Energy runs on weather.
We make that predictable.
预报的边界，就是能源的边界。
天有不测，但风云可算。
Regenesis Energetics combines dynamical climate models, multi-model ensembles, and machine learning to turn atmospheric prediction into asset-level intelligence for renewable generation, load, and grid operations.
Regenesis Energetics 融合气候动力模式、多模式集合与机器学习，把气象预测转化为场站级的功率预测与决策依据，服务于风电、光伏、负荷与电网运行。
Ensemble spread against lead time — uncertainty is a quantity, not a caveat
集合预报的发散随预报时效增长 —— 不确定性可以被算出来，而不是被回避
The problem问题
The grid now runs on the atmosphere.
电网的稳定，正越来越取决于天气。
Wind and solar moved uncertainty from the demand side to the supply side. Generation is no longer scheduled — it is forecast. And the historical statistics that dispatch, bidding, and risk models were built on are being broken by extreme and compound events that have no analogue in the record.
风电和光伏把不确定性从负荷侧搬到了电源侧。发电量不再由调度决定，而是由天气决定。与此同时，调度、报价和风险模型所依赖的那套历史统计规律，正在被越来越频繁的极端天气和复合极端事件打破。
Meanwhile most forecasts still report a single number, as if the future had one value. Bidding, scheduling, and hedging are probability problems. They deserve probabilistic answers.
但今天大多数预报仍然只给一个数字，仿佛未来只有一种可能。报量报价、发电计划、套期保值，本质上都是概率问题——它们需要的是概率化的答案。
The approach方法
From the atmosphere to the decision.
从大气到决策，打通中间这一段。
Most companies own one link in this chain. The value is in connecting them — because the hard part is no longer forecasting the weather. It is knowing what that forecast means for a specific asset, on a specific day, in a specific market.
这条链路上，大多数公司只做其中一环。真正的难点已经不在于预报天气本身，而在于回答一个更具体的问题：这个预报对某一个场站、在某一天、在某个市场里，究竟意味着多少电量、多少钱。
Atmospheric data
气象数据
Reanalysis, observations, and operational ensembles.
再分析资料、实测观测与业务化集合预报。
Multi-model & AI
多模式与 AI 融合
Dynamical ensembles with AI post-processing and calibration.
动力模式集合 + AI 后处理与订正，输出不确定性范围。
Energy mapping
功率映射
Atmospheric fields to power output, under physical constraints.
在物理约束下，把气象要素转换为发电功率。
Decision
决策
Bids, dispatch, hedges, and risk exposure.
报量报价、发电计划、套保与风险敞口。
Steps two and three are where we work. Global weather models are becoming a commodity; the difficulty has moved downstream.
我们的工作集中在第二和第三环。全球气象大模型正在快速商品化，真正的壁垒已经下移到下游。
Horizons时间尺度
One continuous chain, from hours to decades.
从小时到年代际，同一条连续的链。
Predictability comes from a different physical source at every scale. Most forecasting companies work at one end of this chain. The science underneath it is the same, so we work across all of it.
每一个尺度上，可预报性来自完全不同的物理来源。多数预报公司只做这条链的一端。而支撑它们的科学是同一套，所以我们做全链。
Skill decays with lead time, and it does not decay smoothly. What we offer at every scale is not equal confidence — it is equally honest uncertainty.
预报技巧随时效衰减，而且不是均匀衰减。我们在每个尺度上给出的不是同等的确定性，而是同等诚实的不确定性。
Platform平台
One stack, three layers.
一个技术栈，三个层次。
Built to be used independently, designed to compound when used together.
三层可以单独使用，叠加使用时效果相互增强。
Renewable & grid intelligence
新能源与电网智能
Probabilistic generation and load forecasting at the asset level — delivered as distributions, not point estimates.
场站级的概率化功率预测与负荷预测——输出的是概率分布，而不是一条曲线。
- Day-ahead, intraday, and short-term horizons
- Wind and solar generation by asset
- Load and grid-level demand
- Imbalance and bidding risk exposure
- 日前、日内与超短期多时间尺度
- 分场站的风电、光伏功率预测
- 负荷与电网级需求预测
- 偏差考核与报量报价风险敞口
Weather & climate intelligence
气象与气候智能
Calibrated, downscaled prediction across the full range of horizons, with quantified uncertainty throughout.
- Wildfire, heatwave, hurricane, extreme rainfall and flood risk
- Compound events, where hazards arrive together and correlate
- Atmospheric rivers and marine heatwaves as sources of sub-seasonal skill
经过订正与降尺度的气象预测，覆盖全时间尺度，并始终附带不确定性量化。
- 山火、热浪、飓风、极端降水与洪涝风险
- 复合极端事件——多种致灾因子同时发生且相互关联
- 大气河与海洋热浪：次季节尺度可预报性的重要来源
Enterprise decision platform
企业决策平台
Weather, generation, market, and operational data brought into a single decision layer for portfolio-scale operators.
把气象、发电、市场与运行数据汇入同一个决策层，面向组合级的资产运营方。
In development研发中Research研究
Forecasting is a science problem before it is a software problem.
预测首先是科学问题，然后才是软件问题。
Regenesis is built on research in atmospheric predictability — the study of how far ahead prediction remains meaningful, and how uncertainty grows with lead time.
Regenesis 的底层是可预报性研究——预报在多长时效内仍然有意义，以及不确定性如何随时效增长。
That question is no longer academic. As renewable penetration rises, the limit of predictability becomes an operating constraint with a price attached to it. Our work is to measure that limit honestly and deliver forecasts that carry it.
随着新能源占比不断提高，这个问题已经不只是学术问题。可预报性的边界，正在变成一条带价格的运行约束。我们要做的，是老老实实把这条边界量出来，并让每一次预报都带着它。
We don't treat AI as a replacement for physics. Our systems combine dynamical climate models, multi-model ensembles, and machine learning. Ensembles characterize where uncertainty comes from. Dynamical models keep predictions physically consistent. AI corrects systematic bias, downscales to the asset, and maps atmospheric fields to power.
Purely data-driven models perform well inside their training distribution and are least reliable outside it — which is precisely when extreme weather happens, and precisely when the grid needs the forecast most.
Lorenz showed that small perturbations grow — that prediction has a horizon. Most forecasting products ignore this and report a single number. We don't. Knowing the limit of predictability is what makes a forecast usable.
我们不把 AI 当作物理的替代品。Regenesis 的预测系统同时使用气候动力模式、多模式集合与机器学习：集合刻画不确定性的来源，动力模式保证物理一致性，AI 负责订正系统性偏差、降尺度到场站，并把气象要素映射为发电功率。
纯数据驱动的模型在训练分布之内表现优异，在分布之外最不可靠——而那恰恰是极端天气发生的时候，也恰恰是电网最需要预报的时候。
Lorenz 早就证明，微小的扰动会不断放大——预报存在极限。大多数产品选择回避这一点，只给一个数字。我们不回避。知道预报在哪里失效，才知道它在哪里可信。
About关于
Built by scientists and engineers.
由科学家与工程师创立。
Regenesis Energetics was founded by researchers from Princeton, Stanford, Caltech, and Georgia Tech, working alongside engineers who have built and deployed large-scale systems at Microsoft, Amazon, and IBM.
Regenesis Energetics 由来自 Princeton、Stanford、Caltech 和 Georgia Tech 的研究者创立，团队中还有曾在 Microsoft、Amazon、IBM 负责大规模系统开发与部署的工程师。
That combination is deliberate. Weather-to-energy intelligence fails at the seams: where atmospheric science meets model engineering, and where model output meets how power markets actually operate. We built the team to cover all three.
这样的团队构成是刻意的。从天气到能源，问题往往出在接缝处：大气科学与模型工程之间，模型输出与电力市场实际运作方式之间。我们按这些接缝来配置团队。
There is a climate argument here too, and it is a practical one. Better forecasts mean less curtailment and more clean electricity actually delivered to the grid. Mitigation, in this business, is a forecasting problem.
Institutional and company affiliations listed above reflect the prior education and employment of individual team members. They do not imply endorsement, sponsorship, or partnership.
这件事也有它的气候意义，而且是很实际的一层：预报更准，意味着更少的弃风弃光，意味着更多清洁电力真正送上电网。在这个行业里，减排首先是一个预报问题。
上述院校与公司名称仅说明团队成员个人的教育与工作背景，不代表任何形式的背书、赞助或合作关系。
Contact联系
Working on wind, solar, or grid operations?
您的业务与风电、光伏或电网运行相关？
We are onboarding a small number of early partners — asset owners, independent power producers, and trading desks who want probabilistic forecasts built for how they actually bid and dispatch.
我们正在招募少量早期合作伙伴：场站持有方、独立发电商与电力交易团队。如果您需要的是真正贴合报量报价和调度方式的概率化预测，欢迎联系我们。
Get in touch联系我们