Discrete-time linear quadratic stochastic control with equality-constrained inputs: Application to energy demand response
A novel algorithm uses dynamic programming and KKT conditions to manage energy demand with hard equality constraints.
A team of researchers, Leo Seugnet and Shuang Gao, has published a significant paper on arXiv titled 'Discrete-time linear quadratic stochastic control with equality-constrained inputs: Application to energy demand response.' The work tackles a core problem in systems control: managing a population of cooperative agents—like thousands of residential batteries—when their total combined action must meet a precise, hard equality constraint. This is directly motivated by the challenge of demand response in renewable energy grids, where supply from sources like solar must be perfectly balanced with controllable demand.
The researchers established the optimal control solution for systems with additive noise, deriving the control law through a combination of dynamic programming and Karush-Kuhn-Tucker (KKT) conditions. The result is a control policy governed by a discrete-time, Riccati-like recursive equation. In a key application example, they demonstrate how this algorithm can coordinate the charging of a distributed battery network to absorb unpredictable excess solar power generation. Critically, the method achieves exact power tracking to balance the grid while simultaneously respecting the individual State-of-Charge (SoC) objectives and limits of each battery, preventing any single unit from being overused. The 7-page paper has been accepted for publication at the prestigious American Control Conference (ACC).
- Solves stochastic LQ control with hard equality constraints on total input, crucial for grid balance.
- Algorithm uses dynamic programming + KKT conditions, resulting in a Riccati-like recursive equation.
- Enables exact power tracking for demand response, coordinating 1000s of batteries while managing individual SoC.
Why It Matters
Provides a scalable, mathematical framework for precisely managing distributed energy resources, which is essential for integrating more renewable solar and wind power into the grid.