Calculator D5

Pump Scheduling Optimization for Energy Cost Minimization

Scheduling when water pumps turn on and off to move water through pipes while using the least amount of electricity—especially when electricity costs more at certain times of day.

⚠️ Why It Matters

1
Time-varying electricity tariffs
2
Suboptimal pump runtime during peak-price hours
3
Excessive energy expenditure
4
Reduced OPEX margins for utilities
5
Deferred infrastructure upgrades due to budget strain
6
Increased carbon intensity per m³ delivered

📘 Definition

Pump scheduling optimization is a constrained mixed-integer nonlinear programming (MINLP) problem that determines the optimal on/off status, speed setpoints, and sequencing of centrifugal pumps in municipal water distribution networks over a time horizon (e.g., 24 h), subject to hydraulic continuity, pressure constraints, tank level dynamics, and time-varying electricity tariffs. It integrates hydraulic modeling, energy cost modeling, and operational reliability requirements to minimize total electrical energy cost without compromising service quality or system resilience.

🎨 Concept Diagram

Optimization WorkflowSCADA DataHydraulic ModelMINLP Solver

AI-generated illustration for visual understanding

💡 Engineering Insight

Never treat pump scheduling as purely an economic problem—it is a *hydraulic control problem with economic objectives*. A schedule that saves $12,000/year but causes 3 pressure violations/month will be overridden by operators within 48 hours. Always co-optimize for pressure reliability first, then extract cost savings *within* that feasible envelope. Field validation must include surge analysis—not just steady-state hydraulic checks.

📖 Detailed Explanation

At its core, pump scheduling optimization recognizes that moving water requires energy, and energy has a time-dependent price. Municipal systems store water in elevated or ground tanks to decouple supply from demand; this storage acts like a 'battery', allowing pumps to run when electricity is cheap (e.g., overnight) and deliver water when demand peaks (e.g., 6–9 a.m.). The simplest form uses binary decisions (on/off) for fixed-speed pumps, governed by tank level thresholds.

More advanced implementations integrate variable-frequency drives (VFDs), enabling continuous speed modulation. This introduces nonlinearities: pump power scales with the cube of speed (P ∝ N³), and efficiency drops sharply below 60% speed. Hydraulic models must therefore embed accurate pump affinity laws and system curve interactions—not just lookup tables. Real-world solvers often use rolling horizon MPC (Model Predictive Control) with 15-min resolution to balance computational tractability and responsiveness to demand forecast errors.

State-of-the-art deployments couple optimization with digital twin infrastructure: live SCADA feeds update demand forecasts hourly; weather-driven consumption models adjust for temperature/humidity; and cyber-physical safeguards enforce minimum pump runtimes, maximum daily starts, and emergency override protocols. Emerging practice includes co-optimizing with distributed energy resources (e.g., onsite solar PV) and participating in utility demand-response programs—transforming water utilities from passive load consumers into grid-responsive assets.

🔄 Engineering Workflow

Step 1
Step 1: Collect 1-year SCADA data (pump status, flow, power, tank levels, node pressures)
Step 2
Step 2: Calibrate EPANET hydraulic model against field measurements (demand pattern, pipe roughness, pump curves)
Step 3
Step 3: Characterize time-varying electricity tariff structure and verify utility billing logic
Step 4
Step 4: Formulate MINLP problem (objective: ∑(Powerₜ × Tariffₜ); constraints: mass balance, head loss, tank dynamics, pressure bounds)
Step 5
Step 5: Solve using decomposition (e.g., outer approximation) or commercial solver (Gurobi, CPLEX) with warm-start heuristics
Step 6
Step 6: Validate schedule via hydraulic transient simulation (e.g., EPANET-RTX or Bentley WaterGEMS) including pump start/stop surges
Step 7
Step 7: Deploy via PLC/SCADA integration with anti-hunting logic (minimum run/stop durations ≥ 15 min)

📋 Decision Guide

Rock/Field Condition Recommended Design Action
High tariff volatility (>2× ratio between peak/off-peak rates) + large diurnal demand swing (>40%) Deploy variable-speed drives (VSDs) on ≥2 dominant pumps; optimize for both timing and speed—not just on/off
Small network (<50 km pipe) with single elevated tank and fixed-speed pumps only Use binary integer programming with piecewise-linearized pump curves; avoid continuous speed optimization
Critical pressure-sensitive zone (e.g., hospital district) with narrow pressure band (<5 m H₂O tolerance) Add pressure-sensing feedback loop to scheduler; implement robust constraint tightening (±10% margin) in optimization

📊 Key Properties & Parameters

Electricity Tariff Profile

0.05–0.32 USD/kWh (U.S. municipal utilities, 2023–2024)

Hourly or sub-hourly unit cost of electricity (¢/kWh) reflecting demand charges, time-of-use (TOU), or real-time pricing signals.

⚡ Engineering Impact:

Directly drives objective function weighting; mischaracterization leads to >15% suboptimality in annual energy cost savings.

Pump Efficiency Curve (η vs. Q)

55–88% peak efficiency across Q = 0.3–1.0 Qₙₒₘ for ANSI/ASME B73.1 centrifugal pumps

Empirical or manufacturer-provided relationship between pump flow rate (Q) and hydraulic efficiency (η), typically modeled as quadratic or cubic polynomial.

⚡ Engineering Impact:

Nonlinear efficiency penalty at low flow or throttled operation significantly increases kWh/m³—ignoring it overestimates savings by up to 22%.

Tank Level Constraint Band

±15–30% of total tank volume (e.g., 4.5–13.5 m for a 9-m-diameter, 12-m-tall elevated tank)

Minimum and maximum allowable water levels in storage tanks, enforced to ensure hydraulic resilience and meet fire flow requirements.

⚡ Engineering Impact:

Tight bands restrict scheduling flexibility and force inefficient 'valley-filling' pumping; overly loose bands increase surge risk and chlorine decay.

Minimum Node Pressure Requirement

20–60 m H₂O (2–6 bar) depending on jurisdiction and building height

Legally mandated or design-specified minimum static/dynamic pressure (e.g., for fire hydrants or high-rises) at critical network nodes.

⚡ Engineering Impact:

Violations trigger hydraulic infeasibility; conservative assumptions unnecessarily constrain pump combinations and increase runtime.

📐 Key Formulas

Pump Power Consumption

P = (γ × Q × H) / (ηₚ × ηₘ)

Electrical power (kW) drawn by a pump, where γ = specific weight of water (9.81 kN/m³), Q = flow (m³/s), H = total dynamic head (m), ηₚ = pump efficiency, ηₘ = motor efficiency.

Variables:
Symbol Name Unit Description
P Pump Power Consumption kW Electrical power drawn by the pump
γ Specific Weight of Water kN/m³ Weight per unit volume of water (9.81 kN/m³)
Q Flow Rate m³/s Volumetric flow rate through the pump
H Total Dynamic Head m Total energy head imparted to the fluid by the pump
ηₚ Pump Efficiency dimensionless Efficiency of the pump (ratio of hydraulic power output to shaft power input)
ηₘ Motor Efficiency dimensionless Efficiency of the motor (ratio of mechanical shaft power output to electrical power input)
Typical Ranges:
Fixed-speed pump at best efficiency point
65–88%
VFD-controlled pump at 50% speed
42–61%
⚠️ ηₚ < 40% triggers economic shutdown; ηₘ < 88% warrants motor replacement.

Energy Cost Objective

min ∑ₜ₌₁ᵀ (Pₜ × λₜ × Δt)

Total electricity cost over T time steps, where Pₜ is power (kW), λₜ is tariff ($/kWh), and Δt is timestep duration (h).

Variables:
Symbol Name Unit Description
P_t Power kW Electric power consumed at time step t
λ_t Electricity Tariff $/kWh Price of electricity per kilowatt-hour at time step t
Δt Time Step Duration h Duration of each time step
T Number of Time Steps dimensionless Total number of discrete time steps in the optimization horizon
Typical Ranges:
24-h horizon, 15-min timesteps
96 timesteps
Typical U.S. municipal utility annual λₜ avg.
0.12–0.18 USD/kWh
⚠️ λₜ must reflect *true marginal cost*, not average rate—include demand charges if applicable.

🏭 Engineering Example

City of San Jose, CA – Alum Rock Pump Station (2022 Pilot)

N/A (water infrastructure system)
Tank Level Band
12.5–24.0 m (in 30-m elevated reservoir)
Pump Motor Power Rating
250 kW (3× vertical turbine pumps)
Minimum Fire Flow Pressure
35 m H₂O at 12 critical nodes
Annual Energy Cost Reduction
19.3% ($214,000)
Electricity Tariff Peak Rate
0.28 USD/kWh (4–9 p.m.)
Electricity Tariff Off-Peak Rate
0.09 USD/kWh (12–6 a.m.)

🏗️ Applications

  • Municipal drinking water systems
  • Wastewater lift stations
  • Irrigation district pumping plants
  • District cooling water circulation

📋 Real Project Case

Calibration of Lagos Metropolitan Water Network

Nigerian utility upgrading aging infrastructure across 12 zones

Challenge: Persistent model–field mismatch (>25% pressure error) due to undocumented pipe replacements and unac...
Calibration of Lagos Metropolitan Water NetworkZone 1Zone 2Zone 3Zone 4×1.32×1.45×1.58×1.62×1.68Demand Multiplier:CI Mains: C = 92 → 78PVC Laterals: C = 140 → 115Roughness (C-value):Challenge: >25% pressure error (undocumented pipe replacements, unaccounted demand growth)Sensors: 87 pressure loggers • 14 flow metersMain trunkZonal demandPipe roughness
Read full case study →

🎨 Technical Diagrams

Tariff ProfileOff-PeakPeakShoulder
Pump Efficiency CurveBEP

📚 References

[1]
AWWA M18 – Water Utility Energy Management — American Water Works Association
[2]
ISO 50001:2018 – Energy management systems — International Organization for Standardization
[3]
EPANET 2.2 User Manual — U.S. Environmental Protection Agency