Hugging Face's logo

Datasets:
xpertsystems
/
enr005-sample

Tasks:
Tabular Classification
Tabular Regression
Time Series Forecasting
Modalities:
Tabular
Text
Formats:
parquet
Size:
10K - 100K
Tags:
synthetic-data
smart-grid
ami
advanced-metering-infrastructure
smart-meter
der
Libraries:
Datasets
pandas
Polars
License:
Dataset card Data Studio Files
xet
 Community
enr005-sample / README.md
pradeep-xpert's picture
pradeep-xpert
Upload folder using huggingface_hub
d24d799 verified
preview code
|
raw
history blame contribute delete
19.1 kB
metadata 
license: cc-by-nc-4.0
task_categories:
  - tabular-classification
  - tabular-regression
  - time-series-forecasting
tags:
  - synthetic-data
  - smart-grid
  - ami
  - advanced-metering-infrastructure
  - smart-meter
  - der
  - distributed-energy-resources
  - solar-pv
  - bess
  - battery-energy-storage
  - ev-charging
  - wind-micro
  - fuel-cell
  - chp
  - demand-response
  - openadr
  - ieee-2030-5
  - ieee-1547
  - ieee-519
  - ansi-c84-1
  - ansi-c12-19
  - iec-61968
  - cim
  - two-way-power-flow
  - reverse-power-flow
  - voltage-regulation
  - volt-var
  - volt-watt
  - cvr
  - conservation-voltage-reduction
  - microgrid
  - islanding
  - grid-edge
  - grid-edge-analytics
  - power-quality
  - thd
  - harmonics
  - power-factor
  - net-metering
  - prosumer
  - hosting-capacity
  - short-circuit-ratio
  - recloser
  - sectionalizer
  - statcom
  - protection-relay
  - fault-detection-isolation
  - flisr
  - dsm
  - demand-side-management
  - tou
  - cpp
  - rtp
  - dlc
  - vpp
  - virtual-power-plant
pretty_name: ENR005 — Synthetic Smart Grid Dataset (Sample)
size_categories:
  - 10K<n<100K
configs:
  - config_name: ami_telemetry
    data_files: enr005_ami_telemetry.parquet
  - config_name: der_timeseries
    data_files: enr005_der_timeseries.parquet
  - config_name: power_flow
    data_files: enr005_power_flow.parquet
  - config_name: grid_edge
    data_files: enr005_grid_edge.parquet
  - config_name: dr_events
    data_files: enr005_dr_events.parquet
  - config_name: meter_registry
    data_files: enr005_meter_registry.parquet
  - config_name: der_registry
    data_files: enr005_der_registry.parquet

ENR005 — Synthetic Smart Grid Dataset (Sample Preview)

XpertSystems.ai | Synthetic Data Factory | Energy & Climate Vertical

A six-table smart grid AMI/DER dataset spanning the distribution-edge data stack: smart meter (AMI) telemetry, DER asset registry and timeseries (Solar PV, Battery ESS, EV chargers, Wind Micro, Fuel Cell, CHP), OpenADR-compatible demand response event logs, feeder-level two-way power flow, and grid edge device analytics (smart inverters, Volt-VAR, protection relays, microgrid islanding, power quality disturbances). Calibrated benchmark-first against IEEE 2030.5, IEEE 1547, IEEE 519, ANSI C84.1, ANSI C12.19, OpenADR 2.0b, IEC 61968 CIM, and EPRI demand response benchmarks.

This is the sample preview — 30 meters × 5 feeders × 25 DER assets × 2 days × 15-min cadence (14K records across 7 tables, ~2.6 MB). The full product covers 5,000 meters × 20 feeders × 1,500 DER assets × full annual cycle (1.4B records) with N-1 grid stress scenarios, evening EV charging surge, high-solar reverse flow events, and microgrid islanding scenarios pre-built.

Dataset summary

Table Rows (sample) What it contains
ami_telemetry 5,760 Per-meter 15-min telemetry: interval kWh, active/reactive/apparent power, voltage, current, THD, power factor, net metering, outage/tamper flags, data quality
der_timeseries 4,800 Per-DER 15-min operational data: output kW, capacity, SOC (BESS/EV), charge/discharge rates, solar irradiance, inverter status, availability, grid-forming flag
dr_events ~174 OpenADR event participation: program type, signal type, baseline, target/actual reduction, performance %, rebound, opt-out, comfort score, incentive paid
power_flow 960 Feeder-level two-way flow: direction (FORWARD/REVERSE/BIDIRECTIONAL), gross/net/reverse load, voltage regulation, voltage rise from DER, feeder loading %, reactive kVAR, distribution losses
grid_edge 2,400 Edge device ops: Smart Inverter / STATCOM / Volt-VAR / Recloser / Sectionalizer / Cap Bank / DSTATCOM, microgrid mode, islanding detection, protection relay events, PQ disturbance type, hosting capacity, SCR
meter_registry 30 Static meter metadata: feeder assignment, customer class, DER ownership flags, vintage year, tariff code, service voltage, contract demand
der_registry 25 Static DER metadata: type, capacity, install year, manufacturer, interconnection voltage, IEEE 1547 compliance, smart inverter capability

All seven tables are provided in both CSV and Parquet. They join on meter_id, feeder_id, der_id, and timestamp_utc.

Calibration sources

All ten validation metrics target named industry sources, not generator self-metrics:

  • IEEE 2030.5 — smart energy profile (DER/EV/meter communications, microgrid normal operations)
  • IEEE 1547 — interconnection of distributed energy resources (service voltage band, IEEE 1547-2018 anti-islanding)
  • IEEE 519-2014 — harmonic control in electrical power systems (THD limits)
  • ANSI C84.1 — Electric Power Systems and Equipment — Voltage Ratings (Range A 114-126V residential, ±5% from 120V nominal)
  • ANSI C12.19 — utility industry end device data tables (AMI metering data quality KPIs)
  • OpenADR 2.0b — demand response signaling protocol (event lifecycle, enrollment, performance reporting)
  • IEC 61968 CIM — common information model for distribution management
  • EPRI Demand Response Program Performance Benchmarks — typical realized reduction ratio 60-90% of targeted reduction
  • Physics — DER nameplate capacity envelope, BESS SOC bounds [0, 100]%, mass balance (net = gross − DER generation)

Validation scorecard (seed = 42)

10/10 PASS · Grade A+ (100%) across all six canonical seeds (42, 7, 123, 2024, 99, 1).

# Metric Observed Target Tol Type Source
1 residential_voltage_ansi_C84_range_A_rate 1.000 0.98 ±0.03 FLOOR ANSI C84.1
2 power_factor_floor_0_80_rate 1.000 0.99 ±0.01 FLOOR IEEE 519 / tariff
3 thd_ieee_519_8pct_compliance_rate 0.918 0.90 ±0.05 FLOOR IEEE 519-2014
4 bess_soc_in_valid_range_rate 1.000 0.99 ±0.01 FLOOR Physics
5 der_within_capacity_envelope_rate 1.000 0.99 ±0.01 FLOOR Physics
6 feeder_voltage_pu_ansi_band_rate 1.000 0.95 ±0.05 FLOOR ANSI C84.1 / IEEE 1547
7 power_flow_net_equals_gross_minus_der_rate 1.000 0.99 ±0.01 FLOOR Mass balance
8 microgrid_grid_connected_share 0.987 0.98 ±0.03 two-sided IEEE 2030.5 / IEEE 1547-2018
9 dr_performance_mean_pct 73.96 75.0 ±10.0 two-sided EPRI DR Performance
10 ami_data_quality_valid_rate 0.968 0.95 ±0.05 FLOOR ANSI C12.19 / AMI KPI

Schema highlights

ami_telemetry (5,760 rows × 16 cols)

meter_id, feeder_id, customer_class (Residential / Commercial_SMB / Commercial_Large / Industrial), timestamp_utc, interval_kwh, net_metering_kwh, active_power_kw, reactive_power_kvar, apparent_power_kva, power_factor, voltage_v, current_a, thd_pct, outage_flag, tamper_flag, data_quality_flag (VALID / ESTIMATED / MISSING / SUBSTITUTED).

der_timeseries (4,800 rows × 11 cols)

der_id, der_type (Solar_PV / Battery_ESS / EV_Charger / Wind_Micro / Fuel_Cell / CHP), timestamp_utc, der_output_kw (signed: + export, − import for EVs), der_capacity_kw, state_of_charge_pct (BESS/EV only), charge_rate_kw, discharge_rate_kw, solar_irradiance_w_m2, inverter_status (ONLINE / OFFLINE / CURTAILED / FAULT / MPPT), der_availability_flag, grid_forming_flag.

dr_events (~174 rows × 17 cols)

dr_event_id, meter_id, dr_program_type (TOU / CPP / RTP / DLC / AutoDR / VPP_Dispatch), dr_signal_type (SHED / SHIFT / MODULATE / PRICE_SIGNAL / EMERGENCY), dr_event_start_utc, dr_event_end_utc, duration_hours, baseline_kw, dr_target_reduction_kw, actual_reduction_kw, dr_performance_pct, non_event_baseline_kw, rebound_kw, dr_incentive_paid_usd, customer_opt_out_flag, comfort_score, customer_class.

power_flow (960 rows × 14 cols)

feeder_id, timestamp_utc, power_flow_direction (FORWARD / REVERSE / BIDIRECTIONAL), gross_load_kw, der_generation_kw, net_load_kw, reverse_flow_kw, feeder_loading_pct, voltage_regulation_pu, voltage_rise_pu, reactive_kvar, conservation_voltage_reduction_factor, substation_load_kw, distribution_losses_pct, thermal_rating_kw.

grid_edge (2,400 rows × 13 cols)

feeder_id, device_id, edge_device_type (Smart_Inverter / STATCOM / Volt_VAR / Recloser / Sectionalizer / Cap_Bank / DSTATCOM), timestamp_utc, volt_var_mode (VOLT_VAR / VOLT_WATT / CONSTANT_PF / FIXED_Q), reactive_power_support_kvar, feeder_automation_status (MANUAL / AUTOMATIC / FASR / FLISR), islanding_detection_flag, microgrid_mode (GRID_CONNECTED / ISLANDED / TRANSITIONING), protection_relay_event (PICKUP / TRIP / RECLOSE / LOCKOUT / NORMAL), pq_disturbance_type (SAG / SWELL / INTERRUPTION / HARMONIC / FLICKER / NORMAL), hosting_capacity_kw, short_circuit_ratio.

Suggested use cases

  • AMI load disaggregation — train NILM (non-intrusive load monitoring) models to separate base load, HVAC, EV charging, and solar from interval_kwh and net_metering_kwh
  • DER forecasting — short-horizon prediction of der_output_kw for solar PV from irradiance + cloud features; wind micro turbine output from synthetic wind speeds
  • BESS dispatch optimization — learn charge/discharge policies from bess_soc_pct, time-of-day, and feeder loading signals
  • EV charging behavior modeling — predict charging session start time, duration, and energy from customer class, charge rate, and historical patterns
  • Demand response performance prediction — regressor for dr_performance_pct and actual_reduction_kw from baseline, signal type, comfort score, and customer class
  • VPP / aggregator dispatch ML — train portfolio-level DR fleet optimizers across heterogeneous customer classes
  • Two-way power flow classification — predict power_flow_direction from DER generation, gross load, and feeder topology features
  • Voltage rise modeling on high-PV feeders — ML to predict voltage_rise_pu from der_generation_kw, R/X impedances, and feeder loading
  • Volt-VAR control optimization — train smart inverter Volt-VAR curves from PCC voltage and feeder loading conditions
  • Microgrid islanding detection — anomaly/classification on islanding_detection_flag, protection_relay_event, pq_disturbance_type joint signals
  • Hosting capacity prediction — regressor for hosting_capacity_kw given short-circuit ratio, R/X, and existing DER penetration
  • Power quality classification — multi-class on pq_disturbance_type (SAG / SWELL / INTERRUPTION / HARMONIC / FLICKER) from voltage and THD time-series features
  • AMI data quality / outage detection — classifier for outage_flag, tamper_flag, and data_quality_flag transitions
  • Net metering / prosumer billing analytics — model bill components from net_metering_kwh, TOU tier (joinable from registry), and retail rate schedules

Loading examples 

from datasets import load_dataset

ami = load_dataset("xpertsystems/enr005-sample", "ami_telemetry", split="train")
der = load_dataset("xpertsystems/enr005-sample", "der_timeseries", split="train")
flow = load_dataset("xpertsystems/enr005-sample", "power_flow", split="train")
print(ami.shape, der.shape, flow.shape)

import pandas as pd
from huggingface_hub import hf_hub_download

# Load AMI + DR events; join on meter_id
ami = pd.read_parquet(hf_hub_download(
    "xpertsystems/enr005-sample", "enr005_ami_telemetry.parquet",
    repo_type="dataset",
))
dr = pd.read_parquet(hf_hub_download(
    "xpertsystems/enr005-sample", "enr005_dr_events.parquet",
    repo_type="dataset",
))

# Per-class DR performance
print(dr.groupby("customer_class")["dr_performance_pct"].agg(["mean", "std", "count"]))

# Compute net feeder demand profile
import pandas as pd
from huggingface_hub import hf_hub_download

flow = pd.read_parquet(hf_hub_download(
    "xpertsystems/enr005-sample", "enr005_power_flow.parquet",
    repo_type="dataset",
))
flow["t"] = pd.to_datetime(flow["timestamp_utc"])
flow["hour"] = flow["t"].dt.hour
hourly = flow.groupby("hour").agg(
    gross=("gross_load_kw", "mean"),
    der=("der_generation_kw", "mean"),
    net=("net_load_kw", "mean"),
).round(2)
print(hourly)  # see the duck-curve shape

# Check power flow direction transitions per feeder
import pandas as pd
from huggingface_hub import hf_hub_download

flow = pd.read_parquet(hf_hub_download(
    "xpertsystems/enr005-sample", "enr005_power_flow.parquet",
    repo_type="dataset",
))
for fid, g in flow.groupby("feeder_id"):
    g = g.sort_values("timestamp_utc")
    transitions = (g["power_flow_direction"] != g["power_flow_direction"].shift()).sum() - 1
    print(f"Feeder {fid[:8]}: {transitions} direction transitions, "
          f"reverse rate = {(g['reverse_flow_kw'] > 1).mean():.3f}")

Limitations and honest disclosures

This sample is calibrated for structural fidelity, not bit-exact reproduction of any specific utility's AMI archive. Specifically:

  • AMI/DER/power_flow tables auto-truncate to 2 days when hours < 8760 in the generator's main() (lines 791-794, 808-810, 829-830); grid_edge truncates to 1 day (line 839). At this sample's hours=48 setting, AMI/DER/flow all cover the full 48h while grid_edge covers 24h. Time ranges across tables don't align 1:1 — be aware when joining grid_edge with other tables.
  • Power factor distribution clips at 0.80 — generator uses Beta(8,2) clipped to [0.80, 1.00], so values below 0.80 are pushed up to exactly 0.80 (heavy mass at the lower bound). This matches utility residential tariff minimums, but real Commercial_Large and Industrial loads can dip to 0.70-0.85 without correction. Don't use this dataset to study deep low-PF events.
  • THD distribution has fat upper tails. The generator's thd_base + exponential(1.5) shifts ~60% of rows above the IEEE 519 5% residential limit, with ~8% above the 8% short-duration bound. For PQ research, treat these as bus-level coupled THD, not point-of-common- coupling compliance measurements.
  • Distribution losses average 0.46% (Beta(2,20)×5), well below the US industry typical 4-7% range. The generator's loss distribution is intentionally narrow for compact statistical learning; do not use distribution_losses_pct for absolute loss benchmarking.
  • Reverse power flow at sample scale fires rarely. At ~6 meters per feeder × default 35% solar penetration, DER generation seldom exceeds gross load. The full product activates reverse flow via the High_Solar_Day scenario config (solar_penetration=0.65). Wrapper validates the STRUCTURAL identity (reverse_flow_kw = max(0, -net_load_kw) and net_load_kw = gross_load_kw - der_generation_kw) rather than an aggregate rate.
  • hosting_capacity_kw and short_circuit_ratio are sampled per-row (not per-device-or-feeder properties). Use as advisory features in ML pipelines, not as static topology attributes.
  • interconnect_status (CONNECTED / CURTAILED / TRIPPED / ISLANDED) on grid_edge is sampled independently of generation flow at the feeder level. Treat as device-state telemetry, not a causal label for a flow event.
  • solar_irradiance_w_m2 is a simplified mid-US-latitude clear-sky model (latitude hardcoded to 37.5°). All meters share the same solar noon and seasonal cycle. Do not use for geographically- varying PV studies — use the full product or join with ENR-002 for per-site irradiance.
  • DR event participation uses an O(N) pandas lookup per participant in the generator's generate_dr_events (line 426-427). At sample scale (30 meters × 12 events × ~50% enrollment ≈ 180 participants) it's fast (< 0.1s); the full product (5000 meters × 48 events × ~50% enrollment ≈ 120,000 participants) runs slower. Not a data-quality issue, just a perf note.
  • Solar PV der_output_kw is non-negative; EV charger der_output_kw is non-positive (load convention). When aggregating "DER export" use der_output_kw.where(der_output_kw > 0).sum() to avoid loads canceling generation.
  • dr_events.start_utc references the full 12-month timestamp range via the generator's months=12 hardcode (line 401), but the timestamp draw is bounded by len(timestamps) - 16, so at sample scale all events fall within the AMI window. This is by design but is something to be aware of when adapting the generator.

The full ENR005 product addresses these by per-site latitude-aware irradiance, broader PF / THD distributions, reverse-flow and EV-surge scenarios pre-built, multi-month time spans, and full meter/DER fleet scale — contact us for the licensed commercial release.

Companion datasets in the Energy & Climate vertical

  • ENR-001 — Synthetic Power Grid Operations Dataset (transmission-side bus telemetry, line flows, generation dispatch, frequency, contingency)
  • ENR-002 — Synthetic Renewable Energy Generation Dataset (utility- scale solar/wind/hybrid SCADA, weather, forecast, PCC, BESS)
  • ENR-003 — Synthetic Electricity Demand & Load Forecasting Dataset (zone-level demand, multi-horizon forecasts, peak events, EV/DER, TOU, LMP)
  • ENR-004 — Synthetic Upstream Oil & Gas Production Dataset (well- level production, decline curves, PVT, commodity prices, Subpart W methane)
  • ENR-005 — Synthetic Smart Grid Dataset (you are here) — the distribution-edge complement to ENR-001's transmission focus: meter- level AMI, behind-the-meter DER, OpenADR demand response, feeder power flow, and grid edge analytics.

Use ENR-001 + ENR-005 together for full transmission + distribution grid ML workflows; combine with ENR-002 + ENR-003 to add renewables and demand forecasting in the same modeling stack.

For subsurface companion data (seismic, well logs, reservoir simulation, geological formations), see the OIL series (OIL-001 through OIL-004) in our Oil & Gas vertical.

For the broader catalog:

Citation 

@dataset{xpertsystems_enr005_sample_2026,
  author       = {XpertSystems.ai},
  title        = {ENR005 Synthetic Smart Grid Dataset (Sample Preview)},
  year         = 2026,
  publisher    = {Hugging Face},
  url          = {https://huggingface.co/datasets/xpertsystems/enr005-sample}
}

Contact

Sample License: CC-BY-NC-4.0 (Creative Commons Attribution-NonCommercial 4.0) Full product License: Commercial — please contact for pricing.