起步顺序Order of work 先接入现场作业状态,再做有经理确认的重排,最后扩展到更多机组与 SMR 数字孪生。Add live job status first, then re-planning with a manager confirming, then more units and SMR twins.
公开事实Public facts 已核对Checked
- 20–25KOutage AI 关联并排程的每次大修任务数Tasks per outage that Outage A I links and schedulesWorld Energy 报道World Energy article
- 18,059 MW2024 年底在运装机容量Installed capacity at end of 2024公司披露Company filing
- 11,000+2024 年安省与美国员工Employees in Ontario and the U.S., 2024公司披露Company filing
- Hourly与微软 Azure 共建的小时级清洁能源匹配平台Clean-energy matching platform built with Microsoft AzureWorld Energy 报道World Energy article
从一份计划,到一次实时的大修From a plan to a live outage 判断Judgment
- 现状(公开信息)Today, as public:用语言模型读取任务逻辑Read task logic with language models → Outage AI 提出排程Outage A I proposes the schedule → 一份计划草案A draft plan
- 建议Proposed:来自现场的作业实时状态Live job status from the floor → 重排建议A re-plan proposal → 大修经理确认Outage manager confirms
专家组裁决Expert panel 依据公开观点的推演Inference from public views
Andrew Ng
选一件可计量的事:一台机组大修的延误天数;半年内出数字。Track one unit's delay days; numbers in six months.
Ethan Mollick
先在历史大修上测出重排失败点;经理保持在回路。Replay past outages first to find where re-planning fails.
Ziad Obermeyer
缩短工期只是代理指标,安全完整才是目标;两者都要写明。Shorter is the proxy; safe, complete work is the goal.
Erik Brynjolfsson
设大修工期基线;预算含计划员培训与流程变更。Set outage-day baselines; budget for planner training.
Thomas Davenport
这会重塑大修的规划与执行;点名跨两者的负责人。Name one owner for planning and execution.
Foteini Agrafioti
A I 团队、IT 伙伴与电厂运营的权责写成一页。One page of ownership across teams.
Ethan Mollick:尽早用真实大修数据找边界。Test on real outage data early.
Ziad Obermeyer:触碰执行之前,先写清真实目标并建清单。Real target first.
→ 先回放历史大修;上线须经核安全治理评审。Replay first; live use after safety review.
六位专家是谁、怎样思考、有什么铁律 →Who the six experts are, how they think, their rules →
三个目标Three priorities 建议Recommendation
如实统计延误天数Count delay days honestly
改变之前,先为一台机组建大修工期与延误原因基线。Baseline outage days and delay causes for one unit, before changing anything.
加入实时状态与重排Add live status and re-planning
先回放历史大修,再基于实时状态提出重排,由经理确认。Replay past outages, then propose re-plans from live status, a manager confirming.
把模式带到 SMRCarry the pattern to SMR
对小型模块化反应堆的数字孪生与 AI 运维,沿用同一道关卡。Apply the same gate to digital-twin and A I operations for small modular reactors.
场景组合Scenarios 建议Recommendation
| 场景Scenario | 做法Approach | 状态Status |
|---|---|---|
| 执行期间的重排Re-planning during execution 提效Efficiency降风险Risk | 接收现场作业状态并提出重排;由大修经理决定。Take live job status and propose a re-plan; the outage manager decides. | 建议 · 基于 Outage AIProposed · builds on Outage A I |
| 更多机组、整个大修More units, whole outage 提效Efficiency | 把 Outage AI 从排程扩展到多机组的完整大修。Extend Outage A I from scheduling to the full outage across units. | 建议Proposed |
| SMR 数字孪生S M R digital twin 降风险Risk体验Experience | 在小型模块化反应堆项目中使用数字孪生与 AI 运维。Use digital twins and A I operations in small modular reactor projects. | 建议 · 与安全治理确认Proposed · confirm with safety governance |
| 资产健康与负荷预测Asset health and load forecasts 提效Efficiency增收Revenue | 发电资产预测性维护,以及负荷预测。Predictive maintenance for generating assets, and load forecasting. | 建议Proposed |
12 个月路线12-month roadmap 建议Recommendation
基线Baseline
- 一台机组大修工期基线Outage-day baseline, one unit
- 延误原因归类Delay causes catalogued
- 执行使用的关卡规则Gate rules for execution use
回放Replay
- 回放历史大修Replay past outages
- 经理评审建议Managers review proposals
- 记录失败案例Failure log kept
上线Go live
- 接入实时状态Live status connected
- 重排由经理确认Re-plans confirmed by managers
- 与班组每日复盘Daily review with crews
扩展Extend
- 第二台机组Second unit
- SMR 孪生范围界定S M R twin scoping
- 年度重排优先级Annual re-ranking
治理与合规Governance 建议Recommendation
- 核安全治理评审任何进入执行环节的使用Nuclear safety governance reviews any use in execution
- 每次重排都由经理确认A manager confirms every re-plan
- 所有模型清单,对齐其负责任 AI 标准Inventory of every model, aligned with its Responsible A I standard
三层指标Three-layer metrics 建议Recommendation
| 采用Adoption | 每周使用建议的计划员Planners using proposals weekly · 参与评审重排的经理Managers reviewing re-plans |
|---|---|
| 过程Process | 被采纳的重排建议Re-plan proposals accepted · 延误后重排耗时Time to re-plan after a delay · 经理推翻比例Manager override rate |
| 结果Outcome | 对比基线的大修工期Outage days against baseline · 每台机组延误天数Delay days per unit |
目标值由企业自定。Targets are the company's to set.
风险Risks 判断Judgment
更短,不一定更安全Shorter, not safer
若只看节省天数,计划员可能把风险压进排程。If days saved is the only score, planners may push risk into the schedule.
计划与现场脱节Plan drifts from the floor
班组做不到的聪明计划会被忽略;与班组一起测试。A smart plan that crews cannot follow will be ignored; test with them.
自研工具依赖单点One in-house tool, one owner
Outage AI 可能依赖少数人;写明权责与后备。Outage A I may rest on a few people; write ownership and a back-up.
前 90 天First 90 days 建议Recommendation
| 第 1–2 周Weeks 1–2 | 点名规划与执行的负责人Name an owner across planning and execution |
|---|---|
| 第 3–4 周Weeks 3–4 | 选一台机组;定工期基线Pick one unit; set outage-day baseline |
| 第 5–10 周Weeks 5–10 | 起草关卡规则;回放一次历史大修Draft gate rules; replay one past outage |
| 第 13 周Week 13 | 与大修经理首次复盘First review with outage managers |
我们可能错在:We could be wrong if: 延误主要来自备件与许可,重排解决不了;或经理不信任自己没参与制订的排程。delays come mainly from parts and permits no re-plan can fix, or managers will not trust a schedule they did not build.
出处Sources
本页是基于公开资料的分析,与所述企业没有合作关系,不代表其官方立场或内部计划;企业数字多来自其自述或媒体转述;「专家组」是依据六位专家公开观点的推演,不是他们的原话,也不代表他们或其雇主的立场。资料整理日期:2026-11-14。This page is an analysis based on public information. It has no affiliation with the company and does not represent its official position or internal plans. Company figures come mostly from its own statements or media reports. The expert panel is an inference from six experts' public views, not their words, and does not represent them or their employers. Compiled 2026-11-14.
想给你的公司也做一份?Want one for your company?
同样的方法:公开事实、专属判断、12 个月路线和前 90 天。公司大小都可以聊,中文沟通。Same method: public facts, a company-specific judgment, a 12-month roadmap and the first 90 days. Any size of business; we can talk in Chinese or English.