Constant Jacobian Matrix-Based Stochastic Galerkin Method for Probabilistic Load Flow

Yingyun Sun, Rui Mao, Zuyi Li, Wei TianEnergies 2016 · DOI

TL;DR

This paper introduces a novel, efficient method for Probabilistic Load Flow (PLF) analysis that accurately accounts for uncertainties from renewable energy sources. By leveraging the P-Q decoupling property of power systems and pre-calculating constant Jacobian matrices, their modified Stochastic Galerkin (SG) method significantly speeds up computations while maintaining high accuracy, making it more practical for modern power grids.

What Problem Does This Paper Address?

Modern power grids are increasingly incorporating renewable energy sources like wind and solar, which are inherently volatile and intermittent. This variability introduces significant uncertainty into power system operations. Traditional deterministic load flow (DLF) analysis, which assumes fixed values for all parameters, can no longer accurately predict system behavior. Probabilistic Load Flow (PLF) aims to quantify these uncertainties by providing statistical information about system outputs (like voltage and power flow). However, existing PLF methods, such as Monte Carlo simulations, can be computationally expensive, while other analytical or approximate methods may sacrifice accuracy. The challenge is to develop a PLF method that is both accurate in handling complex uncertainties and computationally efficient enough for real-time applications.

Methods

The paper proposes a modified Stochastic Galerkin (SG) method for PLF. Here's a breakdown of their approach:

  • Generalized Polynomial Chaos (gPC) Expansion: They represent uncertain variables (like wind power and load) using gPC expansions. This allows for a spectral stochastic representation of the power system model.
  • Stochastic Galerkin (SG) Method: This is an intrusive spectral method that forms coupled deterministic equations by projecting the system's residual equations onto the gPC basis functions. While known for its accuracy, traditional SG can be computationally cumbersome due to high-dimensional, coupled equations.
  • P-Q Decoupling Modification: The core innovation is adapting the SG method by exploiting the P-Q decoupling property of load flow. This allows them to separate the voltage angle (θ) and voltage magnitude (V) calculations.
  • Constant Jacobian Matrices: Instead of recomputing Jacobian matrices in each iteration of the Newton-Raphson solver (which is typical in SG), they pre-calculate two constant Jacobian matrices (J' and J"). This is a significant computational saving.
  • Sparse Grid Technique: For numerical integration required in the SG method, they employ a sparse grid technique, which is more efficient than traditional tensor product rules, especially for higher-dimensional problems.

Datasets

The proposed method was evaluated on two standard power system test cases:

  • IEEE 14-bus system: Modified to include wind generation.
  • IEEE 118-bus system: A larger system, also modified to include wind generation and increased uncertainty levels.

These are standard benchmarks in power systems research, and the paper uses them to demonstrate the performance of their method against established techniques.

Key Insights and Ideas

  • P-Q Decoupling is Key to SG Efficiency: The paper demonstrates that by exploiting the P-Q decoupling inherent in power flow, the computational burden of the SG method can be drastically reduced. This is a clever application of a well-known power system simplification to a more advanced UQ technique.
  • Constant Jacobians are a Game-Changer: The most impactful insight is the realization that constant Jacobian matrices can be formed for the decoupled system. This eliminates the need for repeated, expensive Jacobian computations within the iterative solver, leading to substantial speedups.
  • Accuracy vs. Computational Cost Trade-off: The study highlights that while higher-order gPC expansions improve accuracy, they also increase computational cost due to a larger number of quadrature nodes. The paper suggests choosing the order based on numerical experiments, implying an optimal balance exists.
  • Intrusive SG Offers Superior Accuracy: Compared to non-intrusive methods like Stochastic Collocation (SCM), the intrusive nature of SG, combined with the proposed modifications, yields better accuracy without prohibitive computational overhead.

Key Results

  • Significant Speedup: The proposed method achieved considerable computational speed improvements. For instance, on the IEEE 118-bus system, their method took only 3.306 seconds, compared to 16.436 seconds for Monte Carlo simulation.
  • High Accuracy Maintained: Despite the speed improvements, the method maintained high accuracy, comparable to Monte Carlo simulations, as evidenced by low relative errors (RE) and average root mean square (ARMS) errors. For example, the order-3 SG-PLF method showed very low mean and maximum RE values for voltage and angle.
  • Scalability: The method demonstrated good performance on both the 14-bus and 118-bus systems, indicating its scalability to larger and more complex power grids.
  • Robustness to Uncertainty Levels: The study showed that the variance of power injection had a minor effect on accuracy, and wind speed correlation had a minor effect on accuracy, suggesting the method is robust to varying levels of uncertainty.

Limitations and Future Work

  • Determining Optimal gPC Order: The paper notes that choosing the optimal gPC expansion order (P) is currently done through numerical experiments. Determining this order analytically or through more systematic methods is suggested as future work.
  • Further UQ Research: The authors believe their method encourages further UQ research in power systems, particularly for handling more complex and sophisticated uncertainties with reduced computational effort. This implies that while their method is advanced, there's still room for exploring even more complex scenarios.

Summary updated

Abstract

An intrusive spectral method of probabilistic load flow (PLF) is proposed in the paper, which can handle the uncertainties arising from renewable energy integration. Generalized polynomial chaos (gPC) expansions of dependent random variables are utilized to build a spectral stochastic representation of PLF model. Instead of solving the coupled PLF model with a traditional, cumbersome method, a modified stochastic Galerkin (SG) method is proposed based on the P-Q decoupling properties of load flow in power system. By introducing two pre-calculated constant sparse Jacobian matrices, the computational burden of the SG method is significantly reduced. Two cases, IEEE 14-bus and IEEE 118-bus systems, are used to verify the computation speed and efficiency of the proposed method.
BibTeX
@inproceedings{sun2016constant,
  author = {Yingyun Sun and Rui Mao and Zuyi Li and Wei Tian},
  title = {{Constant Jacobian Matrix-Based Stochastic Galerkin Method for Probabilistic Load Flow}},
  booktitle = {Energies},
  year = {2016},
  doi = {10.3390/en9030153},
}
Figure