中国的过剩电力问题

四角池

6 t( G) w; D+ l( R7 m7 N7 C7 ?- N

这是两个外国研究者之间的辩论, 其中一个可能是Michael Pettis理论的拥护者, David Fishman在中国田野调查了很多年,是上海农村发展爱好者应该已经有13年了.

其实这也是针对一个硅谷科学家考察拉中国电力以后的看法,就是美国已经不可救药的落后了

Quick answer: This is wrong. China currently has much more dispatchable capacity (MW) than peak load AND much more capability to generate power (MWh) than is consumed by customers, both by large margins.

快速回答：这是错误的。中国目前的可调度容量（兆瓦）远高于峰值负荷，发电能力（兆瓦时）也远超用户用电量，两者都有很大余量。

Installed Capacity vs. Peak Load: 装机容量与峰值负荷：

At the end of 1H 2025, China had roughly 2071 GW of dispatchable capacity (thermal, hydro, nuclear, batteries).2025年上半年末，中国约有2071吉瓦的可调度装机容量（火电、水电、核电、电池储能）。

Meanwhile, national peak load on 17 July 2025 hit a new high of 1508 GW.与此同时，2025年7月17日，全国高峰负荷达到1508吉瓦的新高。

Thus, reserve capacity margin during the recent annual demand peak was ~37%. That's dispatchable capacity...not counting intermittent generators.因此，在最近的年度需求高峰期，备用容量裕度约为37%。这是可调度容量……不包括间歇性发电机。

Of course, in reality, the demand peak arrived at a period when solar generators WERE available (afternoon on a sunny, hot, midsummer day). China now has over 1000 GW of solar (and 650 GW of wind).当然，在现实中，需求高峰出现在太阳能发电机可用的时段（阳光明媚、炎热的仲夏午后）。中国目前拥有超过1000吉瓦的太阳能装机容量（以及650吉瓦的风能装机容量）。

So some of them were performing at this time, but it's a good conservative practice to not include them in a reserve margin estimation.所以当时其中一些正在运行，但保守起见，在储备裕度估算中不将它们纳入是个不错的做法。

Most of the time, load is WELL below 1508 GW and effective reserve margins are higher.大多数时候，负荷远低于1508吉瓦，有效备用率更高。

In mild evenings of shoulder seasons with no particular heating or cooling demand, reserve margins are at their highest, probably over 100%(?) But in the context of considering a data center that needs to operate 24/7/365, we might as well look at the annual peak as our limiter.在过渡季节气候温和的夜晚，没有特定的供暖或制冷需求，备用裕度处于最高水平，可能超过100%（？）。但考虑到数据中心需要全年无休、每周7天、每天24小时运行，我们不妨将年度峰值作为限制因素。

Generation vs. Consumption and Surplus:生成与消耗及盈余：

Generation and consumption will be balanced, yes (after accounting for plent self-use and line losses). But if the generators COULD generate more, then that means you have a lot of unused potential. This is the case in China.发电量和用电量将会达到平衡，确实如此（在考虑到大量的自用和线路损耗之后）。但如果发电机能够发出更多的电，那就意味着存在大量未被利用的潜力。中国就是这种情况。

In 2024, the average annual operating numbers for China's thermal fleet of 1444 GW were just 3442 hours. That works out to an annual capacity factor of ~39% (the average is higher for coal and lower for gas).2024年，中国1444吉瓦火电装机的平均年运行时长仅为3442小时。这意味着年容量系数约为39%（煤炭发电的平均容量系数更高，天然气发电的则更低）。

There's nothing stopping them from generating more, except the power consumption needs of the Chinese economy were already met by other generation sources.没有什么能阻止他们生产更多电力，只是中国经济的电力消耗需求已由其他发电来源满足了。

Frankly, these thermal generators could easily generate 2x their current operating hours and it wouldn't be a problem at all (except for the carbon peaking target of course!).坦率地说，这些热力发电机很容易将当前运行时长翻倍，而且完全不会有问题（当然，碳排放达峰目标除外！）。

If the annual capacity factor of the thermal fleet doubled to a hardly-unreasonable 78%, it would yield an additional 4,970 TWh annually. That's more than the entire annual consumption of the USA.如果火电的年容量系数翻倍，达到并非不合理的78%，每年将额外产生4970太瓦时的电量。这比美国全年的总耗电量还要多。

Yes, China is squatting on an entire USA's worth of untapped power generation solely within the unused capacity of its current thermal fleet.是的，仅利用当前火电装机的闲置容量，中国就拥有相当于整个美国尚未开发的发电潜力。

Additionally the curtailment of wind and solar means there are times throughout the year when these generators are producing power that the grid can't accept - thus it is wasted.此外，对风能和太阳能发电的限制意味着，一年中有时这些发电机所发的电电网无法接纳，从而造成浪费。

National curtailment rates were around 3% last year, but much higher in individual provinces (and they've been rising in 2025).去年全国限电率约为3%，但个别省份的限电率要高得多（并且2025年以来一直在上升）。What About on the Provincial Level?省级层面的情况如何呢？

Despite the national oversupply of both capacity and generation potential, individual provinces might still struggly with local undersupply.尽管全国范围内发电能力和发电潜力均供过于求，但个别省份仍可能面临局部供应不足的问题。

Some provinces use way more power than they can produce locally and are reliant on cross-provincial imports from power-abundant provinces to make up the difference.一些省份的用电量远远超过其本地发电量，因此依赖从电力充裕的省份进行跨省输电来弥补缺口。

But the grid doesn't connect to everywhere at once (physical limitations) nor are the market mechanisms necessarily set up yet to facilitate easy cross-region power trade (policy limitations).但电网无法一次性连接到所有地方（物理限制），而且市场机制也未必已经建立，以促进便捷的跨区域电力交易（政策限制）。

This can create issues that can only be fixed long-term by more investments in the grid and deepening of power market reform.这可能会引发一些问题，而这些问题只有通过对电网进行更多投资以及深化电力市场改革才能得到长期解决。

For example, in 2022 Sichuan's power supply was revealed to be quite vulnerable to droughts because it relied so heavily on local hydropower and hardly had any import capacity. When the local hydropower failed, the power went out.例如，2022年，四川的电力供应被证明极易受到干旱的影响，因为该地区严重依赖本地水电，且几乎没有任何电力输入能力。当本地水电供应中断时，就会停电。

Other provinces that use or import hydropower were also affected. In that case, it didn't really matter that there was surplus generation capacity in Inner Mongolia; it couldn't be transported to Sichuan or Chongqing.其他使用或输入水电的省份也受到了影响。在这种情况下，内蒙古有多余的发电能力其实无关紧要，因为电力无法输送到四川或重庆。

[Actually, even Sichuan's own power couldn't be sold in Sichuan. Some of its giant hydropower facilities were export-only and couldn't really connect to the local grid anyway.][实际上，就连四川本地的电力都无法在四川销售。当地一些大型水电设施仅供外销，而且实际上也无法接入当地电网。]

So, if you are looking to site an energy-intensive industrial facility in China, you might now prioritize building in a region with an abundant non-intermittent capacity base like Inner Mongolia.所以，如果你想在中国选址建设一个能源密集型工业设施，你现在可能会优先考虑在像内蒙古这样拥有丰富“非间歇性”产能基础的地区建厂。

Sichuan used to be viewed as a province with abundant, stable power supply, because hydropower was very cheap and considered minimally variable.四川曾被视为一个电力供应充足且稳定的省份，因为水电价格低廉，且被认为波动极小。

With the higher risk of droughts, Sichuan has lost some attractiveness for steel or aluminum producers (or data center operators, for that matter).由于干旱风险增加，四川对钢铁或铝生产商（就此而言，还有数据中心运营商）的吸引力有所下降。

If you previously built a facility in Sichuan, or Hubei, or another hydropower-dependent region, you might scoff at the idea that China is oversupplied with abundant power, because you've personally faced hydropower-related disruptions.如果你之前在四川、湖北或其他依赖水电的地区建设过设施，你可能会对中国电力供应过剩的说法嗤之以鼻，因为你亲身经历过与水电相关的电力中断。

These two things are true at the same time: A. China is nationally oversuppled with power, which distributes to most provinces on a local level too, but B.以下两件事同时成立：A. 中国全国电力供应过剩，在地方层面上，大多数省份也如此，但B.

China's hydropower-dependent regions face increasing uncertainty because of how droughts affect their capacity base (and are thus now building a ton of backup power and new grid interconnects).由于干旱影响水电产能基础，中国依赖水电的地区面临着越来越大的不确定性（因此，这些地区正在大量建设备用电源并新建电网互联设施）。But What Happened in 2021? 2021年发生了什么？

This is ancient history at this point, but the power shortages in Q3 2021 were not caused by a lack of power generation capacity - but by expensive coal and inflexible market mechanisms.到目前为止，这已经是陈年旧事了，但2021年第三季度的电力短缺并非由发电能力不足导致，而是由煤炭价格高昂和市场机制缺乏灵活性所致。

At the time, coal-fired power generators had to absorb high fuels costs from the mostly-liberalized coal market, but were restricted in how much money they could recoup by raising power rates.After many months of losses, generators literally ran out of money to buy coal, were sick of generating at a loss, and shut down their plants. Pandemonium ensued and industrial power supply was curtailed in 20+ provinces.The NDRC responded by increasing the supply of coal (directing state miners to dig more coal) and liberalizing the price of power to make it more market-driven (essentially, allowing power prices to rise). We wrote about this extensively at the time.当时，燃煤发电企业不得不承受煤炭市场基本放开后高昂的燃料成本，但提高电价所能弥补的成本却受到限制。在连续数月亏损后，发电企业确实没钱买煤了，也不愿再亏损发电，于是关停了发电厂。随后一片混乱，20多个省份实施了工业限电。国家发改委的应对措施是增加煤炭供应（指示国有煤矿增加煤炭开采量），并放开电价，使其更具市场导向（本质上就是允许电价上涨）。当时我们对此进行了大量报道。

On a personal note, this was one of the first incidents that started earning me broader publicity to talk about Chinese power markets in international media. An unfortunate incident that was quite personally beneficial for my career development.就我个人而言，这是最早让我在国际媒体上，因谈论中国电力市场而获得更广泛关注的事件之一。这是一件不幸的事，但对我的职业发展却颇为有益。

I say this is ancient history because at this point, the power markets are much more deregulated, allowing cost-passthrough much more smoothly. I'd be shocked if we ever had a market failure via that mechanism ever again.But it was an instructive and fascinating episode about the perils of overly-cautious deregulation...it wouldn't have happened under the fully-regulated policy regime, nor would it have happened under a fully market-based regime. It was the hybrid system that messed things up.我说这都是老皇历了，因为到了现在，电力市场的管制宽松了许多，成本传导也顺畅得多。如果这种机制再次导致市场失灵，我会很震惊。但这是一段很有启发性且引人入胜的小插曲，它反映出过度谨慎地放松管制会带来的风险……这种情况在完全监管的政策体制下不会发生，在完全以市场为基础的体制下也不会发生。正是这种混合体制把事情搞砸了。What Does This Mean for Data Centers in China?这对中国的数据中心意味着什么？

Electricity supply-- whether in MW terms or in MWh terms -- is not a barrier to scaling up data centers in China.There's a LOT of reserve margin -- much more than would be considered economically efficient in more market-driven power systems like the US -- AND a lot of space to operate existing assets more to meet rising consumption...even before considering the huge ongoing capacity buildout.This is why @ruima recently wrote that data center operators in China consider power to be a "solved" problem.无论是以兆瓦（MW）还是兆瓦时（MWh）来衡量，电力供应都不是中国数据中心扩大规模的障碍。中国有大量的备用余量，比美国等更市场化的电力系统中认为的经济高效水平要多得多，而且在不考虑正在进行的大规模产能建设的情况下，通过进一步运营现有资产也有很大空间来满足不断增长的电力消耗。这就是为什么 @ruima 最近写道，中国的数据中心运营商认为电力问题是一个“已解决”的问题。

More broadly, this highlights an interesting moment in the overcapacity and capital efficiency debate. Many power markets in the world would consider a 15-25% reserve margin to be sufficient -- and 37% to be unncessarily high.And no power plant operator hopes to build a power plant and use it just 39% of the time...Market purists would surely consider China's develoment of power assets in this way to be excessive and wasteful deployment of capital.更广泛地说，这凸显了产能过剩与资本效率辩论中一个有趣的节点。全球许多电力市场认为，15%至25%的备用容量率已足够，37%则高得没必要。而且，没有一家发电厂运营商希望建好一座发电厂，结果其利用率仅39%……市场纯粹主义者肯定会认为，中国以这种方式发展电力资产，属于过度且浪费的资本配置。

Yes, domestic power demand was growing rapidly, even before the AI boom, and this surplus infrastructure probably could have been absorbed within a few years, but you could argue that means they were built a few years too early.In the interim, that capital could have been used for something else, right? Wouldn't it have been more capital-efficient to build just enough, just in time?是的，即使在人工智能热潮之前，国内电力需求也在迅速增长，这些过剩的基础设施可能在几年内就会被消化掉，但你也可以说这意味着它们建得早了几年。在此期间，这些资金本可以用于其他方面，对吧？恰到好处、及时地建设足够的设施，难道不是更符合资本效率吗？

And yet. 然而。

In 2025, power systems built to maximize capital efficiency pre-boom by building just enough just in time now find themselves scrambling and straining. Sure, they'll probably figure out a solution as well. Markets are good at that.But it'll take some precious time, and it's going to a strain.2025年，那些在繁荣前为实现资本效率最大化而采用刚好及时建造足够设施的电力系统，如今发现自己正手忙脚乱、不堪重负。当然，他们或许也能想出解决办法。市场在这方面很擅长。但这需要一些宝贵的时间，而且会带来压力。

Meanwhile, the country that overbuilt its assets and deployed its capital 'inefficiently' will be rewarded by being well positioned to rapidly absorb this unforecasted power demand spike.China's power system will easily stretch (and already is now) to accomodate the AI data center boom.与此同时，那些过度建设资产、“低效”配置资本的国家，将因具备快速应对这一意外电力需求激增的良好条件而得到回报。中国的电力系统能够轻松应对（而且现在已经在应对）人工智能数据中心的蓬勃发展。

Luck? All according to plan? Or the just rewards that come to the prepared...?