Piq Energy’s Unlocking Flexibility Without Curtailment report explores how coordinated home batteries could help connect large electricity users to constrained power grids. Published on October 29, the report describes work with Base Power in which targeted battery dispatch relieved the modeled transmission overloads caused by a proposed 100 MW load near Fort Worth, Texas.
The finding is relevant to AI data centers, whose access to electricity depends on the network’s ability to deliver power as well as the availability of generation. It also puts AI in a second role: Piq uses an agentic planning platform to organize the engineering analysis needed to evaluate potential grid solutions.
Home Batteries as a Transmission-Planning Resource
Base Power develops, installs, and operates residential battery systems. Its business combines household backup with coordinated operation that can support the grid. The same battery therefore serves a local resilience need and participates in a larger managed resource.
The planning question is where that resource can make a difference. Adding demand changes flows across a network, and the most effective battery locations need not be next to the customer seeking a connection. Charging at one location and discharging at another can shift flows away from an overloaded transmission element.
ERCOT’s Aggregate Distributed Energy Resource pilot provides a framework for multiple distribution-connected sites to respond collectively to dispatch instructions. Its registration, telemetry, and qualification requirements show why coordinating a fleet is more involved than adding up the nameplate ratings of individual batteries.
What the Burleson Switch Study Found
The report examines a proposed connection at Burleson Switch in Oncor’s service area. Adding 100 MW produced eight thermal constraints across five transmission facilities. Coordinated dispatch at 11 locations resolved all eight in the model, requiring approximately 81.5 MW of gross battery dispatch.
The larger-load scenario demonstrates diminishing returns:
| Modeled new load | Gross battery dispatch | Thermal constraints resolved |
|---|---|---|
| 100 MW | 81.5 MW | 8 of 8 |
| 200 MW, efficient portfolio | 296.3 MW | 25 of 30 |
| 200 MW, full-resolution portfolio | 1,245.5 MW | 30 of 30 |
The first case supports further investigation of distributed batteries. The second shows why their usefulness depends on location and network conditions: resolving the final few constraints can require a much larger portfolio. A combination of flexible resources and conventional upgrades may be more practical.
Why 81.5 MW Is Not the Data Center’s Power Supply
The underlying technical white paper, dated September 1, distinguishes gross dispatch from net injection. In an illustrated contingency, batteries inject 43.2 MW while others withdraw 38.2 MW, leaving a net injection of 5 MW.
The batteries are redistributing flows rather than supplying the entire 100 MW load. The new customer still needs an energy supply; the modeled intervention addresses transmission overloads. Battery power, measured in MW, also differs from stored energy, measured in MWh, which determines how long a response can last.
Where Agentic AI Fits Into Grid Engineering
Piq’s technology combines grid datasets, simulation engines, and repeatable engineering workflows. Its planning agent, Delphi, helps configure studies, explore network models, compare results, and draft reports while engineers retain decision-making authority.
The company describes a system in which natural-language requests help orchestrate deterministic analysis. Parallel study execution supports more scenario testing, and results remain tied to particular model versions. That distinction is central to the AI contribution: a language model can help organize the work without replacing the power-flow calculations used to evaluate an electrical network.
In Unite.AI’s interview with co-founder and CEO Tom Nudell, he emphasized reproducibility and the work involved in coordinating specialists, defining analyses, and building useful models. This report illustrates that strategy in a specific application. Piq says it completed the project in days, compared with months it estimates a conventional approach would require; that comparison is the company’s assessment, not an independently controlled benchmark.
What Still Needs to Be Demonstrated
The technical study uses DC power-flow analysis and evaluates thermal constraints. AC power flow, voltage, stability, communications, and operational reliability require further assessment before implementation. The optimization excludes prices, so it does not establish the least-cost commercial solution.
Its numerical results apply to the studied connection point, assumed fleet, and network conditions. They cannot be transferred directly to another location. Nor does the model itself establish an approved, operating firm connection.
Base’s discussion of distributed-resource aggregation explains the commercial interest in recognizing batteries at specific transmission locations. For developers, the opportunity is to evaluate flexibility alongside infrastructure expansion. For operators, reliable dispatch and the rules governing its use remain essential.
A More Targeted Approach to Data Center Power Constraints
The report points toward a practical role for AI in infrastructure planning: making it easier to investigate combinations of resources, identify where they help, and expose where their benefits taper off. Faster analysis can improve project choices even when it concludes that batteries alone are insufficient.
Readers can review the findings in Unlocking Flexibility Without Curtailment and examine the assumptions and limitations in the accompanying technical white paper. Together, they offer a case for testing residential battery coordination as part of data center grid planning, with engineering validation determining where it can deliver useful capacity.
Credit: Source link
