DeepManthan AI | ← Back to hub
Interactive Simulator

Pump scheduling simulator

Watch how MPC shifts pumping to off-peak tariff windows, pre-fills the elevated tank overnight, and avoids peak-hour electricity costs — saving 25–40% on pump energy.

Electricity tariff
Off-peak
₹3.2/kWh · pump now
Tank level
72%
Rising · 3 h reserve
Pump speed (VFD)
85%
MPC decision
Pump power
42 kW
Cube law at work
Energy saved today
₹4,180
vs. reactive fixed-speed
12a3a6a9a12p3p6p9p11p

24-hour MPC pump schedule

Cyan = off-peak (pump hard), amber = mid tariff (partial), red = peak (coast on tank). Bar height = pump speed %. Click any bar to jump to that hour.

MPC at 6 AM: Off-peak tariff window. Tank at 72%. Pumping at 85% to continue pre-loading before 7 AM peak tariff begins. Expected saving vs. reactive: ₹4,180 today.

Tariff structure — MSEDCL industrial HT

TOD tariffs penalise peak-hour consumption. The same kWh of pumping work costs 60% less at night than at peak.

ComponentRateAI strategy
Energy charge₹3.2/kWhPump at full speed — cheapest window. Fill tank to 88%+.
TOD multiplier0.80× (20% discount)Off-peak discount fully exploited
Demand charge₹400/kVA/monthVFD keeps power factor >0.95
kVAh billingUnity PF targetCapacitor banks auto-switched

🔌 Monthly bill impact — 500 kVA pump station, 100 MLD supply

AI scheduling vs. fixed-speed reactive operation on the same network.

Baseline bill
₹28.4 L
No TOD optimisation
AI-optimised bill
₹18.1 L
MPC + TOD shifting
Monthly saving
₹10.3 L
36% reduction
Simple payback
14 mo
On AI system capex

Demand forecasting — the ML layer

The MPC needs to know how much water each zone will need over the next 24 hours. ML learns these patterns continuously.

ML inputWhy it mattersUpdate frequency
Historical zone demand by hour/dayBaseline demand curve per DMAContinuous — 90-day weighted
Day of week / holiday calendarIndustrial Zone C drops 60% on SundaysDaily via calendar API
Max temperature forecast+1°C → +0.8% residential demand6-hourly NWP forecast
Pressure sensor readingsLeak detection via model anomalyEvery 5 minutes
Pipe roughness C (calibrated)Keeps friction model accurateWeekly ML re-fit