AI Power System is a new kind of power infrastructure in AI era, integrating energy system and AI system. Designed to supply energy-intensive industrial facilities, like mining, steel, green molecule etc. and industrial parks with reliable, locally generated renewable electricity, the system combines wind power, solar power, battery energy storage, power grid, power electronics, large models and industrial loads such as electrolyzers, computing power, heat and cooling systems, or production equipment. The system works by forecasting renewable generation, optimizing storage dispatch, managing flexible loads and maintaining voltage and frequency within defined operating limits.

Energy-intensive industrial facilities and Industrial parks require large volumes of dependable power, while higher shares of wind and solar generation require faster flexibility, stronger controls and better coordination among generation, storage, demand and the grid. By treating industrial demand as an active part of the power system rather than a passive consumer of electricity, the model can mitigate grid constraints, increase local utilization of renewable energy and provide a replicable pathway for low-carbon industrial production across sectors such as chemicals, steel, fuels, data centers and advanced manufacturing.

Main Outcomes of the Solution

• Economy & Flexibility: Leveraging deep insights into physical AI, it improves secure operation of renewable-heavy industrial power systems through AI modelling, forecasting, automated dispatch, battery energy storage and flexible demand to balance variable generation with continuous industrial loads, which helps lower the green electricity price. In the regions rich in wind and solar resources, the green electricity costs are even lower than industrial and commercial power tariffs.
• System Resilience: Strengthens local power system resilience by integrating grid-forming storage, power-electronics-based controls and independent operating capability, enabling industrial parks maintain voltage and frequency stability and provide black-start capability during grid instability or supply disruption.
• Sustainability: Supports industrial decarbonization by enabling direct renewable power supply for energy-intensive production, reducing reliance on fossil-based electricity and facilitating low-carbon fuels, electrification and cleaner industrial clusters.

Key Deployment and Scaling Factors

1. Integrate planning across renewable generation, storage, power electronics, industrial demand, digital controls, site-level energy and carbon management systems, and the grid.
2. Ensure high-quality weather forecasting, power generation, load and asset data, supported by interoperable systems, reliable telemetry and communications coverage.
3. Deploy accurate forecasting, automated and integrated generation-grid-load-storage dispatch, flexible loads and battery energy storage to manage wind and solar variability.
4. Use grid-forming storage and power-electronics-based controls to maintain voltage and frequency stability, oscillation control and enable black-start capability in independent renewable systems.
5. Establish robust cybersecurity, data governance and AI model governance, especially where integrated generation-grid-load-storage dispatch affects critical industrial processes.
6. Apply clear standards, certification, safety regulations and operating procedures where the system supports hydrogen, ammonia or other clean fuel production.
7. Enable commercial scaling through anchor industrial demand, long-term offtake, transparent pricing, cost recovery mechanisms and clear rules for direct renewable power supply.

Case study

Envision (2025), China

In 2025, Envision commissioned the first commercial phase of its green hydrogen and ammonia facility at the Chifeng Net Zero Industrial Park in Inner Mongolia, China. Public sources describe the first phase of the facility as operational, powered by an off-grid wind and solar-based energy system and capable of producing 320,000 tonnes of green ammonia per year. The wider project has been reported as targeting around 1.5–1.52 million tonnes per year by 2028. In 2026, in cooperation with Tencent, Envision has built the Computing-Power Synergy project at the Chifeng Park, enabling real-time dynamic coordination among wind and solar power, energy storage, computing loads and hydrogen production.

According to Envision, the project realizes 100% green power direct supply through one of the first 2GW-class independent AI power system composed of 100% renewable energy and 100% power electronic equipment, achieving dynamic coupling of wind-solar-storage and hydrogen-ammonia-alcohol production. The system uses Envision Tianji, the Weather Foundation Model, based on physical AI, together with Envision Dubhe, the Energy Foundation Model, to forecast wind and solar output, dispatch battery energy storage, coordinate electrolyzers and ammonia synthesis and support real-time control. Envision reports that AI forecasting improves power prediction accuracy by about 10%, while voltage stability, transient analysis and dispatch boundary identification can be completed within milliseconds. These performance claims should be treated as company-reported rather than independently verified.

The project integrates advanced wind turbines, grid-forming battery energy storage systems, predictive meteorological modeling and an integrated dispatch system based on energy foundation model to balance variable renewable generation with electrolyzer, ammonia synthesis and AI-computing energy demand under islanded operation. Grid-forming AI-enabled energy storage supports voltage control, oscillation suppression and station-level black start capability. The project demonstrates how Physical AI can help build future energy systems, coordinating variable renewable generation with continuous industrial demand at industrial scale.

References:

1. Information provided to World Economic Forum by Envision (May 2026)