Demand Forecast Hydraulic Impact Score (DFHIS)

GMI-WT-025Water TreatmentIntegration-Required

Definition

DFHIS measures the cumulative pressure deviation experienced across the distribution system due to gaps between AI-predicted demand and actual demand, quantified in PSI-hours per cycle.

The Measurement Gap

Measurement Gap and Decision Supported

The Invisible Failure Mode: A water utility deploys an AI demand forecasting system that controls pump scheduling across 6 pressure zones. The forecasting system reports 94% accuracy measured as mean absolute percentage error (MAPE). Operations considers this excellent and relies on the forecast for automated pump control. But the 6% error is not uniformly distributed: it concentrates during morning peak demand ramps and evening demand drops, creating 15–20 minute windows where actual demand diverges significantly from predicted demand. During these windows, system pressure in two zones drops 8–12 PSI below target, causing intermittent low-pressure complaints from customers at higher elevations. The forecast is 94% accurate by MAPE, but the hydraulic impact of the 6% error is significant and concentrated in the periods that matter most.

What Existing Metrics Miss:

Demand forecast accuracy is typically reported as MAPE, RMSE, or R². Statistical measures that treat all errors equally regardless of when they occur or what hydraulic consequence they produce. A 3% error during low-demand overnight hours has negligible hydraulic impact; the same 3% error during peak demand ramp can cause a 10 PSI pressure swing. Without DFHIS, there is no way to distinguish between harmless and consequential forecast errors.

The Decision DFHIS Enables:

Is the demand forecast accurate enough during hydraulically sensitive periods to justify automated pump control, or do specific forecast error patterns require manual intervention, revised pump scheduling logic, or targeted forecast model improvement?

Cross-References:

DFHIS connects to Digital Twin Fidelity Index (DTFI, GMI-WT-007) from the maintenance KPI set. Forecast errors that cause pressure deviations are one source of digital twin divergence. DFHIS also connects to Twin Calibration Divergence Rate (TCDR, GMI-WT-026) because demand forecast errors that propagate through the hydraulic model affect twin calibration. LDCC (GMI-WT-022) is relevant because undetected leaks in detection dead zones create demand that the forecast model cannot predict, directly inflating DFHIS. Process Envelope Utilization Index (PEUI, GMI-WT-001) addresses a parallel problem at the treatment plant: both measure whether AI-optimized operations are pushing against operating boundaries.

Formula

Plain language: For each pump scheduling cycle driven by the AI demand forecast, compare the predicted pressure profile to the actual pressure profile at monitoring points. Count events where the pressure deviated from target by more than a threshold (plus or minus 10 PSI), and score severity by the deviation magnitude and duration.

Notation:
DFHIS = Sum(|Pressure_actual_i - Pressure_target_i| x Duration_i) / N_scheduling_cycles

Where:

  • Pressure_actual_i = measured pressure at monitoring point i during the scheduling cycle (PSI)
  • Pressure_target_i = target pressure for that point per the scheduling plan (PSI)
  • Duration_i = time duration of the deviation event (hours)
  • N_scheduling_cycles = number of pump scheduling cycles in the measurement period
  • Only count deviations exceeding plus or minus 10 PSI from target

Units: PSI-hours per scheduling cycle. Higher = more hydraulic disruption per decision. Zero = perfect forecast-to-reality alignment (the scheduling plan produced the intended pressures).

Proposed formula, pending validation pilots.

At a Glance

Category
Distribution System Intelligence
← Customer Leak Alert Positive Predictive Value (CLAPV)Leak Detection Coverage Confidence (LDCC) →

Assess Your Organization

The AI-GPS Assessment evaluates your AI governance maturity and recommends KPIs like this one for your implementation roadmap.

Take the Free AI-GPS Assessment