Envision claims AI-enabled storage systems can detect battery safety risks days in advance

Envision Energy has unveiled what it describes as a fully integrated energy storage solution that combines hardware, software, and market-facing capabilities, arguing that deep vertical integration is essential for the next phase of battery deployment as energy systems evolve toward an AI-driven energy system.

The company says its approach goes beyond selling battery systems, instead offering an end-to-end platform spanning cell innovation, power conversion, system controls, artificial intelligence, and electricity trading-connecting physical assets directly with real-time intelligence and market decision-making.

The company’s storage portfolio includes battery cell technology, power conversion systems (PCS), medium-voltage stations, energy management systems, and a top-layer software platform designed to support both operations and market participation. According to Envision, this architecture allows AI to be embedded not just at a supervisory level, but directly into the physical operation of storage assets.

“We’re not offering just a product, but the entire solution,” Kevin Huang, Envision SVP & President of Energy Storage Product Line, told pv magazine at the World Forum Energy Summit (WFES) held in Abu Dhabi, UAE, in early January. “From the cells all the way to the software and route to market, it’s one ecosystem.”

Central to Envision’s strategy is what it calls “physical AI,” a concept the company distinguishes from conventional data-centric artificial intelligence. “Traditional AI operates largely at the application and analytics layer,” said Huang. “Physical AI, by contrast, is designed to interact directly with the constraints of real-world energy systems, including grid stability, equipment safety, and operational limits.”

“Energy systems are governed by physical laws and engineering boundaries, which means AI must be tightly integrated with models of electrical behavior and asset performance,” he wento on to say. “In Envision’s view, this allows AI-driven decisions to be made in real time without compromising safety or reliability.”

This Physical AI architecture is underpinned by Dubhe, Envision’s Energy Foundation Model, unveiled during Abu Dhabi Sustainability Week in January. “Dubhe sits at the core of Envision’s Physical AI system, analysing vast streams of real-world energy data to orchestrate renewable generation, storage, grids, and demand in real time, shaping what Envision defines as the AI Energy System,” Huang added.

Rather than deploying a single overarching AI platform, Envision embeds AI technologies across multiple layers of the storage system. This includes grid-support functions such as frequency and voltage response, operational optimization, and market participation. The company also emphasized that AI agents are used to support trading strategies, allowing batteries to respond dynamically to price signals in increasingly volatile electricity markets.

One of the most significant company’s claims relates to asset health and safety.

“By collecting and analyzing large volumes of laboratory and operational data from batteries, PCS units, and control systems, we can train models to detect early warning signs of failure well before conventional monitoring systems,” Kotub Uddin, Envision BESS Chief Technology Officer (CTO) told pv magazine. “Traditional safety systems, such as gas sensors, typically provide only minutes of warning before a critical event. Our Physical AI-based Systemscan identify subtle electrical signatures days in advance, enabling proactive maintenance rather than emergency shutdowns.”

“If you can predict a fault early enough, you don’t need to impetuously shut the plant down,” Uddin explained. “You can schedule maintenance, replace components, and avoid both safety incidents and lost uptime.”

The company pointed to internal analysis showing that voltage irregularities associated with dendrite growth inside battery cells can be detected through AI models trained on large datasets. While Envision stressed that these systems are still evolving, it said predictive detection is already being used to improve operational reliability.

Envision also highlighted its safety track record, noting that it has not experienced major fire incidents in its energy storage batteries. The company attributes this partly to its ability to capture and analyze data across the full system.

Beyond safety and performance, Envision sees AI as a way to fundamentally change the economics of storage. As power markets mature and margins tighten, optimization becomes increasingly important. According to the company, AI can help asset owners extract additional value by improving price forecasting, reducing operational constraints, and enabling more flexible trading strategies.

“Electricity price volatility, driven by growing shares of variable renewable generation, is a key opportunity for storage,” said Uddin. “Our AI trading tools continuously learn from past market behavior, refining strategies based on outcomes rather than relying solely on static optimization rules.”

“Everyone has schedulers and optimizers,” added Uddin. “What AI adds is learning. It can look back at a trade, ask why it wasn’t optimal, and improve the next decision.”

This learning capability, Envision argues, is increasingly important as traders seek fewer operational restrictions from storage assets. By better predicting degradation and performance impacts, AI can support warranty structures that allow more aggressive cycling without compromising long-term asset health.

While large, grid-scale storage projects remain the dominant segment by capacity, Envision sees strong potential for AI-enabled, grid-forming storage in commercial and industrial (C&I) applications. Behind-the-meter systems allow a higher degree of control, making them particularly suitable for AI-driven grid-forming operation rather than static, pre-configured behaviour.

In C&I projects, for example, AI agents are not focused on optimization alone. They continuously infer the characteristics of the grid they are connected to, such as grid strength, inertia, and effective short-circuit level, and adapt grid-forming behaviour in real time.

“This goes beyond setting grid-forming parameters at build or commissioning,” said Uddin. “The system is continuously asking a much more fundamental question: what grid am I connected to right now, and how should I behave?”

“Behind the meter, you can put AI directly into the control loop,” he also stated “That’s where the benefits become very tangible.”

Envision reported that it already has solar-plus-storage projects operating in China and is developing similar systems in markets such as Chile, Egypt, Spain, and Turkey. While many current battery installations remain standalone, the company expects most future deployments to be co-located with renewable energy generation.

Envision also plans to deploy its AI-enabled storage solutions globally, including in Europe and North America. The company acknowledged that AI is rapidly becoming a standard feature across the industry, but argued that its differentiation lies in developing AI models in-house and applying them end-to-end across the system.

Rather than viewing AI adoption as a zero-sum competition, Envision said it expects widespread uptake across the sector. However, it believes that companies able to generate large datasets, integrate hardware and software, and develop proprietary AI models will have a structural advantage.

“We don’t just use AI; we develop it,” Huang concluded, “And what we’re building goes beyond storage – it’s the foundation of a renewable energy system.”

From pv magazine global