[Paper Review] The Creation and Validation of Load Time Series for Synthetic Electric Power Systems

Hanyue Li, Ju Hee Yeo+2 more|arXiv (Cornell University)|Nov 16, 2019

Energy Load and Power Forecasting12 references4 citations

TL;DR

This paper presents a methodology to generate realistic, high-resolution bus-level load time series for synthetic electric power systems using publicly available data on load composition, geographic location, and solar resources. By iteratively aggregating prototypical building-level load profiles and validating against real system data, the approach produces time series that accurately reflect temporal load patterns and enable applications such as duck curve scenario analysis in large-scale synthetic grids like ACTIVSg10K.

ABSTRACT

Synthetic power systems that imitate functional and statistical characteristics of the actual grid have been developed to promote researchers' access to public system models. Developing time series to represent different operating conditions of these synthetic systems will expand the potential of synthetic power systems applications. This paper proposes a methodology to create synthetic time series of bus-level load using publicly available data. Comprehensive validation metrics are provided to assure that the quality of synthetic time series data is sufficiently realistic. This paper also includes an example application in which the methodology is used to construct load scenarios for a 10,000-bus synthetic case.

Motivation & Objective

• To develop a scalable, publicly accessible methodology for generating realistic bus-level load time series in synthetic power systems.
• To ensure the realism and statistical fidelity of synthetic load data through comprehensive validation against real-world system-level time series.
• To enable advanced power system studies by providing high-resolution (hourly, 1-year) load profiles for individual buses in synthetic networks.
• To demonstrate the utility of the methodology through a practical application involving a duck curve scenario in a 10,000-bus synthetic grid.

Proposed method

• The method uses an iterative bottom-up aggregation of prototypical load profiles (residential, commercial, industrial) based on the geographic location and load type composition of each bus.
• Bus-level load size is determined from U.S. census data (population and per-capita MW consumption), with a fixed power factor assumed.
• Solar potential is calculated as a weighted sum of normalized load size and normalized solar resource (NREL data), using equation (4).
• BTM solar capacity per bus is allocated proportionally using equation (5), based on solar potential and total system capacity (30,000 MW in ACTIVSg10K).
• A 24-hour duck curve scenario is generated by subtracting a synthetic BTM solar output time series (peaking 1–3 PM, starting 6–8 AM, ending 6–8 PM) from the original load time series.
• Validation metrics are derived from real system-level load data to ensure statistical similarity in temporal and spatial characteristics.

Experimental results

Research questions

• RQ1 How can realistic, high-resolution bus-level load time series be synthesized for large-scale synthetic power systems using only publicly available data?
• RQ2 What validation metrics are most effective in ensuring the statistical and functional realism of synthetic load time series?
• RQ3 To what extent can synthetic load time series replicate real-world phenomena such as the 'duck curve' in systems with high distributed solar penetration?
• RQ4 How can load and solar resource data be combined to distribute BTM solar capacity across buses in a geographically and economically plausible manner?

Key findings

• The proposed methodology successfully generates 10,000-bus load time series with hourly resolution over a full year, matching the statistical and temporal characteristics of real systems.
• The validation framework confirms that synthetic load time series closely mirror real system data in terms of peak values, daily patterns, and seasonal variations.
• A realistic 'duck curve' scenario was constructed in the ACTIVSg10K synthetic system by integrating BTM solar generation, with peak solar output occurring between 1–3 PM and zero output outside 6–8 AM and 6–8 PM.
• The BTM solar capacity distribution across buses is spatially coherent, with higher capacities assigned to areas with both high load density and high solar irradiance, as confirmed by contour plots.
• The method is generalizable and can be applied to any synthetic power system, regardless of geographic footprint or system size.
• The entire dataset is publicly distributable without confidentiality concerns, enabling broad use in benchmarking and algorithm testing.