Van Hertem,Alexandre Gouveia,Md Umar Hashmi,Reinhilde D’hulst,Dirk Van Hertem +7 more
Practical Method for Behind-the-Meter Solar PV Disaggregation
Dionathan S. Scheid,Adam Taylor, B.Sc.,Eduardo S. Finck,Bibiana P. Ferraz,Sergio Haffner,Luis Alberto Pereira,Mariana Resener,Adam Taylor, B.Sc.,Eduardo S. Finck,Luis A. Pereira,Mariana Resener +10 more
Abstract: The increasing adoption of rooftop solar photovoltaic (PV) generation in power distribution systems (PDS) requires innovative methods to estimate behind-the-meter (BTM) energy consumption and generation, given the widespread use of net metering. Existing approaches often rely on extensive historical data, advanced metering infrastructure (AMI), or smart meters. In contrast, we propose a practical energy disaggregation method that operates solely on monthly net energy imports and exports, estimating hourly gross values by leveraging reference generation profiles and typical load curves for residential, commercial, and industrial consumers. A clustering algorithm is used to generate probabilistic power generation for consumer groups within a region, while the sum of the differences between registered and estimated net monthly data is minimized through an iterative process. Validated with synthetic consumers across thirteen different classifications, the proposed method effectively estimates BTM energy consumption and generation. Thus, it provides utilities with a valuable tool for assessing prosumer behavior and understanding self-consumption patterns, helping prevent the underestimation of actual demand during PV generation periods while supporting grid operation and planning.
References
Long short-term memory
TL;DR: A novel, efficient, gradient based method called long short-term memory (LSTM) is introduced, which can learn to bridge minimal time lags in excess of 1000 discrete-time steps by enforcing constant error flow through constant error carousels within special units.
99K
DeepAR: Probabilistic Forecasting with Autoregressive Recurrent Networks
TL;DR: DeepAR is proposed, a methodology for producing accurate probabilistic forecasts, based on training an auto regressive recurrent network model on a large number of related time series, with accuracy improvements of around 15% compared to state-of-the-art methods.
1.5K
Temporal Fusion Transformers for interpretable multi-horizon time series forecasting
TL;DR: Temporal Fusion Transformer (TFT) as discussed by the authors uses recurrent layers for local processing and interpretable self-attention layers for long-term dependencies to learn temporal relationships at different scales.
969
The National Solar Radiation Data Base (NSRDB)
TL;DR: The National Solar Radiation Data Base (NSRDB) as discussed by the authors is a publicly open dataset that has been created and disseminated during the last 23 years, consisting of solar radiation and meteorological data over the United States and regions of the surrounding countries, and it provides solar irradiance at a 4-km horizontal resolution for each 30-min interval from 1998 to 2016 computed by the National Renewable Energy Laboratory's (NREL's) Physical Solar Model (PSM) and products from the National Oceanic and Atmospheric Administration's (NOAA's) Geostationary Oper
922
Levelized cost of electricity for solar photovoltaic and electrical energy storage
Chun Sing Lai,Malcolm McCulloch +1 more
TL;DR: In this paper, a new metric levelized cost of delivery (LCOD) is proposed to calculate the LCOE for the EES, which can be used to assist policymakers to consider the discount rate, the type of storage technology and sizing of components in a PV-EES hybrid system.
462