TL;DR
This paper introduces a novel stochastic programming approach to assess the annual peak-regulation potential of Virtual Power Plants (VPPs), specifically considering the uncertainties of renewable energy sources (wind and solar) and the flexible demand response from large-scale cold storage facilities. The method aims to optimize peak shaving while minimizing user electricity loss, offering a more realistic assessment than traditional deterministic models.
What Problem Does This Paper Address?
The increasing integration of distributed energy resources, particularly intermittent renewables like wind and solar, along with the growing demand from diverse load types, is creating challenges for grid stability and symmetry. Specifically, accurately assessing a Virtual Power Plant's (VPP) ability to manage peak electricity demand (peak shaving) is crucial for grid reliability. Traditional methods often rely on deterministic forecasts of renewable energy, which can be inaccurate and lead to over- or underestimation of peak-regulation capabilities. Furthermore, the demand-side response (DR) from large-scale facilities like cold storage, which have significant cooling demands and temperature sensitivities, has not been adequately integrated into VPP peak-regulation assessments, especially on an annual basis. This paper addresses the need for a more robust method that accounts for renewable energy uncertainty and the unique characteristics of flexible loads like cold storage to provide a realistic annual peak-regulation potential assessment for VPPs.
Methods
The paper proposes a stochastic programming (SP) model for VPPs. Key methodological contributions include:
- Scenario Generation for Renewable Uncertainty: The authors use a time-series probabilistic scenario set to model the uncertainties of wind power (WP) and photovoltaic (PV) generation. This involves:
- Estimating cumulative probability distributions of historical WP and PV output data using the
ksdensityfunction. - Employing the Frank copula function to model the joint distribution of wind and solar outputs, capturing their spatiotemporal correlation.
- Using K-means clustering with the elbow method to reduce the numerous generated scenarios into a representative set of typical scenarios.
- Flexible Load Modeling:
- Human Comfort Indicators (PMV): For office areas, the Predicted Mean Vote (PMV) index is used to quantify thermal comfort, influencing cooling load adjustments.
- Selling Pressure Indicator (SPI): For cold storage, a novel index is developed to evaluate economic losses and preservation degradation due to temperature deviations, considering varying sensitivities of different storage types.
- Stochastic Programming Model: An SP model is formulated with dual objectives: maximizing the stochastic expected value of the peak-regulation effect and minimizing the stochastic expectation value of user electricity loss (UELD). The UELD index is a composite measure combining human comfort and sales pressure.
- Optimization: The YALMIP toolbox in MATLAB is used to implement the multi-scenario stochastic mixed-integer linear programming (MILP) model, leveraging the Gurobi solver for efficient resolution.
Datasets
The study utilizes historical actual measurement data of wind and PV power from a power station in Guangdong, China. The data covers two periods:
- December 1, 2021, to February 29, 2021.
- June 1, 2022, to August 31, 2022. The time scale of the data is 1 hour. This data is used to generate typical scenarios for summer and winter.
Key Insights and Ideas
- Realistic Uncertainty Modeling: By using a multi-scenario approach based on copula functions and K-means clustering, the paper moves beyond simple deterministic forecasts to capture the complex spatiotemporal correlations and uncertainties of renewable energy sources, leading to a more robust assessment.
- Integrated Flexible Load Assessment: The introduction of the Selling Pressure Indicator (SPI) for cold storage, alongside the PMV index for office comfort, provides a more comprehensive way to model and quantify the impact of flexible loads on peak regulation and user losses.
- Annual Peak-Regulation Potential: The method's ability to simulate seasonal variations and then aggregate them to estimate annual potential is a significant step towards practical VPP operation planning.
- Trade-off Between Peak Shaving and User Loss: The dual-objective function highlights the inherent trade-off between achieving aggressive peak shaving and minimizing electricity losses for users, allowing for Pareto-optimal solutions.
Key Results
The case study, conducted in Guangdong, China, demonstrated the effectiveness of the proposed SP method:
- Peak Load Reduction: The optimal flexible load allocation resulted in a peak hour load reduction of 18.61% and a trough-hour load increase of 17.92% for the annual assessment.
- User Electricity Loss: The electricity loss degrees for users were found to be 0.209 in summer and 0.167 in winter, indicating minimal user impact.
- Comparison with Other Methods:
- The proposed SP method showed a more conservative and realistic peak-regulation potential compared to the "Traditional Approach" (DO), which tended to overestimate potential.
- The Robust Optimization (RO) method, while conservative, underestimated the peak-regulation potential and led to higher electricity purchases.
- The SP method effectively balanced peak-shaving effectiveness with minimal user loss, outperforming DO and RO in this regard.
Limitations and Future Work
The authors acknowledge several limitations and suggest future research directions:
- PMV Parameter Approximation: The current model uses approximate processing for multiple PMV parameters, not fully considering their time-varying characteristics or uncertainties related to extensive user behaviors.
- Subjective Weight Coefficients: The weight coefficients for objective functions are based on subjective experience, which could be refined.
- Focus on Electric Cooling: The current research focuses solely on the peak-shaving potential of electric cooling systems.
- Future Work: Future studies could explore:
- Refining the PMV modeling with more dynamic and uncertain parameters.
- Developing more objective methods for determining weight coefficients.
- Expanding the scope to include multi-energy systems and multi-energy VPPs.
Summary updated