Use load forecasts, demand management and equipment analysis to identify efficiency opportunities. Recommendations depend on local tariffs, equipment constraints and authorized operating windows.
Measure value. Verify results.
Evaluate each project using measured energy, resource availability and service performance. Compare a documented baseline with observed results under the same operating boundary. No fixed returns or financialized compute assets are implied.
AI concept · Not a project photograph or product specification24-hour compute-power model
Illustrative model, not live data, a DCP benchmark or a return promise. Compare power, grid energy and electricity cost at equal daily work.
Compute energy (unchanged)
480 kWh
Scheduled grid energy
1,788 kWh
Energy-charge difference
81.6 CNY
Grid peak (before → after)
125 → 114 kW
Swipe the chart or use the hour slider above to inspect values.
12:00 · Original 115 kW · Scheduled 103 kW · Solar 100 kW
Swipe the chart or use the hour slider above to inspect values.
12:00 · Original 15 kW · Scheduled 3 kW
Same tariffs and daily energy; grid energy charges only.
- Grid
- 1,788 kWh
- On-site solar
- 762 kWh
- Total site energy
- 2,550 kWh
Solar consumed + grid imports = site demand. Surplus solar earns no revenue in this model.
Assumptions, limitations and hourly data
Synthetic day, 1-hour steps. Fixed load: 95 kW at 08–17, 105 kW at 18–21, otherwise 70 kW. Original compute: 20 kW × 24 h. Movable work is ready at 00:00 and due by 24:00. Site limit: 160 kW. Assign 1 kWh increments by marginal grid cost. Illustrative tariffs: 0.4 at 00–06 and 23, 1.1 at 17–21, otherwise 0.7 CNY/kWh, not a regional tariff. Excludes migration overhead, demand charges, storage, depreciation, taxes, export revenue and emissions. Peak reduction is not the objective; grid peak may increase. Solar follows a synthetic sine curve.
| Hour | Original | Scheduled | Solar | Grid·before | Grid·after | Tariff |
|---|---|---|---|---|---|---|
| 0 | 90 | 114 | 0 | 90 | 114 | 0.4 |
| 1 | 90 | 114 | 0 | 90 | 114 | 0.4 |
| 2 | 90 | 114 | 0 | 90 | 114 | 0.4 |
| 3 | 90 | 114 | 0 | 90 | 114 | 0.4 |
| 4 | 90 | 114 | 0 | 90 | 114 | 0.4 |
| 5 | 90 | 114 | 0 | 90 | 114 | 0.4 |
| 6 | 90 | 114 | 0 | 90 | 114 | 0.4 |
| 7 | 90 | 78 | 26 | 64 | 52 | 0.7 |
| 8 | 115 | 103 | 50 | 65 | 53 | 0.7 |
| 9 | 115 | 103 | 71 | 44 | 32 | 0.7 |
| 10 | 115 | 103 | 87 | 28 | 16 | 0.7 |
| 11 | 115 | 103 | 97 | 18 | 6 | 0.7 |
| 12 | 115 | 103 | 100 | 15 | 3 | 0.7 |
| 13 | 115 | 103 | 97 | 18 | 6 | 0.7 |
| 14 | 115 | 103 | 87 | 28 | 16 | 0.7 |
| 15 | 115 | 103 | 71 | 44 | 32 | 0.7 |
| 16 | 115 | 103 | 50 | 65 | 53 | 0.7 |
| 17 | 115 | 103 | 26 | 89 | 77 | 1.1 |
| 18 | 125 | 113 | 0 | 125 | 113 | 1.1 |
| 19 | 125 | 113 | 0 | 125 | 113 | 1.1 |
| 20 | 125 | 113 | 0 | 125 | 113 | 1.1 |
| 21 | 125 | 113 | 0 | 125 | 113 | 1.1 |
| 22 | 90 | 78 | 0 | 90 | 78 | 0.7 |
| 23 | 90 | 114 | 0 | 90 | 114 | 0.4 |
A clear path from inputs to evidence.
Place eligible inference, caching, retrieval and lightweight tasks near their data sources. Evaluate network transfer, latency and resource use before deciding which workloads to move.
Track device health, available capacity, response records, maintenance and service levels. Use the evidence for capacity planning and project-level cost allocation.
Clear responsibilities. Coordinated operation.
Agree on equipment, interfaces, retention periods, approvals and operating metrics at project initiation. Performance is verified through site measurements and acceptance testing.