About

AI-Driven PV Forecasting for Grid Optimisation and Flexibility Management

EO x Grid is a feasibility study focused on improving electricity distribution grid operations through accurate, space-based photovoltaic (PV) generation forecasting. The project is funded under the European Space Agency’s (ESA) ARTES 4.0 Downstream Applications programme and explores how Earth Observation (EO) data combined with artificial intelligence can support grid stability, renewable integration, and emerging flexibility markets.

The Challenge

Electricity distribution grids are undergoing rapid change due to the growing share of distributed energy resources, particularly rooftop and small-scale photovoltaic systems. A large proportion of these assets do not provide real-time generation data to grid operators. As a result, distribution system operators (DSOs) face limited visibility of actual PV production at the local level.

This lack of observability reduces the accuracy of grid forecasting and planning, increases operational uncertainty, and makes it more difficult to manage congestion, voltage issues, and flexibility resources. Existing forecasting solutions often lack the spatial granularity and timeliness required at the low- and medium-voltage level, especially for PV assets without direct measurements. These limitations can lead to inefficient grid operation, higher balancing costs, and conservative grid development decisions.

Project Vision

EO x Grid aims to develop an EO- and AI-driven PV forecasting service tailored to the operational and planning needs of electricity utilities. By leveraging space-based data and advanced machine-learning models, the project seeks to deliver accurate and timely PV generation forecasts at local grid scale.

The vision is to enable grid operators to make better-informed decisions related to grid operation, planning, and participation in local flexibility markets, even in areas where direct PV production data is unavailable.

Target Users

The primary end-users of EO x Grid are electricity utilities, in particular distribution system operators. Strategic partners include smart-grid software providers offering ADMS and DERMS solutions. Secondary users include PV and battery asset operators, aggregators, balance responsible parties, and energy traders who depend on reliable short-term PV forecasts for operational and market decisions.

Funding Context

EO x Grid is implemented as a feasibility study under the European Space Agency’s ARTES 4.0 Downstream Applications programme. The study assesses both the technical performance and the commercial viability of the proposed service as a foundation for a future demonstration project and commercial deployment.

Project Objectives

The main goal of EO x Grid is to assess the feasibility of a highly accurate PV forecasting service that integrates Earth Observation data from Meteosat Third Generation (MTG) satellites with artificial intelligence techniques.

Key objectives include:

• validating the achievable forecasting accuracy for nowcasting and day-ahead PV generation,
• enabling PV forecasts aggregated at transformer or feeder level, including assets without historical or real-time data,
• identifying the minimum technical requirements and number of reference points needed to achieve the desired accuracy,
• and evaluating the operational and economic value of the service for electricity utilities.

Target performance indicators range from below 5% forecast error for very short-term forecasts to around 20% for day-ahead horizons, with prediction horizons of up to 48 hours.

Core Approach and Key Activities

The EO x Grid service combines several key elements. High-frequency satellite data from Meteosat MTG is used to capture cloud dynamics and irradiance conditions with low latency. This EO data is combined with numerical weather prediction data and grid-related information within an AI-based forecasting framework.

Where available, selected PV assets act as reference points, while the forecasting models extrapolate results across the wider grid using sensitivity analysis and asset metadata. Forecasts are validated at multiple aggregation levels, from individual assets to transformers and feeders, ensuring relevance for both operational and planning use cases.

Alongside technical validation, the project evaluates the commercial feasibility of the service, including integration into existing utility workflows and smart-grid software platforms.

Innovation and Differentiation

The core innovation of EO x Grid lies in its use of Meteosat MTG data to achieve high-resolution, low-latency PV forecasting at local grid level. This enables forecasting for PV assets that lack direct measurements, without the need for extensive IoT deployment.

Compared to existing solutions, EO x Grid targets higher accuracy and faster updates for distribution-level applications. Its focus on transformer-level aggregation and meter-less observability directly addresses one of the most critical gaps faced by DSOs today.

Expected Impact and Benefits

The EO x Grid service is expected to deliver tangible benefits for electricity utilities. More accurate PV forecasts improve the anticipation of grid imbalances and support more stable grid operation. Better planning and operational awareness can reduce balancing and procurement costs and allow more efficient use of local flexibility resources.

By improving predictability, the service supports higher integration of PV into existing grids and reduces the need for costly monitoring infrastructure. Environmental benefits include improved utilisation of renewable generation and reduced reliance on fossil-based balancing resources, contributing to lower greenhouse gas emissions.

Next Steps

The feasibility study represents the first step towards a fully operational EO x Grid service. Following successful completion, the next phase foresees a demonstration project to validate the solution in real-world operational environments. This would be followed by commercial roll-out to electricity utilities, supported by partnerships with smart-grid software providers to ensure scalability and market uptake.

Project Partners

The EO x Grid consortium brings together complementary expertise across AI, power systems research, and grid operation.
Abelium d.o.o. is the lead organisation, contributing expertise in AI-driven data analytics, PV forecasting, and project management.

The Institute for Innovation and Development of the University of Ljubljana (IRI UL) provides research expertise in grid-level forecasting and energy systems analysis.

Elektro Primorska d.d., a Slovenian distribution system operator, contributes operational knowledge, real-world data, and user requirements, ensuring the relevance of the solution for utility applications.