An end-to-end data analytics and commercial optimization assessment of two years (35,040 half-hourly intervals) of electricity consumption data for a UK commercial facility operating under a Time-of-Use (ToU) tariff structure.
The analysis evaluates building energy performance, isolates an unmonitored operational Building Management System (BMS) control failure, quantifies weather-normalized efficiency shifts using Ordinary Least Squares (OLS) regression, decomposes energy cost drivers, resolves kVA capacity exceedances, and outlines a prioritized implementation roadmap.
| Metric | Year 1 (2024/25) | Year 2 (2025/26) | Operational Variance |
|---|---|---|---|
| Total Import Volume | 990,104 kWh | 1,072,975 kWh | +8.37% (+82,871 kWh) |
| Total Spend | £186,038 | £215,663 | +15.92% (+£29,625) |
| Weather-Expected Volume | 990,104 kWh | 981,203 kWh | +9.35% Weather-Adj. Shift (+91,772 kWh) |
| Unoccupied Mean Load | 93.95 kW | 105.33 kW | +12.12% (+60.4% post-Jan 12 fault) |
| Peak Demand (kVA) | 303.21 kVA | 343.79 kVA | +13.38% (11 breaches >320 kVA limit) |
| Average Power Factor | 0.938 | 0.917 | -2.24% (dipped to 0.855 min) |
- Step-Change Identification: The site was operating 8.98% more efficiently in Year 2 until Monday, January 12, 2026, when a control override/setback failure occurred.
- Unoccupied Baseload Jump: Overnight and weekend baseline demand stepped up from 82.22 kW to 131.89 kW (+49.67 kW continuous excess load).
- Direct Financial Loss: Incurred 190,311 kWh in excess consumption and £36,038 in unnecessary energy cost across 5.5 months. Rectifying this fault yields £36,000–£72,000/year in recurring savings.
Executive Performance Summary: Control Anomaly Impact
Performance Metric Pre-Anomaly Baseline Post-Anomaly Period Absolute Shift Impact / Variance
Unoccupied Mean Demand (kW) 82.22 kW 131.89 kW +49.67 kW +60.4% baseload increase
Occupied Mean Demand (kW) 143.33 kW 183.04 kW +39.71 kW +27.7% load increase
Average Power Factor 0.938 0.893 -0.045 -4.8% Inductive motor load degradation
Daily Excess Energy Shift —— —— +1,119.5 kWh/day Occupancy-adjusted daily excess
Effective Electricity Rate —— £0.1894 / kWh —— Post-anomaly weighted average
Daily Financial Burn Rate —— —— +£211.99/day Ongoing daily financial loss
Total Excess Consumption —— —— 190,311.37 kWh Cumulative extra energy
Direct Financial Cost of Fault —— —— £36,038.18 Cumulative financial impact
- Year 2 was 10.9% milder in heating degree days (
$\text{HDD}_{15.5} = 1,851.9$ vs$2,079.0$ ). - An OLS regression model trained on Year 1 baseline data (
$R^2 = 0.682$ ) projected an expected Year 2 consumption of 981,203 kWh. - The true weather-normalized operational deterioration was +91,772 kWh (+9.35%), confirming that efficiency loss was driven by internal control failures rather than weather dynamics.
Ordinary Least Squares (OLS) regression model
=================== 1. MODEL VALIDATION ===================
R-squared : 0.6819 (Target: >0.75)
CV(RMSE) : 8.42% (Target: <15.0%)
NMBE : -0.00% (Target: +/- 5.0%)
================== 2. MODEL COEFFICIENTS ==================
Base Load (Intercept) : 2064.80 kWh/day (p=0.000)
Heating Sensitivity (HDD) : 33.30 kWh/HDD (p=0.000)
Cooling Sensitivity (CDD) : 51.29 kWh/CDD (p=0.016)
Workday Added Load : 634.92 kWh/day (p=0.000)
=================== 3. DECOMPOSED IMPACT ===================
Actual Year 1 Total : 990,104.36 kWh
Expected Year 2 Total : 981,203.28 kWh (Baseline Model + Y2 Weather)
Actual Year 2 Total : 1,072,975.33 kWh
------------------------------------------------------------------------------------------------------------------------------------
└─ Weather Impact (expected_y2 - actual_y1) : -8,901.07 kWh (Expected reduction)
└─ Weather-Normalized / Non-Weather Operational Shift (actual_y2 - expected_y2) : +91,772.05 kWh (+9.35%) True efficiency loss
└─ Total Raw Change (weather_impact + operational_shift) : +82,870.98 kWh (+8.37%)
- The +£29,625 (+15.92%) annual spend increase was mathematically decomposed into:
- Unit Rate Inflation: +£12,014 (+6.46%) from ToU rate increases across Red, Amber, and Green bands.
- Volume Growth: +£16,523 (+8.88%) driven by the post-Jan 12 control fault.
-
Capacity Charges: +£1,088 (+0.58%) from higher agreed capacity rates (£0.095
$\rightarrow$ £0.105/kVA/day) and kVA exceedance penalties.
=== ANNUAL COST COMPONENTS BREAKDOWN ===
period_year energy_import_cost allocated_capacity_charge excess_capacity_charge total_cost
0 Year 1 (Jul 24 - Jun 25) £174,655.30 £11,382.82 £0.00 £186,038.12
1 Year 2 (Jul 25 - Jun 26) £203,192.36 £12,469.30 £1.78 £215,663.44
=== COST VARIANCE & INFLATION DECOMPOSITION ===
Year 1 Total Spend: £186,038.12
+ Unit Rate Inflation Effect: +£12,013.88 (+6.46%)
+ Volume Increase Effect: +£16,523.18 (+8.88%)
+ Capacity Charge Increase: +£1,088.26 (+0.58%)
Year 2 Total Spend: £215,663.44 (+15.92%)
- The facility experienced 11 half-hourly capacity exceedances exceeding the 320 kVA Agreed Import Capacity (peaking at 343.79 kVA).
- Real active power (
$303.50\text{ kW}$ ) never exceeded 320 kW. All exceedances were caused by degraded power factor ($\text{PF} \approx 0.855\text{--}0.898$ ) during morning plant restart windows (08:00–09:30). -
Corrective Proof: Restoring site power factor to
$\ge 0.95$ via a 60 kVAR capacitor bank reduces peak apparent power to$319.47\text{ kVA}$ , completely eliminating 100% of capacity exceedances without requiring load curtailment.
_vs._Apparent_Power_(kVA)_&_Power_Factor_Boundaries.png)
.png)
| # | Action Item | Priority | Est. CapEx (£) | Est. Annual Savings (£/yr) | Simple Payback |
|---|---|---|---|---|---|
| 1 | BMS Setback & Override Audit | P1 - High | £0 – £500 | £36,000 – £72,000 | < 1 Week |
| 2 | Morning HVAC Plant Staggering | P2 - Medium | £0 – £1,000 | £2,000 – £5,000 | 1–2 Months |
| 3 | Power Factor Correction (PFC) Unit | P2 - Medium | £4,000 – £8,000 | £3,000 – £6,000 | 12–18 Months |
| 4 | Automated BMS Exception Alerts | P3 - Ongoing | £1,500 – £3,000 | £5,000 – £10,000 | < 6 Months |
| TOTAL | Full Portfolio Investment | — | £9,000 (Avg) | £69,500 / year | 1.6 Months |
Launch the pipeline instantly in your browser with zero setup:
├── building_energy_performance_analytics.ipynb # Interactive Google Colab Notebook
├── energy_dataset.xlsx # Raw synthetic half-hourly dataset
├── requirements.txt # Python dependencies
├── README.md # Executive summary & documentation
├── plots/
├── Daily_Electricity_Import.png
├── Monthly_Peak_kVA_Demand_vs_Agreed_Import_Capacity.png
├── Active_Power_(kW)_vs._Apparent_Power_(kVA)_&_Power_Factor_Boundaries.png
├── Average_Diurnal_Power_Factor_Profile_(Pre-_vs._Post-Jan_12,_2026).png
- Python 3.9+
- Data Processing:
pandas,numpy - Statistical Modeling:
statsmodels(OLS Regression) - Visualization:
matplotlib,seaborn - Spreadsheet Ingestion:
openpyxl
Distributed under the MIT License. See LICENSE for details.