Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Reliability-Aware Control of Distributed Energy Resources using Multi-Source Data Models

Gejia Zhang, Robert Mieth

TL;DR

This work tackles the challenge of predicting distribution component failure risk as a function of weather context and operational decisions, using multi-source weather data fusion. It introduces two tree-based ensemble models, Weighted MS Decision Tree Ensemble (WMSDTE) and Two Cross-Stage MS Boosting (TCSMSB), designed to handle multicollinearity via soft site/feature weights and dimensionality reduction. The predicted failure probabilities are embedded in a sequential convex programming (SCP) framework that co-optimizes operational cost and reliability (via expected energy-not-served) for a distribution grid with PV and other DERs, including PV integration. A semi-synthetic case study on a modified IEEE 33-bus feeder with Copernicus weather data demonstrates convergence of the optimization and reliability-cost improvements under different PV scenarios, highlighting practical impact and scalability of the approach.

Abstract

Distributed energy resources offer a control-based option to improve distribution system reliability by ensuring system states that positively impact component failure rates. This option is an attractive complement to otherwise costly and lengthy physical infrastructure upgrades. However, required models that adequately map operational decisions and environmental conditions to system failure risk are lacking because of data unavailability and the fact that distribution system failures remain rare events. This paper addresses this gap and proposes a multi-source data model that consistently maps comprehensive weather and system state information to component failure rates. To manage collinearity in the available features, we propose two ensemble tree-based models that systematically identify the most influential features and reduce the dataset's dimensionality based on each feature's impact on failure rate estimates. These estimates are embedded within a sequential, non-convex optimization procedure, that dynamically updates operational control decisions. We perform a numerical experiment to demonstrate the cost and reliability benefits that can be achieved through this reliability-aware control approach and to analyze the properties of each proposed estimation model.

Reliability-Aware Control of Distributed Energy Resources using Multi-Source Data Models

TL;DR

This work tackles the challenge of predicting distribution component failure risk as a function of weather context and operational decisions, using multi-source weather data fusion. It introduces two tree-based ensemble models, Weighted MS Decision Tree Ensemble (WMSDTE) and Two Cross-Stage MS Boosting (TCSMSB), designed to handle multicollinearity via soft site/feature weights and dimensionality reduction. The predicted failure probabilities are embedded in a sequential convex programming (SCP) framework that co-optimizes operational cost and reliability (via expected energy-not-served) for a distribution grid with PV and other DERs, including PV integration. A semi-synthetic case study on a modified IEEE 33-bus feeder with Copernicus weather data demonstrates convergence of the optimization and reliability-cost improvements under different PV scenarios, highlighting practical impact and scalability of the approach.

Abstract

Distributed energy resources offer a control-based option to improve distribution system reliability by ensuring system states that positively impact component failure rates. This option is an attractive complement to otherwise costly and lengthy physical infrastructure upgrades. However, required models that adequately map operational decisions and environmental conditions to system failure risk are lacking because of data unavailability and the fact that distribution system failures remain rare events. This paper addresses this gap and proposes a multi-source data model that consistently maps comprehensive weather and system state information to component failure rates. To manage collinearity in the available features, we propose two ensemble tree-based models that systematically identify the most influential features and reduce the dataset's dimensionality based on each feature's impact on failure rate estimates. These estimates are embedded within a sequential, non-convex optimization procedure, that dynamically updates operational control decisions. We perform a numerical experiment to demonstrate the cost and reliability benefits that can be achieved through this reliability-aware control approach and to analyze the properties of each proposed estimation model.
Paper Structure (33 sections, 2 theorems, 78 equations, 6 figures, 2 tables, 4 algorithms)

This paper contains 33 sections, 2 theorems, 78 equations, 6 figures, 2 tables, 4 algorithms.

Table of Contents

  1. Introduction
  2. Tree-based Ensemble Model
  3. Weighted feature selection
  4. Weighted MS decision tree ensemble (WMSDTE)
  5. Probabilistic feature subset splitting in binary decision trees
  6. Sampling weights for measurement sites
  7. Sampling weights for feature categories
  8. Construction of WMSDTE
  9. Two cross-stage MS boosting (TCSMSB)
  10. Category- and site-wise grouping
  11. Dual-subset feature boosting
  12. Reliability-aware DER Control
  13. Reliability-aware active distribution grid model
  14. Computational approach
  15. Case Study
  16. ...and 18 more sections

Key Result

Theorem 1

biau2012analysis$\mathbf{X}_{\rm{all}}$ is uniformly distributed on $[0, 1]^{\kappa}$, $\rm{tf}^{\star}(\mathbf{x}_{\varsigma})$ is L-Lipschitz on $[0, 1]^{\varsigma}$ and, for all $\bm{x}_{\rm{all}} \in [0, 1]^{\kappa}$, the variance of $Y$ given $\mathbf{X}_{\rm{all}}$ is bounded by some positive where and Here $\mathcal{I}_{\rm{ter,n}}$ denotes the leaf nodes and $L$ is the L-Lipschitz const

Figures (6)

  • Figure 1: Illustration of WMSDT from data (left), via categories (middle), to decision trees (right). Darker shading denotes higher weight for $\mathcal{D}_i$ or $\mathcal{M}_i$.
  • Figure 2: Illustration of the sampling interval $\mathcal{SR}_i^{[k]}$.
  • Figure 3: Convergence and objective function (low PV)
  • Figure 4: Convergence and objective function (high PV)
  • Figure 5: DER active power dispatch in percent utilization of maximum power limit resulting from CM and MCRM in the "high PV" scenario. In CM, DG and DR are inactive. "BESS C" denotes battery charging; "BESS D" denotes battery discharging.
  • ...and 1 more figures

Theorems & Definitions (7)

  • Remark 1
  • Theorem 1
  • Definition 1: Uniform sampling
  • Definition 2: $\bm{\delta}$-targeted sampling
  • Proposition 1
  • Proof 1
  • Remark 2