Abstract
The increasing uncertainty level caused by growing renewable energy sources (RES) and aging transmission networks poses a great challenge in the assessment of total transfer capability (TTC) and available transfer capability (ATC). This paper proposes a novel data-driven sparse polynomial chaos expansion (DDSPCE) method for estimating the probabilistic characteristics of probabilistic TTC (PTTC). The proposed method requires no pre-assumed probabilistic distributions of random inputs, exploiting data sets directly. A sparse scheme is integrated to improve computational efficiency. Numerical studies on the modified IEEE 118-bus system demonstrate that DDSPCE achieves accurate estimation with high efficiency and highlights the importance of incorporating discrete random inputs.
Introduction
In modern power systems, Total Transfer Capability (TTC) represents the maximum power that can be reliably transferred over a transmission network without violating thermal, voltage, or stability limits. As grids transition toward intermittent renewable energy sources like wind and solar, TTC is no longer a static value but a stochastic variable.
Traditional deterministic assessments often rely on fixed safety margins, which can lead to overly conservative operations or unforeseen risks. Probabilistic TTC (PTTC) methods offer a more nuanced view by treating capacity as a distribution rather than a single point. However, these methods often struggle with non-standard data distributions and the high dimensionality of large power grids.
Methodology: The DDSPCE Framework
The authors introduce the Data-Driven Sparse Polynomial Chaos Expansion (DDSPCE) method to overcome these hurdles. The framework evolves standard PCE in three key ways:
- Data-Driven Basis: Instead of using pre-defined polynomials (e.g., Hermite for Gaussian data), the method employs a moment-based approach. It calculates statistical moments directly from raw data to construct an orthonormal polynomial basis, allowing the model to "learn" the specific shape of uncertainty without fitting a predefined distribution.
- Sparsity via Least Angle Regression (LAR): To manage high-dimensional systems, the LAR algorithm identifies and selects only the most influential polynomial terms. This creates a "sparse" surrogate model that requires significantly fewer system simulations to achieve high accuracy.
- Hybrid Modeling: The framework explicitly incorporates discrete random variables (modeled via Bernoulli distributions) to account for sudden branch outages, integrating them into the same expansion used for continuous weather data.
Key Results and Discussion
The DDSPCE method was validated using a modified IEEE 118-bus system featuring 111 continuous variables (wind, solar, loads) and 4 discrete line outages.
Accuracy and Efficiency
The results demonstrate that DDSPCE provides an exceptionally close match to the "gold standard" Monte Carlo Simulation (MCS). As shown in the probability and CDF plots, the DDSPCE curves almost perfectly overlap with MCS results, indicating high fidelity in capturing the system's probabilistic behavior.
In terms of speed, the proposed method was approximately 18 times faster than MCS. For the 118-bus system, DDSPCE achieved accurate results in roughly 9,370 seconds, compared to over 173,484 seconds for MCS.
The Impact of Discrete Events
A critical finding is that ignoring discrete events like line outages leads to a significant overestimation of transfer capability. Incorporating these events reduced the mean TTC by up to 9.14% and increased variance by over 25%. This underscores the necessity of including discrete uncertainties to set safe Transmission Reliability Margins (TRM).
Conclusion
The DDSPCE method offers a robust, efficient, and data-driven approach to PTTC assessment. By avoiding distribution assumptions and utilizing sparse modeling, it provides grid operators with a high-fidelity "digital twin" for real-time capacity estimation. Future work will explore further dimension reduction for even larger systems and global sensitivity analysis to pinpoint the primary drivers of grid uncertainty.
Summary updated