Announcing Two Open Working Papers on Transparent Energy-System Intelligence
Announcing Two Open Working Papers on Transparent Energy-System Intelligence
Wangari Global has released two open working papers on transparent, robust, and privacy-aware forecasting for energy-system decision-making.
Wangari Global has released two open working papers on transparent, robust, and privacy-aware forecasting for energy-system decision-making.
At Wangari Global, we are working to make energy-system intelligence more transparent, robust, and useful in real-world decision-making. On 24 June 2026, we released two open working papers that explore how advanced forecasting can help energy operators plan, manage flexibility, and respond to uncertainty—without treating accuracy as the only measure of value.
The papers address two connected challenges. First, energy systems need forecasts that remain useful when the conditions behind demand, generation, and prices change. Second, distribution system operators need forecasting tools that can support local flexibility while respecting the privacy and sovereignty of underlying smart-meter data.
Causal-Augmented Forecasting for Energy-System Decision Support
Our first paper, Causal-Augmented Forecasting for Energy-System Decision Support, presents an approach that combines time-series forecasting models with structural causal models. The aim is not only to predict what may happen next, but also to help users understand the drivers behind a forecast and consider the likely effect of interventions or changing conditions.
The research applies this approach to local energy systems, including microgrids that combine renewable generation, electrical and thermal storage, and ancillary-services participation. Using 3.5 years of European public energy data, the paper evaluates forecasting performance, calibrated uncertainty estimates, and robustness during the 2022 European energy-price shock. It also shows how forecasts can inform system sizing, predictive control, and ancillary-services decisions through an energy-system digital twin.
A key design principle is deterministic-first AI: numerical analysis and figures are produced through auditable computation, while language models are limited to explanation. A validation guard is used to prevent numerical claims that cannot be traced back to computed results. The accompanying evaluation harness is available openly to support reproducibility.
Read the paper: Causal-Augmented Forecasting for Energy-System Decision Support
Transparent Federated Load Forecasting for DSO Flexibility
Our second paper, Transparent Federated Load Forecasting at the Secondary-Substation Level for DSO Flexibility, examines day-ahead load forecasting for distribution system operators (DSOs). Accurate baselines can help DSOs participate in local flexibility markets and manage flexible connection agreements, but conventional approaches can be difficult to reproduce and may require sensitive smart-meter data to be pooled centrally.
The paper proposes an interpretable, edge-deployable forecaster trained through federated learning across secondary substations. Under this approach, model parameters are shared for learning, while raw consumption data remains at each local node. The work evaluates the method on public UK Power Networks low-voltage feeder data and considers its application to congestion and flexibility scheduling.
The research demonstrates how privacy-preserving federation, transparent modelling, and efficient deployment can be brought together at a practically relevant grid level. The published evaluation harness supports open review and further development.
Read the paper: Transparent Federated Load Forecasting at the Secondary-Substation Level for DSO Flexibility
Why this matters
As electricity systems become more decentralised, operators need tools that can work with changing patterns of renewable generation, demand, market prices, and local network constraints. We believe that trustworthy energy intelligence should be more than a black-box prediction. It should be explainable, appropriately uncertainty-aware, reproducible, and designed with real operational and data-governance requirements in mind.
Together, these two working papers set out Wangari Global’s approach to this challenge: combining rigorous, auditable methods with practical decision support for the energy transition.
Both papers are published openly under a Creative Commons Attribution 4.0 licence. We welcome engagement from distribution system operators, energy innovators, researchers, and collaborators working on transparent and resilient energy systems.
At Wangari Global, we are working to make energy-system intelligence more transparent, robust, and useful in real-world decision-making. On 24 June 2026, we released two open working papers that explore how advanced forecasting can help energy operators plan, manage flexibility, and respond to uncertainty—without treating accuracy as the only measure of value.
The papers address two connected challenges. First, energy systems need forecasts that remain useful when the conditions behind demand, generation, and prices change. Second, distribution system operators need forecasting tools that can support local flexibility while respecting the privacy and sovereignty of underlying smart-meter data.
Causal-Augmented Forecasting for Energy-System Decision Support
Our first paper, Causal-Augmented Forecasting for Energy-System Decision Support, presents an approach that combines time-series forecasting models with structural causal models. The aim is not only to predict what may happen next, but also to help users understand the drivers behind a forecast and consider the likely effect of interventions or changing conditions.
The research applies this approach to local energy systems, including microgrids that combine renewable generation, electrical and thermal storage, and ancillary-services participation. Using 3.5 years of European public energy data, the paper evaluates forecasting performance, calibrated uncertainty estimates, and robustness during the 2022 European energy-price shock. It also shows how forecasts can inform system sizing, predictive control, and ancillary-services decisions through an energy-system digital twin.
A key design principle is deterministic-first AI: numerical analysis and figures are produced through auditable computation, while language models are limited to explanation. A validation guard is used to prevent numerical claims that cannot be traced back to computed results. The accompanying evaluation harness is available openly to support reproducibility.
Read the paper: Causal-Augmented Forecasting for Energy-System Decision Support
Transparent Federated Load Forecasting for DSO Flexibility
Our second paper, Transparent Federated Load Forecasting at the Secondary-Substation Level for DSO Flexibility, examines day-ahead load forecasting for distribution system operators (DSOs). Accurate baselines can help DSOs participate in local flexibility markets and manage flexible connection agreements, but conventional approaches can be difficult to reproduce and may require sensitive smart-meter data to be pooled centrally.
The paper proposes an interpretable, edge-deployable forecaster trained through federated learning across secondary substations. Under this approach, model parameters are shared for learning, while raw consumption data remains at each local node. The work evaluates the method on public UK Power Networks low-voltage feeder data and considers its application to congestion and flexibility scheduling.
The research demonstrates how privacy-preserving federation, transparent modelling, and efficient deployment can be brought together at a practically relevant grid level. The published evaluation harness supports open review and further development.
Read the paper: Transparent Federated Load Forecasting at the Secondary-Substation Level for DSO Flexibility
Why this matters
As electricity systems become more decentralised, operators need tools that can work with changing patterns of renewable generation, demand, market prices, and local network constraints. We believe that trustworthy energy intelligence should be more than a black-box prediction. It should be explainable, appropriately uncertainty-aware, reproducible, and designed with real operational and data-governance requirements in mind.
Together, these two working papers set out Wangari Global’s approach to this challenge: combining rigorous, auditable methods with practical decision support for the energy transition.
Both papers are published openly under a Creative Commons Attribution 4.0 licence. We welcome engagement from distribution system operators, energy innovators, researchers, and collaborators working on transparent and resilient energy systems.