Abstract:This paper studies some compression methods to accelerate the scenario-based chance-constrained security-constrained economic dispatch (SCED) problem. In particular, we show that by exclusively employing the vertices after convex hull compression, an equivalent solution can be obtained compared to utilizing the entire scenario set. For other compression methods that might relax the original solution, such as box compression, this paper presents the compression risk validation scheme to assess the risk arising from the sample space. By quantifying the risk associated with compression, decision-makers are empowered to select either solution risk or compression risk as the risk metric, depending on the complexity of specific problems. Numerical examples based on the 118-bus system and synthetic Texas grids compare these two risk metrics. The results also demonstrate the efficiency of compression methods in both problem formulation and solving processes.
What problem does this paper attempt to address?
The problem that this paper attempts to solve is how to accelerate the solution process of scenario - based stochastic - constrained security - constrained economic dispatch (SCED) problems in large - scale power systems through scenario compression methods. Specifically, several compression methods are studied to reduce the number of scenarios that need to be processed, thereby improving computational efficiency while maintaining the effectiveness of the solution or evaluating the risks introduced by compression.
### Background of the Paper and Problem Definition
With the increase in renewable energy, the security - constrained economic dispatch of power systems faces unprecedented challenges. Traditional deterministic SCED methods are difficult to deal with these uncertainties, so multiple uncertainty optimization methods have been proposed, such as stochastic optimization (SO), robust optimization (RO), and chance - constrained optimization (CCO). CCO combines the advantages of SO and RO, ensuring the feasibility of the solution at a specific probability, allowing decision - makers to balance risks and costs according to their preferences.
### Specific Expression of the Problem
The paper focuses on how to accelerate the SCED problem in CCO through scenario compression methods. Specifically:
1. **Convex Hull Compression**: By using only the vertices of the convex hull, the same solution as using the entire set of scenarios can be obtained. This takes advantage of the characteristics of the vertices after convex hull compression and reduces the number of scenarios that need to be processed.
2. **Box Compression**: By compressing the scenarios into a hyper - rectangular space, the number of vertices that need to be processed is further reduced. Although this method may relax the original solution, the resulting risks can be evaluated through a compression - risk - verification scheme.
3. **Compression - Risk - Verification**: A new compression - risk - verification scheme is proposed, which directly quantifies the compression risk based on the complexity of the sample space, rather than just based on the complexity of the solution. This allows decision - makers to select appropriate risk - measurement criteria according to the complexity of specific problems.
### Mathematical Models and Methods
#### 1. Chance - Constrained SCED Problem
The conventional SCED problem is reformulated in the form of chance constraints, mainly considering wind - power - generation prediction errors as system uncertainties. The mathematical model is as follows:
\[
\begin{aligned}
& \min_{g, \eta} c(g) \\
& \text{s.t.} \quad 1^\top g = 1^\top d - 1^\top \hat{w} \\
& \quad g \preceq g \preceq \bar{g} \\
& \quad f(\hat{w}, \tilde{w}) = Hg \left( g - 1^\top \tilde{w} \eta \right) - H_d d + H_w (\hat{w} + \tilde{w}) \\
& \quad P_{\tilde{w}} \left( f \preceq f(\hat{w}, \tilde{w}) \preceq \bar{f}, \quad g \preceq g - 1^\top \tilde{w} \eta \preceq \bar{g}, \quad R_d \preceq -1^\top \tilde{w} \eta \preceq R_u \right) \geq 1 - \epsilon \\
& \quad 1^\top \eta = 1
\end{aligned}
\]
where \( g \) is the generator output level, \( \eta \) is the affine policy of automatic generation control, \( \hat{w} \) is the deterministic wind - power prediction value, \( \tilde{w} \) is the uncertain prediction error, and \( \epsilon \) is the risk level.
#### 2. Scenario Method
Through the scenario method, the chance - constraint inequality is converted into a scenario - based inequality, as follows:
\[
\begin{aligned}
& \min_{g, \eta} c(g) \\
& \text{s.t.} \quad f \preceq f(\hat{w}, \tilde{w}_i) \preceq \bar{f} \\
& \quad g \preceq g - 1^