TL;DR
This paper proposes a hierarchical framework for distributed energy resources (DERs), organized as microgrids, to participate in frequency regulation markets. It introduces methods for DERs to bid into these markets and a distributed algorithm for efficiently allocating regulation signals, ensuring economic dispatch while respecting operational constraints.
What Problem Does This Paper Address?
The increasing integration of Distributed Energy Resources (DERs) like solar panels and batteries into the power grid presents an opportunity for them to provide ancillary services, particularly frequency regulation. Frequency regulation is crucial for maintaining grid stability by ensuring generation and load are balanced in real-time. Traditionally, this service has been provided by large, centralized power plants. However, DERs, while offering benefits like reduced emissions and faster response times due to their inertialess nature, face challenges in participating in these markets. These challenges include:
- Market Entry Barriers: Individual DERs might be too small to meet minimum size requirements for market participation.
- Information Asymmetry: System operators (like Regional Transmission Organizations - RTOs) need aggregated information about DER capabilities (capacity, cost, ramp rates) to clear the market and allocate signals.
- Coordination Complexity: Effectively disaggregating the regulation signal from the RTO to individual DERs within a microgrid, while considering their dynamic operational constraints (like ramp rates and power flow limits), is complex.
- Uncertainty: DER availability and load within microgrids can fluctuate, making it difficult to guarantee performance and bid accurately.
This paper aims to bridge this gap by providing a framework that allows DERs, aggregated into microgrids, to participate effectively in frequency regulation markets by addressing these challenges.
Methods
The paper proposes a hierarchical framework with two layers:
- Top Layer (Aggregators): Aggregators, representing microgrids, communicate in a distributed manner. They are responsible for determining their bids for the market clearing stage and for disaggregating the regulation signal received from the RTO.
- Bottom Layer (DERs within Microgrids): Individual DERs within a microgrid adjust their power output to meet the aggregated signal from their aggregator, respecting power flow constraints within the microgrid.
Key methodological contributions include:
- Meaningful Abstractions: The authors develop abstractions for microgrid capabilities, including:
- Capacity Bounds: Determining the maximum up/down regulation a microgrid can provide, considering both deterministic and probabilistic (chance-constrained) scenarios for variable loads. This involves solving optimization problems that respect power flow constraints.
- Cost Functions: Abstracting the cost of providing regulation, accounting for the operating points of controllable DERs and the impact of previous regulation signals (considering ramp rates).
- Ramp Rate Functions: Characterizing how quickly a microgrid can change its power output, which is crucial for real-time tracking and depends on the network topology and operating point.
- Distributed Coordination Algorithm (gdac): A novel distributed algorithm that combines gradient descent with dynamic average consensus. This algorithm allows aggregators to optimally disaggregate the regulation signal from the RTO. It addresses the challenge of limited information by enabling aggregators to estimate the required regulation from others and ensures that only one aggregator needs to know the total required regulation. The algorithm is proven to asymptotically converge to the optimal solution.
Datasets
The simulations in this paper utilize a reduced-order model of the University of California, San Diego (UCSD) microgrid. This model is based on a larger, real-world microgrid and is adapted to create 12 different microgrids, grouped into four categories based on generation and load characteristics. The paper also refers to the PJM (Pennsylvania-New Jersey-Maryland) Interconnection's dynamic regulation test signal (RegD) for simulation purposes.
Key Insights and Ideas
- Hierarchical Control is Key: Breaking down the complex problem of DER participation into manageable layers (RTO, aggregators, DERs) is essential for scalability and efficiency.
- Abstraction is Crucial for Market Participation: DERs cannot directly reveal their internal complexities to the market. Developing robust abstractions for capacity, cost, and ramp rates allows them to participate effectively in market clearing.
- Probabilistic Bidding for Robustness: Accounting for load uncertainties using chance-constrained optimization provides more realistic and less conservative capacity bids, enabling greater participation.
- Distributed Algorithms for Scalability: Centralized solutions for signal disaggregation are not scalable. A distributed approach, like the proposed gdac algorithm, is necessary for coordinating a large number of microgrids.
- Economic Dispatch with Ramp Constraints: The paper highlights that simply meeting the regulation signal might not be economically optimal or even feasible due to ramp rate limitations. The proposed algorithm explicitly optimizes for economic dispatch while respecting these constraints.
Key Results
- Feasible Market Participation: The proposed abstractions and framework enable microgrids to generate meaningful bids for market clearance.
- Efficient Signal Disaggregation: The gdac algorithm demonstrates efficient and distributed disaggregation of regulation signals. Simulations show it converges significantly faster than existing saddle-point dynamics methods for similar problems.
- Cost Savings: In simulations, the proposed algorithm achieved a lower cost ($9104) compared to current practices ($9638) for the same regulation task, attributed to more optimal allocation of contributions based on abstracted cost functions.
- Robustness to Uncertainty: The chance-constrained approach for capacity bounds leads to more realistic bids and allows for participation even with variable loads.
Limitations and Future Work
- Network Topology: The current analysis primarily focuses on tree networks for ramp rate characterization, with some discussion for networks with non-overlapping loops. Future work aims to extend this to more general network topologies.
- AC Power Flow: The paper uses simplified AC power flow equations. Future work will incorporate full AC power flow equations.
- Non-Smooth Dynamics: The proposed distributed algorithm exhibits non-smooth dynamics, which can lead to chattering. Future research will investigate smooth distributed algorithms to mitigate this.
- Initialization: While the algorithm's convergence is proven, specific initial conditions are required for optimal performance. The paper discusses practical initialization strategies.
- Market Strategy: The current bidding strategy is conservative, focusing on feasibility rather than profit maximization. Future work could explore more strategic bidding.
Summary updated