AI research for low‑carbon energy has exploded—thousands of ML/DL papers now focus on forecasting and optimization—but scholarly growth far outpaces synthesis and evidence of real‑world cost, emissions, and market impacts.

Main Finding

AI research applied to low‑carbon and renewable energy has rapidly transitioned from a niche topic into a large, fast‑growing research domain dominated by machine learning and deep learning methods—primarily used for forecasting/prediction and optimization—with a growing explicit linkage to sustainable development (5,192 documents identified).

Key Points

• Rapid expansion: annual scientific output rose sharply from 239 documents in 2021 to 2,366 in 2025.
• Technical concentration: the field’s core methods are machine learning and deep learning.
• Primary applications: forecasting (e.g., demand, generation) and optimization (e.g., grid operation, storage dispatch, resource allocation).
• Sustainability connection: the reviewed corpus contains 5,192 documents explicitly linked to sustainable development.
• Analysis approach: the study combined bibliometric mapping and a strategic (SWOT) appraisal to both quantify and interpret the research landscape.
• Knowledge state: broad and fast‑growing literature, but consolidation (synthesis, standards, evaluation of real‑world impacts) lags behind volume growth.

Data & Methods

• Data source: Scopus bibliographic database.
• Coverage/timespan: up to and including 2025 (annual counts cited for 2021–2025).
• Bibliometric techniques used:
  • Descriptive performance analysis (publication counts, growth trends).
  • Science mapping: bibliographic coupling, citation analysis, co‑authorship networks, co‑citation analysis, keyword co‑occurrence.
• Qualitative extension: a coded core sample of documents was reviewed and used for strategic interpretation (including a SWOT appraisal).
• Corpus size: at least 5,192 documents tied explicitly to sustainable development; total corpus larger (growth figures indicate thousands of annual items by 2025).

Implications for AI Economics

• Productivity and cost effects: improved forecasting and optimization can reduce operating costs, lower system inefficiencies, and shift marginal costs in power systems—affecting prices, firm profitability, and investment decisions in generation and storage.
• Investment signals: faster, more accurate forecasting and optimization may de‑risk variable renewable investments and storage, potentially accelerating capital flows into low‑carbon infrastructure.
• Market structure and competition: AI‑enabled operational advantages (better dispatch, bidding strategies) could change competitive dynamics—raising concerns about market power if advanced AI capabilities concentrate in a few firms.
• Regulatory and market design needs: regulators will need to update market rules and transparency requirements (e.g., algorithmic bidding, data sharing, testing of AI systems to avoid systemic risks).
• Externalities and rebound risks: efficiency gains could induce rebound effects (higher electricity use or economic activity) unless paired with policy measures that lock in emissions reductions.
• Distributional consequences: gains may be uneven across regions and firms—areas with better data, digital infrastructure, or human capital will capture more benefit, implying equity considerations for policy.
• Labor and skills: demand for data scientists, engineers, and specialists in energy‑AI integration will increase; transitions for traditional energy-sector jobs require retraining and social policies.
• Measurement and evaluation gaps: bibliometric growth does not equate to deployment impact—economists should prioritize causal/empirical studies quantifying real‑world impacts of AI on emissions, costs, prices, and employment.
• Research & policy priorities:
  • Conduct firm- and grid-level impact evaluations (causal inference) of AI deployments.
  • Study how AI changes investment timing and capacity choices under uncertainty.
  • Assess market power and design algorithm‑aware regulation.
  • Incorporate distributional analysis and workforce transition planning into policy responses.
  • Promote data governance and standards to enable replicable, cross‑jurisdictional benefits.

Limitations to note for interpreting the study’s conclusions: reliance on Scopus may miss gray literature and proprietary implementations; bibliometrics reveal research activity and intellectual structure but not necessarily real‑world deployment outcomes or net emissions impacts.