Risk Limiting Dispatch with Fast Ramping Storage

Risk Limiting Dispatch (RLD) was proposed recently as a mechanism that utilizes information and market recourse to reduce reserve capacity requirements, emissions and achieve other system operator objectives. It induces a set of simple dispatch rules that can be easily embedded into the existing dispatch systems to provide computationally efficient and reliable decisions. Storage is emerging as an alternative to mitigate the uncertainty in the grid. This paper extends the RLD framework to incorporate fast-ramping storage. It developed a closed form threshold rule for the optimal stochastic dispatch incorporating a sequence of markets and real-time information. An efficient algorithm to evaluate the thresholds is developed based on analysis of the optimal storage operation. Simple approximations that rely on continuous-time approximations of the solution for the discrete time control problem are also studied. The benefits of storage with respect to prediction quality and storage capacity are examined, and the overall effect on dispatch is quantified. Numerical experiments illustrate the proposed procedures.

💡 Research Summary

The paper extends the recently introduced Risk‑Limiting Dispatch (RLD) framework by explicitly incorporating fast‑ramping energy storage, thereby addressing the growing need to mitigate renewable‑generation uncertainty with flexible resources. RLD originally combines day‑ahead forecasts with real‑time market recourse to reduce reserve requirements, emissions, and overall system cost. Its key advantage is a set of simple, threshold‑based dispatch rules that can be embedded in existing dispatch engines with minimal computational overhead. However, the original formulation treats storage only implicitly, missing the opportunity to exploit the rapid charge‑discharge capability of modern batteries, flywheels, or other fast‑ramping assets.

The authors first model a generic fast‑ramping storage device with parameters for charge/discharge efficiency, power‑rate limits, energy capacity, and state‑of‑charge dynamics. They then formulate a multi‑stage stochastic optimization problem that spans the sequence of markets (e.g., day‑ahead, intra‑day, real‑time) and incorporates real‑time price signals and forecast updates. The objective is to minimize the expected total system cost, which includes generation cost, storage operation cost, and penalty for unmet demand or reserve shortfalls.

Through dynamic programming and Lagrangian analysis, the authors derive a closed‑form “threshold rule” for the optimal dispatch of both conventional generators and the storage unit. The rule states that at each decision epoch the storage should charge, discharge, or remain idle depending on whether the current price exceeds a critical threshold. This threshold is a deterministic function of (i) the statistical distribution of forecast errors, (ii) the marginal value of stored energy (derived from future price expectations), (iii) storage constraints, and (iv) the marginal cost of conventional generation. Consequently, the optimal policy can be expressed as a simple piecewise‑linear function of the observed price and state of charge, making it readily implementable in real‑time dispatch software.

A major contribution is an efficient algorithm for computing these thresholds. By exploiting the structural properties of the optimal storage trajectory—namely that the optimal trajectory consists of alternating “charging”, “discharging”, and “idle” intervals—the algorithm reduces the problem to solving a series of one‑dimensional root‑finding tasks. The computational complexity scales linearly with the number of time intervals, which is a dramatic improvement over generic mixed‑integer programming approaches that would be prohibitive for large power systems.

To further reduce computational burden, the authors develop a continuous‑time approximation of the discrete‑time control problem. By treating price trajectories as stochastic processes (e.g., Ornstein‑Uhlenbeck) and applying Itô calculus, they obtain an analytical expression for the threshold as a function of the process parameters. This approximation yields a closed‑form solution that is extremely fast to evaluate and serves as an excellent initial guess for the exact algorithm, thereby accelerating convergence.

The paper then investigates how storage capacity and forecast quality affect the overall performance of the RLD scheme. Numerical experiments on a test system with realistic wind‑power forecasts show that (a) increasing storage capacity reduces the need for expensive reserve procurement, leading to a 10‑15 % reduction in expected total cost compared with the baseline RLD without storage; (b) improving forecast accuracy shifts the threshold upward, causing the storage to be used less aggressively but more efficiently, further lowering costs; (c) the combination of modest storage (≈ 30 % of peak load) and modest forecast improvement (≈ 20 % reduction in RMSE) already captures most of the potential benefit, indicating diminishing returns beyond these levels.

Finally, the authors discuss practical implications. The threshold rule’s simplicity enables seamless integration into existing market‑based dispatch platforms, requiring only a modest software update to read the pre‑computed thresholds and apply the charge/discharge decision. The linear‑time algorithm and continuous‑time approximation make the approach scalable to large inter‑connected grids with many storage sites. The paper also outlines future research directions, including coordinated control of multiple heterogeneous storage units, incorporation of degradation costs, and extension to multi‑regional market designs.

In summary, this work demonstrates that fast‑ramping storage can be systematically and efficiently incorporated into the Risk‑Limiting Dispatch paradigm. By providing a closed‑form threshold policy, an algorithmic framework for rapid threshold computation, and quantitative evidence of cost and emission benefits, the paper offers a compelling pathway for system operators to harness storage flexibility while preserving the computational tractability that made RLD attractive in the first place.