What Lessons Do The Chinese Hydropower Shortages In Summer 2022 Hold For The World?
By Mike Thomas, David Fishman 24 May, 2023
With extreme weather on the rise, is “just enough” power the way to futureproof power systems? Experts Thomas & Fishman from The Lantau Group share their insights from the Yangtze multi-month drought

Sichuan Province, located in southwestern China, is probably most well-known internationally for either its fiery, mouth-numbing cuisine, or its pandas. But, for the Chinese power sector, Sichuan definitely means something else: hydropower.
Arising in western Sichuan atop the Tibetan plateau, numerous major and minor rivers zigzag west-to-east across the province, descending 4000 meters in elevation as they traverse the province before emptying into the Yangtze River on Sichuan’s eastern border. Dozens of large and small hydroelectric dams dot these waterways, altogether totalling some 120 gigawatts of capacity (exceeding the entire generation capacity of all but a few countries).
Sichuan relies on hydro for >80% of its power & also exports its electricity to central & eastern China…
Sichuan relies on hydro for over 80% of its annual power consumption needs, while also exporting significant volumes of electricity to central and eastern China. In recent years, five particularly massive hydro export projects have been constructed along Sichuan rivers, with ultra-high voltage (UHV) power lines connecting those projects to power customers thousands of kilometres to the east. It’s not an exaggeration to say that Sichuan hydropower has been the key secret ingredient enabling provinces in renewables-poor regions of eastern China to meet their progressively stricter decarbonisation goals.
After exporting all this power, where does that leave Sichuan’s own power supply? Increasingly tight, it turns out. Over the past decade, Sichuan’s local industrial demand has risen rapidly, thanks to the cheap abundant hydropower in the rainy season. But, while hydropower is huge in Sichuan, it has highly variable wet and dry seasonality. Aside from the hydro, Sichuan has only a relatively small fleet of thermal power plants that see most of their use in the dry season when hydro performance lags.
…but Sichuan’s growing grid is facing tight power demands due to highly variable wet & dry seasonality…
Sichuan can cater for some variability in hydro inflows, but as demand grows and alternative generation does not, the seemingly oversupplied Sichuan starts to look a lot less comfortable. At the same time, most of the newly added hydro capacity like the Baihetan Power Plant has been earmarked for exports, with very limited grid infrastructure to link into the Sichuan grid. As already as 2020, provincial grid planners had begun voicing concerns about Sichuan’s dwindling power reserve margin.
…the multi-month drought in 2022 was a wake-up call when hydropower output halved
Such concerns proved to be prescient in July/August 2022 when a multi-month drought halved hydropower output in Sichuan, while record-breaking temperatures across China simultaneously boosted cooling power demand to unprecedented levels. By mid-August, as power load continued to set new daily records and major rivers withered away to a trickle, Sichuan was forced to halt power supply to its industrial sector to meet its commitments to residential and agricultural power users.
At the same time, similar measures to protect priority customer classes were implemented in neighbouring Chongqing, which imports much of its power from Sichuan. These measures allowed the direct effects of the heat wave on human livelihood to be greatly mitigated, while shifting the economic impact to the industrial sector, where economic losses added up to many billions of yuan.
It is tempting to characterize the exact conditions that brought about the Sichuan power shortages as a freak confluence of factors, a perfect storm that was impossible to predict. Certainly, Chinese media took that tack at the time. The reality, however, is more complex.
Expecting the Unexpected
Let’s look at how the power industry treats the likelihood of extreme weather. Suppose we look through many centuries of historical data and determine that a certain rare event like a severe drought occurs once every 100 years, on average. Thus, we may infer that this event has a 1% chance of occurring in any given year. Based on this premise, how many times should we expect to see this “once in a hundred year event” at least once over the next 100 years? The answer, derived by distribution analysis, may be surprising.
There’s a 63% chance a ‘one-in-100-yr extreme weather’ event will happen…
…we’re more likely to see this in our lifetimes than not
Over a 100-year period, the likelihood that this rare event or sequence of factors will happen at least once is actually a substantial 63%. We’re more likely to see this event in our lifetimes than not! Indeed, the likelihood that it happens at least twice is still a very healthy 26%, and even the chance that it happens at least three times in that 100 year period is just around 8%! Those three times could be spread out or quite bunched together. There is no assurance of ‘spacing’ in the realm of probabilistic events.
However, it has become abundantly clear, thanks to evidence not just from China, but all over the world, that even after including the peculiarities of probability analysis, these so-called “once in a century” weather events are appearing with increasing frequency, intensity, and regularity. This was definitively demonstrated in a 2022 Carbon Brief meta-study, which looked at 400-peer reviewed studies of extreme weather and climate research and confirmed an unmistakable trend: the extreme weather is appearing more and getting worse. A once in a century event may now actually be a one in 50 years event, or lower.
The challenge now lies in determining what ‘surprise’ we choose to prepare for in advance
The challenge, then, lies in determining what surprises we choose to prepare for in advance. Not only did Sichuan rainfall decrease by over 50% in July and August, during what is normally the peak rainy season, but power demand was soaring at the exact same time due to the record-setting heat.
While reservoir levels were already running low in mid-July, the idea that the drought would last through the rest of what is normally a very wet time of the year in Sichuan must have seemed unbelievable at the time, no matter how clear hindsight is now. Could something have been done differently?
Surprise is the natural response of the unprepared. Both temperature and hydrology are inherently variable – and have been forever. To be surprised and caught so short didn’t happen because Sichuan generators and grid operators didn’t know that temperatures and rainfall could vary. Surprise arises when factors combine in ways that one is not prepared for. This usually means that the series of factors combined in a way that was deemed to have a very low probability of occurrence – and was therefore unnecessary to prepare for.
Preparing to avoid ‘surprises’ has a 2-sided solution
1) Focusing on the probability of an event & 2) magnitude of harm that would result if a low-probability event occurs
Preparing to avoid surprises is a problem with a two-sided solution. On one hand, one must focus on the probability of one or a sequence of events occurring at all. On the other hand, one must account on the magnitude of harm that would result if a low-probability event occurs. A common event with very limited harm may pose limited concern. An extremely rare event with the potential for extraordinary harm may merit additional attention.
So, when we are trying to predict combinations of future factors, it matters how we frame the problem – should we focus on the probability of the event occurring, or the magnitude of the potential harm? Obviously, it would ideally be both, but when we look back at significant disruptive weather events, the way we talk about the event often focuses on the element of surprise and how much of an unexpectedly “perfect storm” the event’s occurrence was, and not the degree of harm caused. While such a narrative framing may be more compelling for lay readers, it is unhelpful in terms of preventative planning.
If we want our energy systems to be futureproofed, they need to be upgraded for the reality of 2023
In an era where climate change is wreaking havoc with weather forecasts derived from climatic assumptions and processes of the past, a shift of focus to how bad even seemingly rare combinations of events seems even more prudent. If such a shift of focus occurs, the conclusion is clear: If we want our energy systems to be resilient and future-proof, then they need upgrades, because whatever was considered sufficiently robust twenty or thirty years ago is clearly insufficient by the reality of 2023.
The Chinese Policy Response
If this reality was not fully appreciated by Chinese energy policymakers previously, it certainly appears to be now. China, still in the process of developing and reforming its power sector and grid architecture, is now afforded the key opportunity to course-correct in real time, adding, enhancing, and revising its grid to ensure that the power system is ready for the “new normal” of 2023 and beyond. Crises beget responses, or as Churchill famously remarked “never let a good crisis go to waste”.
Chinese energy policy makers are now revising its grid to ensure power systems are ready for the “new normal” & beyond
Europe too, for example, is rewriting the rules for energy security and fuel dependency as we speak. But when Chinese policymakers are resolved to change something, they move fast. Recall that the coal pricing problems and power shortages in late 2021 to near-immediate changes to power pricing and dispatch, jump-starting sector reforms that had been stalled for years, at least on the surface (we do not discount the possibility these policies had already been planned and the power crisis simply presented a convenient opportunity for implementation). After this year’s drought event exposed the unexpected vulnerability of the hydropower fleet, Sichuan and other Chinese regions dependent on hydropower were surely re-examining their capacity mix and asking questions like:
Is a scenario where we lose 50% of our hydropower generation so rare as to be beyond consideration in our system design?
If we really lost 50% of our hydropower, how severe would the impact be on business and personal livelihood?
As we okay with the frequency with which we think we could lose 50% of our hydropower? If no, what must be done to make the system more robust?
On top of renewable expansion, Sichuan is building a gas-fired plant – a degree of protection against future droughts
Back in 2020, perhaps in response to the concerns of local grid planners, Sichuan approved an 8GW UHV line that would connect to massive wind and solar bases in Xinjiang that is under construction now. A similar line will connect to Chongqing as well, while the transmission lines between Chongqing and Chengdu (Sichuan’s capital) will also see their capacity doubled.
Sichuan’s domestic work plan at the end of 2022 also announced 7.7 GW of gas-fired capacity to be built in the province, as well as 8.5 GW of transmission capacity that will focus on linking major hydropower dams to the provincial grid and more regional HV lines connecting Sichuan to its neighbours’ grids. When this infrastructure is completed, all of China’s degree of protection against future droughts will be greatly enhanced.
Guangdong is also building a coal plant as power disruption can be minimised by timely use of ‘demand-side response’
Other provinces also quickly demonstrated that a lesson had been learned from the Sichuan drought. In the weeks after the power crisis was resolved, Guangdong unexpectedly approved 17GW of new coal-fired capacity not originally in its 14th Five Year Plan, slated to begin construction as soon as possible. Trying to understand this decision, we estimated the effect on Guangdong’s power supply if hydropower in Yunnan Province were ever to see a 50% decrease in production during the summer rainy season (with a corresponding 50% decrease in its hydropower exports to Guangdong).
Perhaps unsurprisingly, we found such a scenario resulted in daily 330GWh shortfall in Guangdong. During the August 2022 drought, Yunnan’s hydropower production was also impacted, although the effect on Guangdong this time was minimized thanks to the timely use of demand-side response that trimmed loads and forestalled the need for broader power cuts. Apparently, the way Guangdong looks at the risk of a hydropower disruption has been revised, and the capacity has been approved to be built, even if these plants will likely see very low capacity factors over their lifetimes.
Experience Elsewhere Varies…
Is that same lesson being learned in other parts of the world? In the United States, at least, we fear it is not, and we’re not the only ones.
But it’s clear the US hasn’t learned from China…
In the United States, electricity industry overseers like the North American Electric Reliability Corporation (NERC) have been ringing alarm bells for years that utilities in California, Texas, and elsewhere are going into battle with a climate adversary that is more volatile than ever before, underprepared with outdated expectations as to the frequency and severity of extreme droughts, floods, blizzards, and heat waves.
Unlike China, which is still in the process of building its grid infrastructure, most US infrastructure was built before climate change was a strong design consideration. Furthermore, utilities in the US, which have a shareholder mandate to operate profitably, are loathe to make investments based on conservative forecasts of extreme weather’s likelihood, instead preferring to look to historical data, which under-predicts the likelihood of such extreme weather events.
…it was a mystery even when CAISO forecasted supply & demand in summer 2022 to reach peak load 1 in 10 years
When California’s Independent System Operator (CAISO) forecast supply and demand for summer 2022, it evaluated the probability of reaching a peak load of 51.4GW to be 1 in 10, or something that most people think of as happening only once ever ten years (though in fact can happen two or three times in a given 10 year period with a considerable probability). But why is 1 in 10 years a ‘standard’?
That has to be one of the biggest mysteries of all. Perhaps it’s the way things have been done, so that’s the way they’ll always be done. In effect it has come to be an assessment of what society can accept given the nature of the risks and uncertainties that we have learned about and lived through over many decades.
It is arguably just an ossification of a planning convention from ages past in systems that were far less interdependent and with tolerances, in fact, for performance that were far less sensitive as they are today within our digital world. At the same time, the assessment also noted “these results do not take into account growing risks of more extreme events stemming from more disruptive climate change events”, even specifically highlighting: “the timeframe of greatest operational risk is during the late summer if the ISO and the west experience a widespread heat wave…concurrent with the diminishing effective load-carrying capability of solar resources and the wane of hydro generation.”
Despite being aware of the grid’s fragility, the risk was deemed acceptable according to the old ways
Despite this seeming awareness of the grid’s fragility, this risk was apparently deemed to be acceptable according to the old ways. In early September 2022, California hit a daily load peak of 52GW, forcing PG&E to warn half a million power consumers that they were at risk of outages and asking them conserve energy. CAISO issued an EEA 3 alert, the highest level of alert before utilities have to start implementing load shedding. Hanging by a thread, California’s grid shuddered and heaved, but made it through the event without major disruptions. The disaster of state-wide outages like in 2020 was narrowly averted…barely.
Though California & Texas hit peak demand, they managed to narrowly avoid state-wide outages
Meanwhile, Texas’ ERCOT continues to prepare its power grid for summer loads based on average historical weather data for the last 15 years, not a conservative view of future weather patterns. Earlier in 2022, NERC said Texas was at “elevated risk” for blackouts during the summer, a statement that was tested several times through this year’s ever more-aggressive summer heat waves. During one week in July 2022, Texas set several new records for peak demand, rising to 78.4GW just as a combination of wind lulls, cloud cover, and unplanned thermal outages forced utilities to ask power customers to conserve power and temporarily cut power to industrial facilities. While a repeat of the broad February 2021 outages was averted, Texas again revealed itself to be constantly walking the narrowest of lines between a perfectly balanced grid and rolling blackouts.
Barely Secure by Design
To be sure, many grid operators and utilities consider having “just enough” power to be the greatest victory of efficiency and forecasting. From their perspective, China’s solution of building its way out of the issue with more generation assets and grid infrastructure would be inefficient and inelegant, a ham-fisted approach to ensuring reliability that is expensive and over-designed. But we doubt this approach is so unacceptable to industry watchdogs and reliability experts in the US, who have now been shouting for years about the unappreciated fragility of key portions of the US grid, especially in regions where hydropower plays a big role.
Having “just enough power” shouldn’t be a victory…
…instead the power sector should increase costs & provide protection from potentially much higher costs of disruption
Yes, meeting unexpectedly larger challenges will increase costs, but the increase in the costs provides more protection from the potentially much higher costs of disruption. Over many years, we have formed estimates of loss for load-related costs. In the battle to increase precision, we already have a great deal of insight into the likely costs to societies associated with short, medium, and longer-term disruptions. Indeed, it often feels like new reliability studies are done periodically just to reaffirm conclusions that have already been reached before, but that no one is prepared to act upon.
Security of supply is a public good and given a choice, no individual customer or group wants to pay for it. Changing a planning criterion or standard — or even changing the way a standard is interpreted or assessed — means incurring significant cost for certain in preparation against a risk that has some probability of not happening anytime too soon. Perhaps that is what Churchill was really getting at when seeing opportunity in a crisis to take steps in a new direction that is suddenly much easier to appreciate.
It’s better to err on the side of overbuilding to forestall the next drought- or heatwave calamity…
If successful, China’s current approach will seemingly combine the best philosophies of an earlier era in the US (reliability of supply above all else) with typical Chinese regulated outcomes (cross-subsidize residential power customers, force state-owned enterprises to absorb the costs of T&D upgrades or stranded assets while eating high fuel costs with minimal passthrough to power customers). Eventually, China plants to implement capacity markets in its regional markets to better balance generator profitability and grid reliability, another feature both California’s CAISO and Texas’ ERCOT refuse to implement.
Regions like Cali & Texas must start prioritising ‘reliability’ & make the corresponding changes to power systems
Will China’s approach to future-proofing its grid from future disruptions to generation succeed? We think so. At this relatively early stage of its liberalisation, after decades of a highly regulated and centralised electricity sector, we understand why it’s preferable to err on the side of overbuilding. We still see the indications of a flexibility and pragmatism in electricity and grid planning, a determination that power sector decarbonization and liberalization can only happen upon a foundation of energy security, and a willingness to make adjustments when the facts of our changing climate call for it. Whether these measures turn out to be sufficient in forestalling the next drought- or heatwave-induced calamity remains to be seen, but at least no one will be able to accuse them of sleepwalking or stumbling into it.
Spending on grid resiliency is far less costly than being unprepared for a 100-yr-drought that knock’s half of the hydro grid
This, unfortunately, is exactly what might be happening in California and Texas, which are facing similar challenges, but not adopting similar tactics. In these regions, now is the time to prioritize reliability, to take a harder look at older standards and probability assessments are that probably non-conservative and make the corresponding changes.
If the outcome is that more money must be spent on grid resiliency, even at the risk of redundancy, it’s still going to be far less costly than the outages or disruptions caused by next year’s unanticipated “100 year drought” that knocks out half of the grid hydro.
Further reading
- Rivers are Running Dry Today – Rivers are our lifelines; they support cities, food & economies for centuries. This summer they were tested with severe drought/floods – will they fail? CWR’s Tan & Lam dive into challenges facing 5 major river basins that are the industrial/agricultural heartlands for China, Pakistan, US and Europe.
- China’s Growing Water Risk Factor – China is especially vulnerable to water-related climate risks but that also means increasing capacity for adaptation for which it has big ambitions. Moore, Director of China Programs & Strategic Initiatives at UoP, expands
- China’s Path to Peak Emission Regions – Are Chinese provinces set to meet carbon emission peak ambition by 2030? Dr Zhanfeng Dong from the Chinese Academy for Environmental Planning lists out leaders & laggards and unveils what’s on their action list
- First-ever 14FYP for Water Security – 8 Key Thoughts – China’s first-ever 14 Five Year Plan for Water Security signals that it is well ahead in water adaptation & the IPCC’s “climate resilient development”. CWR’s Debra Tan & Dr CT Low share more key thoughts
- Sponge City Is Transforming Urban Flood Management – Dr. Chan, Dr. Thadani & Li explain how Sponge city program (SCP) combats urban flood risks in China & the role played by social media
More on Latest
- CDP Global Water Report 2022 “Riding the Wave” – Key Takeaways – CDP’s new report reveals billions in potential gains for Southeast Asian companies that disclose water risks. Souza, CDP’s Regional Manager for Water Security, shares more key findings
- No River, No Power – are we energy secure when rivers run dry? – CWR’s new report revealed 16 Asian countries’ power generation assets with installed capacity greater than the G7 ex-US face sizable trifecta exposure from escalating climate risks and rivers running dry. There are clear national energy security implications unpack this reliance on 10 climate-sensitive rivers now!
- SWEL: The Waveline Magnet – Meet The Waveline Magnet, the only wave energy convertor that could rival fossil fuels. We chat to Zakheos, Co-CEO of SWEL, to unravel its potential & why it’s not a major player yet

Read more from Mike Thomas →

Read more from David Fishman →