⚡ electrical engineering

Risk-Aware Multi-Market Scheduling of Virtual Power Plants with Dynamic Network Tariffs

This paper proposes a two-stage stochastic optimization framework for Virtual Power Plants that integrates detailed device and network constraints with risk-aware multi-market bidding and dynamic network tariffs, demonstrating that while dynamic tariffs unlock local flexibility, strong tariff signals can significantly reduce expected profitability despite mitigating profit volatility.

Original authors: Lorenzo Zapparoli, Paul Fäth, Blazhe Gjorgiev, Giovanni Sansavini

Published 2026-04-30

📖 5 min read🧠 Deep dive

CC BY 4.0

Original authors: Lorenzo Zapparoli, Paul Fäth, Blazhe Gjorgiev, Giovanni Sansavini

Original paper licensed under CC BY 4.0 (http://creativecommons.org/licenses/by/4.0/). ✨ This is an AI-generated explanation of the paper below. It is not written or endorsed by the authors. For technical accuracy, refer to the original paper. Read full disclaimer

Imagine a Virtual Power Plant (VPP) not as a giant factory with smokestacks, but as a smart, super-organized neighborhood manager. This manager doesn't own a single power plant; instead, they control a collection of everyday things in a neighborhood: rooftop solar panels, electric car chargers, home batteries, and heat pumps.

The goal of this paper is to figure out the best way for this "neighborhood manager" to sell electricity and services to the grid without getting into trouble, while dealing with two big headaches: uncertainty (we don't know exactly how much sun will shine or how many people will charge their cars) and changing rules (the price the grid charges for using its wires changes throughout the day).

Here is how the paper breaks it down, using simple analogies:

1. The Two-Stage Game: Planning vs. Acting

The researchers designed a "two-stage" game plan for the VPP, similar to planning a road trip:

Stage 1 (The Day Before): The manager makes a bet. They look at the weather forecast and price predictions and say, "I promise to sell 100 units of power tomorrow at this price, and I promise to have 50 units ready to help the grid if it gets shaky." They lock these bets in the market.
Stage 2 (The Day Of): Reality hits. Maybe the sun was weaker than expected, or a neighbor plugged in their EV earlier than planned. Now, the manager has to act. They decide which batteries to drain, which heat pumps to turn down, and how to fix any mistakes (called "imbalances") to avoid paying fines.

The paper uses a mathematical framework to make sure the bets made in Stage 1 are smart enough to handle the surprises in Stage 2.

2. The Risky Business: Being "Risk-Neutral" vs. "Risk-Averse"

The paper compares two different personalities for the neighborhood manager:

The Risk-Neutral Manager (The Gambler): This manager says, "I'll take the biggest gamble to make the most money." They bet heavily on selling power when prices look high.
- The Result: When the gamble pays off, they make a lot of money. But when the weather or user behavior surprises them, they end up with huge penalties for not delivering what they promised. Their daily profit is like a rollercoaster—wild swings up and down.
The Risk-Averse Manager (The Conservative): This manager says, "I'd rather make a little less money than lose a lot." They use a safety tool called CVaR (Conditional Value-at-Risk), which is like an insurance policy against the worst-case scenarios.
- The Result: They make less money on average, but their daily profit is very stable. They avoid the "rollercoaster" by keeping their promises closer to what they can actually physically deliver.

The Trade-off: The study found that being "risk-averse" costs the manager about 264 CHF in potential profit, but it saves them from the stress and financial danger of wild profit swings. To achieve this safety, the risk-averse manager uses their batteries much more often (146% more) to smooth things out, acting like a shock absorber for the neighborhood.

3. The Dynamic Tariff: The "Happy Hour" for Electricity

The paper also tests what happens when the grid operator introduces Dynamic Network Tariffs. Think of this as the grid operator changing the price of "using the road" (the wires) throughout the day.

The Signal: "It's cheap to drive on our roads between 10 AM and 2 PM (low tariffs), but it's very expensive between 5 PM and 9 PM (high tariffs)."
The Reaction: The VPP manager tries to shift the neighborhood's energy use to the "cheap hours." They charge electric cars and run heat pumps during the day and slow down in the evening.
The Catch: The study found a "Goldilocks" zone.
- Small price changes (10–20% difference) successfully get people to shift their habits without hurting the VPP's profits much.
- Huge price changes (up to 100% difference) force people to shift, but it actually destroys the VPP's profit (by up to 65%) because the cost of buying/selling power becomes too chaotic. The "flexibility" gained isn't worth the financial damage.

4. The Secret Sauce: Benders Decomposition

The math behind this is incredibly complex because the manager has to track 97 different locations, hundreds of devices, and 1,000 different "what-if" scenarios all at once. A normal computer would crash trying to solve this.

The authors used a trick called Benders Decomposition. Imagine trying to solve a giant jigsaw puzzle. Instead of one person trying to fit all 10,000 pieces at once, you split the puzzle into:

The Master Problem: One person decides the big picture (the bets).
The Subproblems: Many people work on small, separate sections (the specific scenarios) in parallel.
They pass notes back and forth until the whole picture fits. This allowed the researchers to run a very detailed, realistic simulation that would have been impossible otherwise.

Summary

The paper concludes that:

Safety pays off: Being cautious (risk-averse) makes the Virtual Power Plant more reliable and less volatile, even if it means slightly less profit.
Batteries are key: To stay safe, the VPP needs to use its batteries more aggressively to balance out mistakes.
Don't overdo the price signals: Grid operators can encourage people to use power at different times by changing prices, but if the price changes are too extreme, it hurts the system's profitability without gaining much extra flexibility.

In short, the paper provides a blueprint for how to run a neighborhood energy network that is smart, safe, and profitable, even when the weather and human behavior are unpredictable.

1. Problem Statement

The paper addresses the challenge of scheduling Virtual Power Plants (VPPs) composed of Distributed Energy Resources (DERs) in an increasingly complex electricity market environment. As renewable penetration grows, VPPs must navigate:

Multi-Market Participation: Simultaneous participation in sequential markets (Day-Ahead Energy, Reserve Capacity, and Reserve Activation/Imbalance markets).
Uncertainty: Volatility in renewable generation, load demand, user behavior (e.g., EV charging), and market prices.
Network Constraints: The need to respect physical grid limitations (voltage and thermal limits) at the distribution level, which are often ignored in aggregated models.
Dynamic Network Tariffs: The emerging regulatory mechanism where Distribution System Operators (DSOs) use time-varying tariffs to manage local congestion, adding a layer of price signals distinct from market prices.

Research Gaps: Existing literature often relies on highly aggregated DER models, simplified grid topologies, or ignores the joint impact of dynamic tariffs and market bidding under uncertainty. Furthermore, few models explicitly incorporate risk preferences (risk-aversion) alongside detailed device-level constraints and network physics.

2. Methodology

The authors propose a two-stage stochastic optimization framework designed to co-optimize market bidding and physical dispatch.

A. Mathematical Formulation

Decision Structure:
- Stage 1 (Day-Ahead): The VPP operator determines bids for the Day-Ahead Energy Market and Reserve Capacity Market. These decisions are made before uncertainty is realized.
- Stage 2 (Delivery Day): The operator dispatches individual DERs, bids in the Reserve Activation market, and manages imbalances based on realized scenarios (prices, generation, load).
Objective Function: The model minimizes a risk functional $R[\cdot]$ $R [\cdot]$ applied to the total operating cost.
- Risk-Neutral: Minimizes Expected Value ( $E[\cdot]$ ).
- Risk-Averse: Minimizes Conditional Value-at-Risk (CVaR) at a confidence level $\alpha$ (set to 0.9). This focuses on minimizing the average cost of the worst-case scenarios (tail risk).
Cost Components: The objective function accounts for revenues (Energy, Reserve Capacity, Reserve Activation) and costs (DER operations, network tariffs, imbalance penalties).

B. Constraints

Financial Constraints: Ensure consistency between market commitments and physical dispatch. They link day-ahead bids, reserve offers, and actual imbalances.
Physical Constraints:
- Network: Modeled using linearized DistFlow equations to capture voltage and flow limits on a radial distribution network.
- Device-Level: Detailed models for four DER types:
  - Distributed Generators (DG): Capacity factors based on weather.
  - Heat Pumps (HP): Thermal dynamics (temperature, capacitance) to maintain comfort.
  - Electric Vehicles (EV): Charging events, state-of-charge (SoC) constraints, and Vehicle-to-Grid (V2G) capabilities.
  - Battery Energy Storage (BESS): Charging/discharging efficiencies and SoC constraints.
Uncertainty Modeling: Uses a scenario-based approach with 1,000 scenarios generated via Latin Hypercube Sampling. Errors are modeled for solar irradiance, ambient temperature, EV behavior, load profiles, and market prices.

C. Solution Algorithm

Benders Decomposition: To handle the computational intractability of a monolithic stochastic program with thousands of variables and constraints, the authors employ Benders decomposition.
- Master Problem: Optimizes first-stage variables (bids) and risk variables (CVaR).
- Subproblems: Solve for second-stage variables (dispatch) for each scenario independently.
- Benefit: This allows parallel computation and makes the problem solvable with large scenario sets and detailed grid models that would otherwise exceed memory limits.

3. Key Contributions

Integrated Risk-Aware Framework: A novel two-stage stochastic model that explicitly incorporates CVaR to manage financial risk while optimizing across energy and ancillary service markets.
High-Resolution Modeling: Unlike previous studies using aggregated units, this model uses device-level constraints for heterogeneous DERs (PV, HP, EV, BESS) coupled with linearized power flow equations for a realistic low-voltage network.
Dynamic Tariff Analysis: The first study to jointly analyze the impact of dynamic network tariffs and market participation on VPP profitability and flexibility provision under uncertainty.
Scalable Solution Method: Demonstrates the feasibility of solving large-scale stochastic programs (1,000 scenarios, ~41M variables) using Benders decomposition, overcoming the memory limitations of extensive-form formulations.

4. Results

The model was tested on a realistic case study of a Swiss low-voltage network (97 buses, 150kW peak load, 150kWp PV, 85kW HPs, 75kWh BESS, 40 EVs).

A. Impact of Risk Aversion

Profit vs. Stability: The risk-neutral strategy yielded a higher expected profit (22.1 CHF) but with high volatility (207.5 CHF). The risk-averse strategy (CVaR) reduced volatility significantly (31.0 CHF) but resulted in an expected loss (-242.3 CHF) due to conservative bidding.
Operational Shift: The risk-averse strategy aligned bids more closely with physical dispatch to avoid imbalance penalties.
- BESS Utilization: Increased by 146% compared to the risk-neutral case to hedge against imbalances.
- EV & HP: Slight increases in cycling and consumption to create thermal flexibility.
Reserve Capacity: The VPP did not bid in the Reserve Capacity market due to unfavorable price incentives and the restrictive 4-hour booking duration under uncertainty.

B. Impact of Dynamic Network Tariffs

Load Shifting: Dynamic tariffs successfully incentivized load shifting. Consumption increased during low-tariff hours (10 AM–2 PM) and decreased during high-tariff hours (5 PM–9 PM).
Diminishing Returns:
- Moderate Signals (10–60%): Effective at shifting load with minimal profit loss.
- Strong Signals (>20% increase): Induced limited additional flexibility gains but drastically reduced expected profitability (up to 65% reduction).
Conclusion: While dynamic tariffs unlock local flexibility, excessively strong signals erode the economic viability of the VPP without providing proportional operational benefits.

5. Significance and Conclusion

This paper provides a critical advancement in VPP scheduling by bridging the gap between high-level market strategies and low-level physical grid constraints.

For Market Operators: It highlights that risk-averse strategies, while less profitable on average, provide more stable and reliable dispatch, reducing system stress from imbalances.
For Policymakers: It suggests that dynamic network tariffs should be calibrated carefully. Moderate signals can effectively manage congestion, but overly aggressive pricing may destroy VPP profitability without achieving significant additional grid relief.
Future Work: The authors suggest moving beyond independent Gaussian error assumptions (using copulas for joint tail risks) and extending the framework to include intraday markets.

The study demonstrates that a risk-aware, high-fidelity approach is essential for the reliable integration of distributed resources into future electricity systems.