A Time-Series Distribution Test System Based on Real Utility Data

I. INTRODUCTION

• Class I feeders modify real networks, while Class II systems are synthetic; existing models support distribution-system studies but have limitations for time-series analysis.Class I feeders are stationary snapshots, and synthetic-system findings may not generalize easily to practical networks.Source: §I · p. 1
• Existing U.S. test systems generally lack real time-series measurements; the European feeder offers only one day of data and lacks common components and U.S. system structure.The paper identifies a need for real circuit models with long-term time-series loads and common distribution components.Source: §I · p. 1
• The proposed real, fully observable Midwest grid includes one-year measurements from more than 1120 customers, network parameters, and detailed topology, with customer loads aggregated at secondary transformers for privacy.The component inventory includes overhead lines, underground cables, load-tap-changing and secondary transformers, capacitor banks, and circuit breakers.Source: §I · pp. 1–2

A. System Description

• The radial test system comprises three feeders supplied by a 69 kV substation.It belongs to a municipal utility in the Midwest U.S. and has smart meters at all customers.Source: A. System Description · p. 2
• The system has 240 primary network nodes and 23 miles of primary feeder conductor.Customers connect to primary network nodes through secondary distribution transformers.Source: A. System Description · p. 2
• Smart meters at all customers make the municipal-utility network fully observable.The network is a real distribution grid located in the Midwest U.S.Source: A. System Description · p. 2

B. Time-series Data Description

• Hourly energy data from 1120 customers spans January through December 2017 and captures seasonal load variation.Hourly average kW demand is approximated from energy assuming constant demand within each one-hour interval; customer power factors are randomly selected between 0.9 and 0.95.Source: B. Time-series Data Description · p. 2
• Grossly erroneous and missing smart-meter samples are removed using a statistical approach.The measurements are hourly energy consumption in kWh.Source: B. Time-series Data Description · p. 2

C. Load Tap Changing Substation Transformer

• The 69/13.8 kV, 10,000 kVA delta-wye substation transformer has an on-load tap changer and is modeled with equivalent sub-transmission short-circuit impedance.Its three independent single-phase tap changers are wye-wye connected and regulate Bus 1 voltage; each has 16 steps.Source: C. Load Tap Changing Substation… · pp. 2–3

D. Secondary Distribution Transformers

• Most three-phase secondary transformers serve commercial customers at 240 V phase-to-phase.Source: D. Secondary Distribution Transformers · p. 3
• Most single-phase centertapped secondary transformers serve residential customers at 120/240 V.Source: D. Secondary Distribution Transformers · p. 3
• The system's secondary transformers include three-phase and single-phase centertapped types.Source: D. Secondary Distribution Transformers · p. 3

E. Line

• The system includes overhead lines and underground cables with varied phasing, using real conductor and construction information to calculate series impedance and shunt capacitance.The conductor parameters are listed in Table III.Source: E. Line · pp. 3–4
• Fig. 4 shows one year of active and reactive power consumption at the substation transformer.Source: Figure 4 · p. 4
• Fig. 5 shows distributions of one-year nodal voltages for Feeder C.Source: Figure 5 · p. 4
• Table II presents parameters of distribution transformers.Source: Table II · p. 4
• Table III presents parameters of conductors.Source: Table III · p. 4

1) Overhead Lines:

• Conductor self and mutual impedances are calculated using modified Carson’s equations, then used to build the primitive impedance matrix and derive the phase impedance matrix by Kron reduction.Source: 1) Overhead Lines · p. 4
• Overhead-line shunt capacitance is ignored because its shunt admittance is small.Source: 1) Overhead Lines · p. 4
• Fig. 6 shows tap positions and voltage magnitudes at Bus 1 from July 1 to July 7, 2017.Source: Figure 6 · p. 5
• Fig. 7 presents the cumulative distribution function of power-flow solution error.Source: Figure 7 · p. 5

2) Underground Cables:

• Equations 1–4 are also used to calculate underground-cable series impedance.Source: 2) Underground Cables · p. 5
• Unlike overhead-line shunt capacitance, underground-cable shunt capacitance cannot be ignored and is calculated from cable geometry and material properties.The equation defines ε as medium permittivity, Rb as the radius through neutral-strand centers, RDc as phase-conductor radius, k as strand count, and RDs as strand-conductor radius.Source: 2) Underground Cables · p. 5

F. Shunt Capacitor Bank

• The test system has two shunt capacitor banks, located on Feeders B and C.Source: F. Shunt Capacitor Bank · p. 5
• Each capacitor bank is rated at 50 kVAr and uses a grounded-wye connection.Source: F. Shunt Capacitor Bank · p. 5
• The utility switches on capacitor banks during normal operation to provide reactive-power support.Source: F. Shunt Capacitor Bank · p. 5

G. Circuit Breaker

• The three feeders have nine circuit breakers for protection and reconfiguration; six are normally closed and three normally open.Source: G. Circuit Breaker · p. 5

III. A NUMERICAL EXAMPLE OF THE TEST SYSTEM

• One-year time-series power flow was run with the OpenDSS simulation program through a Matlab-OpenDSS COM interface, collecting bus voltages and line currents.Source: §III · p. 5
• Bus voltages on Feeder C’s primary main feeder ranged from 1.005 p.u. to 1.035 p.u. over the year, satisfying the voltage quality requirement.Fig. 5 presents the annual statistical results.Source: §III · p. 5
• When Bus 1 voltage fell below its setting, the regulator raised the tap position from 2 to 3, increasing voltage during a July peak-demand week.Fig. 6 shows the tap positions and voltage profiles for phases A, B, and C from July 1 to July 7.Source: §III · p. 5
• In the simulation error check, 95% of percentage-error samples were below (3 × 10−3)%.The error compares bus power calculated from solution voltages and currents with expected power; Fig. 7 shows its CDF.Source: §III · pp. 5–6

• The test system combines a real Midwest U.S. distribution grid model with real one-year smart-meter measurements, distinguishing it from existing test systems.Source: §IV · p. 6
• The paper provides the real network topologies and electric equipment parameters alongside the field measurements.Source: §IV · p. 6
• The OpenDSS model is available on the authors’ website for researchers to validate and demonstrate theoretical work using a real grid model with field measurements.Source: §IV · p. 6