Haibo Sichuang: Thinking and Layout of Computing and Power Collaboration

"The end of AI is computing power, and the end of computing power is electricity"-this widely circulated judgment in the scientific and technological circles is moving from industry consensus to industry reality.

In 2026, the Report on the Work of the Government first incorporated the coordination of computing and electricity into the new infrastructure, and the "15th Five-Year Plan" also made it clear that "vigorously developing the deep integration of

From the perspective of electricity load, the computing center will be the fastest growing area in the next five years.". It is estimated that by the end of the 15th Five-Year Plan, the electricity consumption of the computing power center will account for more than 8% of the total electricity consumption of the whole society. It is estimated that in 2030, the power load of the data center will reach 1.

"Once the killer C-end or B-end applications appear, the power consumption will increase exponentially." Zhang Jianhui judged that in the next five years, the demand for energy storage in the computing power center scenario is expected to exceed the traditional ". From the perspective of cost tolerance, the electricity cost of the computing power center accounts for a relatively limited proportion of the total investment. Taking 3000P computing power center as an example, the server investment is about 1 billion yuan, and the annual electricity consumption is less than 100 million degrees. Compared with hardware input and computing power output, the marginal fluctuation of electricity price has less impact on the overall operating cost. This means that, compared with independent energy storage, the collaborative scenario can bring more considerable profit space for energy storage power stations.

At present, many projects are called "computing power collaboration" cases, but in essence, they are only building computing power centers in green power-rich areas and using low-cost green power, not computing power collaboration in the strict sense.

Real computing power synergy is based on an important premise: part of the computing power demand can be delayed to meet. Large model training, scientific computing, video rendering, data mining and other tasks have the characteristics of schedulability. Based on this characteristic, the computing power center can operate at full load during the low period of electricity price, and reduce or stagger the load during the peak period of electricity price, so as to realize the flexible dispatch of computing power load with electricity price.

On the other hand, as a "time porter", energy storage can store low-cost electricity such as wind power and photovoltaic power, release it during the peak period of computing power, and convert low-cost green power into high-value computing power output. Computing power and electric power thus realize the dual optimal allocation of time and space.

Zhang Jianhui said that the real coordination of computing power should realize the dynamic dispatch of power demand and the flexible response of computing power to power supply, and realize the two-way coordination of "computing with power, computing with power" through pre-judgment and self-regulation.

In the future, relying on the flexible regulation function of energy storage system, the value transformation of electric energy and computing power can be realized. China's green power enrichment area is expected to become a "computing power output base" and an important engine of economic growth.

In the future, energy storage will no longer be limited to dealing with sudden power outages

Power generation side distribution and storage takes green power direct connection and energy storage configuration as the core, which can meet 80% -90% of the green power consumption demand of the computing power center, while reducing the comprehensive energy cost.

Computer room side distribution and storage focuses on the stringent requirements of high-end computing power for power quality. Especially in the computing power scenario for AI reasoning, its power changes quickly and fluctuates greatly, which puts forward higher standards for the continuity and stability of power supply. The ultimate goal of

computing power synergy is to build an integrated and coordinated development model of "new energy + energy storage + data center". Energy storage can achieve comprehensive improvement in green transformation, economy and power supply reliability by locking in the abandoned electricity and spot low-cost electricity of wind and solar new energy.

AI's reverse enabling

computing synergy is not the one-way support of energy storage to computing power, on the contrary, computing power is also enabling the energy storage industry.

At present, Haibo Sichuang has applied AI technology to the whole life cycle of energy storage power plants, such as project planning, product design, safety warning, fault prediction, operation and maintenance, and power trading. In the operation and maintenance stage, AI can generate multi-dimensional performance evaluation in real time, realize automatic fault warning, automatically generate solutions, guide the site to solve problems accurately and efficiently, and protect the safety of power plants.

Electricity trading is another important scenario for AI's deep empowerment. Relying on massive operation data, the system needs to predict the trend of node electricity price every day, and make decisions by considering multiple variables such as new energy output, meteorology, power grid congestion, market quotation strategy and so on.

Traditional models have been difficult to adapt to complex decision-making needs, through AI continuous iterative deduction, can significantly improve the accuracy of trading strategies and the level of power plant revenue. "AI-driven power trading capability has become one of the core competitiveness of the company," said Zhang Jianhui.