Utilizing Analytics to Model Distribution System Losses with Smart Grid Data

Transcription

1 Utilizing Analytics to Model Distribution System Losses with Smart Grid Data Scott Albrechtsen BC Hydro, Load Analysis February 19, 2014

2 2 Author Scott Albrechtsen is a Senior Load Advisor at BC Hydro. He holds a Master s degree in Applied Economics from the University of Arizona (2007). Scott joined the BC Hydro Load Analysis Team 6 years ago doing predictive modeling and data mining for BC Hydro Rates, Load Forecast and Distribution Planning. He is a SAS Certified Programmer and Vancouver SAS User Group (VanSUG) Vice-President.

3 3 Agenda Business Problem Pi SCADA Distribution Feeder metering data Smart Meter (SMI / AMI) hourly data Grid Topology Estimating Loads Streetlights Non-SMI The Arithmetic Technical Losses Non-technical Losses

4 4 Business Problem How do we model hourly distribution grid system losses? Losses at the substation, feeder, feeder section, or transformer Technical & Non-Technical Losses

5 5 SCADA Feeder Metering Data We have hourly metering at different points of the distribution grid Substations, Feeders, Feeder Sections, Customers SCADA metering system (Plant Information Pi ) A single feeder (circuit) is examined here

6 Hourly Pi SCADA Feeder metering data (One Week) 6

7 7 Smart Meter (SMI / AMI) hourly data We have hourly Smart Metering at most customer points within the Distribution Grid

8 8 Smart Meter (SMI / AMI) hourly data We aggregate many customer kwh / hour loads along various segments of the distribution grid kwh / Hour

9 140 Smart Meter (SMI / AMI) hourly data Aggregated kwh / Hour

10 Grid Topology Example : Feeder XXX 10 Feeder 25122XXX

11 11 Grid Topology Example : Feeder XXX Example for one distribution feeder with 2,162 Residential customers and 89 commercial customers Customer Type Analogue Metered Customers SMI Meter Customers Street Light Accounts Traffic Light Accounts Total Commercial Customers Residential Customers 140 2, ,245 Total 163 2, ,334

12 12 Estimating Loads? Not all distribution nodes & customers have hourly metering. Even in a Smart Meter (SMI) environment We must estimate unmetered loads, non-smi loads, and SMI meters with data issues

13 13 Transformer # Watts of Lighting kwh / Hour Street-lighting

14 14 How do we estimate non-smi loads? 2,500 Customer kwh Billed Bi-Monthly (Pre-SMI) 2,000 1,500 Total kwh 1, May Jun Jun Jun Jun Jun Jul Jul Jul-09 Day 6 Customer Hourly kwh Consumption (Post-SMI) 5 4 Total kwh Jun Jun Jun Jun Jun Jun Jun Jun Jun Jun May May May May Jun Jun Jul Jul Jul Jul Jul Jul Jul Jul Jul Jul Jul Jul Jun Jun Jun Jun-09 Day

15 How do we estimate non-smi loads? 15 2,500 Customer kwh Billed Bi-Monthly (Pre-SMI) 6 Customer Hourly kwh Consumption (Post-SMI) 2,000 5 Total kwh 1,500 1,000 Total kwh May Jun Jun Jun Jun Jun Jul Jul Jul-09 Day 0 19-Jun Jun Jun Jun Jun Jun Jun Jun Jun Jun May May May May Jun Jun-09 Day 23-Jul Jul Jul Jul Jul Jul Jul Jul Jul Jul Jul Jul Jun Jun Jun Jun-09 We can expand bi-monthly, monthly, or daily register kwh data into hourly data via Load Research Load Profiles

16 Load Research Load Profiles? 16

17 17 Feeder Load Components Modeled Load Customers Street Light load SMI Meter Customers Traffic Light Load

18 18 Feeder Load Components SMI Meter Customers Modeled Load Customers Street & Traffic Lights (very small)

19 Feeder Load Components 19

20 20 Pi SCADA Metering vs. Customer Loads Pi SCADA Metering All Customer Loads

21 21 Calculated Losses Losses (Technical & Non-Technical) kwh / Hour

22 22 What are Technical & Non-Technical Losses? Technical Losses Losses through primary drivers, secondary conductors, and distribution transformers They are a function of customer load Non-Technical Losses: Abnormalities and electricity theft A prevalent issue in British Columbia

23 23 Estimating Technical Losses Primary Losses Secondary Losses Feeder ID A (Primary) B (Primary) C (Primary) Series Feeder Reactor XXX E-08 Bin (12 kv) A B C Feeder ID A coefficient (coil) A coefficient (secondary) C coefficient (core losses, units in kw) XXX E E JJJ E E AAA E E AAA E E AEX E E AEX E E AEX E E AEX E E AEX E E Best-fit A coeff (Coil) Best-fit A coeff (Secondary) Average Constant Decay Constant Decay C coeff (Core) 4 kv kv kv kv kv kv kv kv kv A Coefficient = Constant * (Load ^ Decay)

24 24 Equations from Engineering We use these equations to determine Technical Losses as a function of Primary, Secondary Loads Feeder XXX: Primary Loss = ( *(Total load 2 ) + ( )*(Total Load) ) Transformer Core Loss = ( E-07) *(Secondary Load 2 ) ) Secondary Loss = (1.761E-06) *(Secondary Load 2 ))

25 25 Estimated Technical Losses Transformer Core Losses Secondary Losses Primary Losses

26 26 Losses Non-Technical Losses Technical Losses

27 27 Extension The methodology can be extended to the whole distribution system with feeder metering (hundreds of points)

28 28 Extension The methodology can be extended to the whole distribution system with feeder metering (hundreds of meters)

29 29 Extension Where do we have high non-technical losses?

30 30 Conclusions Modeling distribution system losses is a straightforward approach that requires: A high degree of data quality Computing capacity Analytical tools Important Caveats! The Pi Metering data quality must be acceptable The Grid Topology of the Feeder must be accurate