Vol.183 Industry Watch 33 | The Power Struggle in the AI Era: Reframing China’s New Power System
Summary
- China’s power demand and new-energy penetration have reached a scale that the traditional dispatch system can no longer absorb with ease. National monthly electricity consumption topped 1 trillion kWh for the first time this July, equivalent to Japan’s annual consumption or that of France and Germany combined; new-energy vehicles exceeded 50% of sales, while new-energy capacity approached 50% of total installed capacity and generation surpassed 20%. At the same time, 李翔 noted that oil imports and coal-fired power’s share of generation were “flat or down” year on year, signaling a qualitative shift in the energy mix.
- The industry’s bottleneck has shifted from whether more new energy can be generated to whether volatile power can be absorbed steadily, locally, and economically. As 张海瑞 summarized it, solar has moved from guaranteed grid access, to self-consumption with surplus fed into the grid, and now to “largely not being allowed or unable to connect to the grid.” After Document No. 136 eliminated mandatory storage and accelerated marketization, new-energy pricing became “levelized cost plus control and management plus market supply and demand,” shifting industry value beyond standalone solar-and-storage equipment toward microgrids, smart distribution, real-time communications, and energy operations.
- China’s grid has prioritized safety and reliability over pure economics, creating major cross-regional dispatch capacity. Resources in the west and loads in the east are distributed in opposite directions, giving rise to a massive grid built around ultra-high-voltage transmission; 张海瑞 said that over the past 10 to 20 years, “safety and reliability had to take precedence over economic efficiency.” Overseas systems place more emphasis on balancing economics with security, and under extreme conditions, a small fault that is mishandled can escalate into a nationwide incident like Spain’s blackout.
- Negative power prices are making visible some of the balancing costs that new energy previously imposed on the grid. Solar output peaks around midday just as factories cut production, and the resulting imbalance has already produced negative prices in parts of China’s spot markets. Price as the “baton” will force generators, storage operators, and end users to adjust together. Pure equipment investors face curtailment pressure and falling orders, while system providers that integrate solar, storage, loads, and trading are seeing a new opening.
- The core technology shift in the new power system is from “low-frequency, large-granularity” control to “high-frequency, massive small nodes.” Thermal and hydro systems can be centrally dispatched around large plants and grid nodes, while power-electronic solar, storage, and charging loads can change instantly with a cloud or a wave of vehicles. 张海瑞 argues that high-frequency data collection, system-state assessment, and real-time closed-loop control are essential; 博联智电’s control cycle and response time have reached the millisecond level—“this is not simply adding storage.”
- Microgrids provide the local physical loop, while virtual power plants aggregate nodes across the system and participate in trading—but the infrastructure is still incomplete. A virtual power plant is like “ride-hailing for electricity,” aggregating distributed solar, storage, and loads into a virtual entity with generation and balancing capabilities. Commercial deployment still depends on medium- and low-voltage grid simulation, monitoring, automatic control, communications, settlement, and profit allocation; markets below 220 kV remain particularly immature.
- AI and power reinforce each other, creating two parallel investment needs. Compute centers require massive volumes of green power, with reliability and power-quality requirements reportedly “several levels higher” than those of residential users. In the other direction, the vast number of small and medium-sized power systems can no longer be served by traditional experts and the bulk grid alone; AI must become a “virtual power expert.” The opportunity therefore spans both high-quality power infrastructure and specialized AI control layers for electricity.
- Power exports divide into two radically different markets, and China’s supply chain can address both ends. Africa and Southeast Asia need to solve the reliability problem of “having power sometimes and not having it at other times,” while Chinese manufacturers moving overseas will bring power capacity with them. Europe, Australia, and the United States, by contrast, need green power plus storage to reduce high electricity costs. By 张海瑞’s estimate, the average Chinese household spends about RMB1,000 a year on electricity, compared with more than RMB2,000–3,000 in Australia and even more in Europe.
Deep dive
1. Trillion-kWh demand collides with instantaneous balancing, reviving the value of grid services
李翔 opened by placing energy in a much longer cycle: “All these technological advances and changes in the country ultimately come down to changes and progress in energy.” The metric he uses is the amount of energy available per person, especially the steady increase in usable solar energy.
The structural shift is already visible in concrete figures: national monthly electricity consumption topped 1 trillion kWh for the first time this July; new-energy vehicles exceeded 50% sales penetration, new-energy capacity approached 50% of installed capacity, and new-energy generation surpassed 20%. Construction has also begun on the Yarlung Zangbo River hydropower station.
李翔 highlighted the opposite movement: even as electricity consumption continues to expand, oil imports and coal-fired power’s share of generation were “flat or down” year on year. China has begun promoting and implementing a sizable medium-term energy-storage program, serving the volatility of the wind-and-solar supply chain as well as energy security and long-term structural adjustment.
刘鹏琦 compared the grid with an e-commerce platform: it matches supply and demand while also providing logistics-like transmission and services. 张海瑞 accepted the analogy but stressed that electricity moves from production to consumption “instantaneously”; generation must broadly match usage in real time.
2. Electricity is not an ordinary commodity: balancing and fault isolation cannot stop for a moment
张海瑞 breaks the system into three parts: power plants, the grid, and users. Plants convert energy; the grid handles transmission, transformation, and reliability; households and businesses consume the final product. Any mismatch in between must be balanced in real time through pumped hydro, primary or secondary frequency regulation, load adjustment, and other mechanisms.
Electricity travels through electromagnetic waves, making it fast, highly dynamic, and difficult to control. Monitoring, protection, and control are therefore not ancillary services. As 张海瑞 put it: “Once the entire power system is running, regulation and control are taking place without interruption.”
The other hard boundary is to contain faults within the smallest possible area. If a fault on the user side is not identified and isolated promptly, it can spread to other users and ultimately the entire system. Protection systems must stop the fault from propagating while returning the affected equipment to service as quickly as possible.
3. China uses ultra-high voltage and reliability to offset its reverse resource distribution
张海瑞 sees one of the biggest differences between China and other countries in the fact that China has “one strong high-voltage grid.” In the construction choices of the past 10 to 20 years, “safety and reliability had to take precedence over economic efficiency,” producing one of the world’s leading large-scale high-voltage networks.
This was not simply a matter of institutional preference. Shanxi, Inner Mongolia, and western China have abundant thermal, new-energy, and hydropower resources, while major electricity consumers have long been concentrated along the eastern seaboard. A super-large grid built around ultra-high-voltage transmission was designed for long-distance delivery and unified management.
刘鹏琦 asked whether major overseas blackouts meant grids were inherently less stable. 张海瑞 did not attribute them simplistically to new energy, pointing instead to differences in each country’s energy-transition history, resource endowment, and trade-off between security and economics. The Paris Agreement and China’s targets of peaking carbon emissions by 2030 and reaching carbon neutrality by 2060 are both pushing this structural transition.
Spain’s blackout is 张海瑞’s warning case. An initial power-system fault does not inevitably cause a nationwide outage; the decisive factor is that “after the small fault occurred, it was not handled promptly, or was even handled improperly.” The failure then intensified until the system entered an irreversible state.
4. Market-based power prices spread generation, transmission, distribution, and reliability costs back across the system
Under cost-based pricing, the price of a kilowatt-hour includes generation, transmission and transformation, grid balancing and protection, reliability assurance, and terminal services. Market pricing answers directly who can generate and who can consume: “Whoever bids the highest can use this electricity.” When supply is ample, users prioritize the cheapest available power.
When solar output peaks around midday while factories suspend production, surplus electricity creates a supply-demand imbalance and the spot market may turn negative. Multiple locations in China have seen this over the past 2 years, exposing the tension between cheap new-energy generation and expensive system balancing.
China’s reform did not start suddenly. Market mechanisms began to be introduced in the 2000s, followed by the 2015 No. 9 document on power reform and reforms to power sales and distribution. Trading among large users, large plants, and nodes above 220 kV is now relatively marketized, while price transmission, competition, and trading mechanisms for small and medium-sized industrial and commercial users below 220 kV remain underdeveloped.
Households have not been excluded from the market; grid companies simply act as their primary agents in power purchases and sales. Medium- and long-term contracts and day-ahead trading are relatively well established, while spot markets operate in only some provinces and are still being adjusted. Overseas markets may be more open, but prices are also more prone to sharp swings with supply and demand.
5. China’s power build-out has moved from building grids around generation to building zero-carbon systems around load
张海瑞 defines the period around 2000–2010 as the era of “building high-voltage grids around power sources.” Hydropower from Three Gorges and large thermal plants in Shanxi and Inner Mongolia were built first, then power was transmitted east. The immediate priority was to relieve widespread electricity shortages.
Around 2010–2020, the focus shifted to “building large wind-and-solar bases around the grid.” Once supply tightness eased, centralized wind and solar expanded rapidly in western China. The generation mix began shifting beyond thermal, hydro, and nuclear power toward new energy, which was incorporated into the system through the grid.
After 2020, the model became “building integrated source-grid-load-storage systems around load.” Users are not only consuming more electricity; they are expected to use more green power, provide balancing capacity, and gradually move toward zero-carbon operations. New-energy vehicles turning gasoline demand into charging demand are the clearest example of this change on the consumption side.
The objective does not discard the traditional requirements. Users still demand “safety and reliability, economic efficiency,” 张海瑞 said; the system now adds “green and low-carbon.” The challenge is to make generation, control and management, and consumption more flexible and economical, while ensuring that the products ultimately produced are genuinely made with green power.
6. Solar has moved from guaranteed grid access to grid exclusion, making absorption the primary constraint
Solar’s first stage ran roughly from 2010 to 2020: everything generated could be fully connected to the grid and receive subsidies, while the grid absorbed some of the cost of taking in the new energy. In the second stage, systems shifted to self-consumption with surplus fed into the grid; rooftops, fishery-solar hybrids, and agriculture-solar hybrids first served local loads.
The third stage, in 张海瑞’s summary, has now largely arrived: solar is “largely not allowed or unable to connect to the grid.” As the grid’s balancing headroom narrows, projects may install anti-backflow devices, while electricity fed into the grid may face negative prices. New energy increasingly has to be absorbed locally first.
Microgrids are the physical solution. Within a relatively small area, they integrate generation, dispatch, and power distribution and consumption. They can supplement the bulk grid, or serve as the primary power system in remote areas and on islands, creating a local source-grid-load-storage loop.
A virtual power plant is the organizational and trading layer. It aggregates distributed solar, storage, and large loads into a virtual entity with generation and balancing capabilities. 刘鹏琦 compared it first with ride-hailing and then with cloud-computing virtualization; 张海瑞 added that electricity remains tied to the physical world, so control, communications, settlement, and profit allocation must all be solved together.
7. Document No. 136 removes mandatory storage while pushing new energy to reprice itself through market supply and demand
Extreme weather and the Russia-Ukraine conflict caused power shortages and price spikes in some regions, while China continued to advance power-market reforms and new-energy-storage policy. Represented by Document No. 136 in the first half of this year, the measures were followed by provincial implementation. 张海瑞 sees them as a clear signal that new energy will continue to advance.
Before Document No. 136, local governments generally required new-energy projects to install storage, hoping storage would act as a stabilizer for volatility. The document both removes mandatory storage and accelerates marketization. Solar is no longer judged only by its levelized cost; control and management costs and actual supply-and-demand prices must also be included.
The industry is therefore moving from “standalone investment and construction into energy management and operations,” and from solar grid connection toward solar-plus-storage power supply. Solar and storage companies focused solely on equipment production and investment are directly exposed; absorption difficulties may reduce orders. New demand is shifting toward energy management, smart distribution, real-time communications, AI, and virtual power experts.
刘鹏琦’s summary is that technology and pricing must be reshaped together: storage and source-grid-load-storage systems reduce physical volatility, while market trading encourages generators and users to adjust proactively. 张海瑞 agreed that price is the “baton,” but warned that China’s spot-market mechanism is not yet fully mature.
8. The new power system pushes control toward high frequency and massive numbers of small nodes
张海瑞’s central technology judgment is that conventional electromechanical systems such as thermal and hydro power are giving way to power-electronic systems represented by new energy, objectively creating a shift “from low frequency to high frequency.” Legacy low-frequency monitoring and control cannot fully match the faster fluctuations of power electronics.
刘鹏琦 used two instantaneous changes to illustrate the problem: a passing cloud can alter solar output, while a wave of vehicles entering a charging station can lift the load. Strategies set a day ahead or by the hour could once maintain rough balance; the system now needs higher-frequency monitoring, control, and real-time closed loops.
The change is not only about speed, but also “from large granularity to small granularity.” In the past, it was enough to dispatch large plants, large nodes, and large users. Now even a household with solar, storage, and loads needs to be controlled, while traditional power experts and centralized grid services cannot cover the massive number of new demand points.
The chain must therefore run from data collection, monitoring, and system-state assessment through generation, consumption, and charging control. 张海瑞 stresses that the endpoint is not merely seeing the data, but “how to form an effective closed loop” for flexible, real-time allocation of power.
9. AI creates the highest-grade new load—and must also become a virtual power expert
AI compute centers consume massive amounts of electricity, first raising the question of where that power will come from. 张海瑞 believes they need more green power and safer, more reliable electricity, with reliability and power-quality requirements at least “several levels higher” than those of residential users.
In the opposite direction, high-speed, high-frequency, fine-grained small and medium-sized power systems also need AI. Vast numbers of industrial parks and users were previously served mainly by the grid and human experts; their scale and response-speed requirements now exceed the capabilities of traditional organizations. Monitoring, analysis, assessment, and control must be AI-enabled.
博联智电 calls this role the “virtual power expert” and says it began related projects 3 years ago. 张海瑞 also defines AI’s boundary: it must be tightly integrated with power engineering and cannot operate independently of physical controls, safety constraints, and real-time data.
10. 博联智电 turns ordinary distribution equipment into millisecond-level control gateways
张海瑞 previously worked on microgrids, new-energy grid integration, research, and product development at Huawei Digital Energy and Huawei’s 2012 Lab. After the 2015 No. 9 document, the team began focusing on the user-side power Internet of Things and saw cloud, network, edge, and terminal technologies mature into commercial applications. By 2022, it concluded that demand had emerged for high-performance real-time power-control systems, which it called PRTC.
The products do not rely on unfamiliar terminals. Instead, they redesign smart meters, circuit breakers, and gateways, analogous to the upgrade “from traditional phones to smartphones”: the form factor and user habits remain familiar, while high-frequency data collection, algorithmic models, industrial communications, virtual buses, and control technologies are added internally.
Data across devices is synchronized, programmable, and coordinated according to logic. The edge performs data collection, computation, analysis, and transmission; upper-layer software manages equipment, data, and customer operations. Terminals, edge systems, and the cloud ultimately form a complete closed loop. The equipment layer and system communications and control have been accelerated and optimized, allowing data to be processed and transmitted rapidly.
The architecture supports both greenfield deployment and retrofits, with existing equipment directly augmented or replaced. A 国家电投 project performs wide-area synchronization and centralized aggregation of distributed solar, while a 上海电气 industrial park uniformly monitors charging, consumption, and generation and carries out online analysis and control.
11. Millisecond-level closed loops first deliver cost savings and efficiency gains, then replicate overseas across two demand profiles
博联智电 says its control cycle and response time have reached the millisecond level, enabling safe optimization and management of source-grid-load-storage systems within 50 Hz AC networks, increasing generation, improving returns, and building microgrid capabilities. 刘鹏琦 believes the system should be imperceptible to generators and users; on a longer time scale, it could create safety problems such as outages and restarts after switching.
Asked by 刘鹏琦 about a long-term revenue-sharing model, 张海瑞 did not provide a specific mechanism. He instead summarized two current sources of value: reducing electricity costs and avoiding waste; and rapidly deploying systems in areas without power or with underdeveloped supply, improving electrification and enabling replication at scale.
Underserved markets are buying reliability. Many parts of Africa and Southeast Asia still “have power sometimes and don’t have it at other times.” Even when Chinese factories move in to take advantage of cheaper labor, unstable electricity can remain a constraint, so generation, grid, and control systems must go overseas “together with manufacturing, and serve it together.”
Europe, Australia, and the United States are buying cost reductions. By 张海瑞’s estimate, the average Chinese household spends about RMB1,000 a year on electricity, Australia more than RMB2,000–3,000, and Europe even more after the Russia-Ukraine conflict. 博联智电 says its technology development, market-access certifications, and sales channels were designed for global markets from the start.