On October 19, at the closed-door technology conference titled "Artificial Intelligence and Future Energy Systems," Chairman Zhang Lei of Envision Group proposed the concept of "Physical Artificial Intelligence," elaborating on the trend of AI evolving from a tool to a decision-making entity in energy systems, and predicting that the competitiveness of future energy companies will shift from the scale of physical assets to the scale of intelligent assets.
Zhang Lei believes that the essential difference between AI technology and previous technological revolutions is that it is no longer just a passive tool, but an entity with self-perception and decision-making capabilities. This marks a transition of AI from "automation" to "autonomy." He compared AI to a "child" that needs to be nurtured, emphasizing that human-machine collaboration will create new possibilities.
Regarding the challenges currently faced by the energy system, Zhang Lei pointed out that as the proportion of renewable energy increases, the complexity and market uncertainty of the power system have significantly increased. Although this complexity puts pressure on traditional management models, it provides an ideal scenario for AI applications. The parallel computing capability of AI can process massive data in real-time, identify hidden patterns, and optimize decisions, thereby addressing the challenges of complex energy systems and power markets.
The core of the "Physical Artificial Intelligence" concept is to deeply integrate AI reasoning capabilities with physical laws and system boundaries, enabling it to operate reliably in real physical environments. This differs from purely data-driven AI models, emphasizing the importance of physical constraints and causal relationships in the model. Zhang Lei stated that China has rich application scenarios and data resources in this field, and has the potential to lead globally.
In terms of technical implementation, Envision Technology has made progress in weather and energy modeling. The "Tianji" weather large model has improved the accuracy of medium- and long-term weather forecasting, providing a foundational support for the reliable operation of renewable energy. The "Tianshu" energy large model can control the power system in real-time, optimize the efficiency of power trading, and promote the development of green energy. The application of these models demonstrates the value of "Physical Artificial Intelligence" in practical scenarios.
Zhang Lei presented a new judgment on the competitive landscape of the future energy industry. He believes that the core competitiveness of energy companies will shift from traditional physical assets such as power generation capacity and the scale of transmission and distribution networks to "artificial intelligence assets" such as the ability and quantity of intelligent models. This transformation means that the logic of value creation in the energy industry is undergoing a fundamental change—from capital-intensive to technology-intensive.
From an industry impact perspective, this view has reference significance for the strategic planning of energy enterprises. If AI assets become the core competitiveness, companies need to make long-term investments in data accumulation, algorithm development, and scenario applications. However, it should be noted that the physical infrastructure of the energy system remains indispensable; "intelligent assets" are more likely to empower and add value to physical assets, rather than completely replace them.
From a technical implementation perspective, the concept of "Physical Artificial Intelligence" emphasizes the deep integration of domain knowledge and AI technology. Energy systems involve multiple physical disciplines such as electromagnetism, thermodynamics, and fluid mechanics, as well as complex engineering constraints and safety boundaries. How to effectively embed these hard constraints into AI models, ensuring that their outputs are both optimized and reliable, is the technical challenge. Envision Technology's practice in weather and energy models provides preliminary verification, but large-scale application still requires solving issues such as model generalization, real-time performance, and interpretability.
From a development perspective, the intermittency and volatility of renewable energy indeed create demand for AI applications. The uncertainty of wind and solar power generation requires more intelligent prediction, scheduling, and trading strategies. The reform of power marketization is also increasing system complexity, and traditional scheduling methods struggle to meet the real-time optimization needs of a large number of devices. In this context, AI technology is expected to become the "operating system" of the new power system.
However, the particularity of the energy system also imposes higher requirements on AI applications. Unlike internet applications, the energy system relates to infrastructure security and social stability, and AI decision-making errors could lead to serious consequences. Therefore, "Physical Artificial Intelligence" not only needs high accuracy, but also needs to meet engineering standards such as reliability, interpretability, and security. How to establish a regulatory framework and accountability mechanism for AI in the energy sector is also a topic that the industry needs to explore.
