RF Safe

Transfer learning strategies for solar power forecasting under data scarcity.

Scientific reports · 2022 · Unclear · Insufficient evidence

← Research stream
UnclearInsufficient evidencePaper 2415

Transfer learning (TL) strategies using a stacked LSTM model outperformed a conventional non-TL model and a smart persistence model for hourly production forecasting across 6 solar plants. Reported improvements included 12.6% better RMSE and 16.3% better fore…

Study snapshot

PublicationScientific reports · 2022
Study typeother
Reported directionUnclear
Evidence ratingInsufficient

Abstract

Accurately forecasting solar plants production is critical for balancing supply and demand and for scheduling distribution networks operation in the context of inclusive smart cities and energy communities. However, the problem becomes more demanding, when there is insufficient amount of data to adequately train forecasting models, due to plants being recently installed or because of lack of smart-meters. Transfer learning (TL) offers the capability of transferring knowledge from the source domain to different target domains to resolve related problems. This study uses the stacked Long Short-Term Memory (LSTM) model with three TL strategies to provide accurate solar plant production forecasts. TL is exploited both for weight initialization of the LSTM model and for feature extraction, using different freezing approaches. The presented TL strategies are compared to the conventional non-TL model, as well as to the smart persistence model, at forecasting the hourly production of 6 solar plants. Results indicate that TL models significantly outperform the conventional one, achieving 12.6% accuracy improvement in terms of RMSE and 16.3% in terms of forecast skill index with 1 year of training data. The gap between the two approaches becomes even bigger when fewer training data are available (especially in the case of a 3-month training set), breaking new ground in power production forecasting of newly installed solar plants and rendering TL a reliable tool in the hands of self-producers towards the ultimate goal of energy balancing and demand response management from an early stage.

Machine Enhanced Logic

AI evidence extraction

Structured fields below were generated from the paper’s abstract and metadata to make the evidence easier to inspect. Automated extraction may be incomplete or incorrect; verify important details against the original paper.

Study typeother
Effect directionunclear
PopulationNot supplied
Sample size6
ExposureNot supplied
Evidence strengthInsufficientConfidence: 74% · Peer reviewed: yes

Main findings

Transfer learning (TL) strategies using a stacked LSTM model outperformed a conventional non-TL model and a smart persistence model for hourly production forecasting across 6 solar plants. Reported improvements included 12.6% better RMSE and 16.3% better forecast skill index with 1 year of training data, with larger gaps when less training data were available (e.g., 3-month training set).

Outcomes measured

  • solar plant power production forecasting accuracy
  • RMSE
  • forecast skill index
View raw extracted JSON
{
  "study_type": "other",
  "exposure": {
    "band": null,
    "source": null,
    "frequency_mhz": null,
    "sar_wkg": null,
    "duration": null
  },
  "population": null,
  "sample_size": 6,
  "outcomes": [
    "solar plant power production forecasting accuracy",
    "RMSE",
    "forecast skill index"
  ],
  "main_findings": "Transfer learning (TL) strategies using a stacked LSTM model outperformed a conventional non-TL model and a smart persistence model for hourly production forecasting across 6 solar plants. Reported improvements included 12.6% better RMSE and 16.3% better forecast skill index with 1 year of training data, with larger gaps when less training data were available (e.g., 3-month training set).",
  "effect_direction": "unclear",
  "limitations": [],
  "evidence_strength": "insufficient",
  "confidence": 0.74,
  "peer_reviewed_likely": "yes",
  "keywords": [
    "transfer learning",
    "solar power forecasting",
    "stacked LSTM",
    "data scarcity",
    "smart meters",
    "RMSE",
    "forecast skill index",
    "hourly production",
    "smart persistence"
  ],
  "suggested_hubs": []
}
AI can be wrong. Always verify against the paper.

AI-extracted fields are generated from the abstract and metadata and may be incomplete or incorrect. This material is for informational purposes and is not medical advice.

Identifiers and source

DOI: 10.1038/s41598-022-18516-x
PubMed: 36030346

Read the original paper →

Related papers

← Back to the Research Library