Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

bayesgrid: An Open-Source Python Tool for Generating Probabilistic Synthetic Transmission-Distribution Grids Using Bayesian Hierarchical Models

Henrique O. Caetano, Rahul K. Gupta, Carlos D. Maciel

Abstract

In this work, we present bayesgrid, an open-source python toolbox for generating synthetic power transmission-distribution systems for any geographical location worldwide, using the publicly available data from OpenStreetMap (OSM). The toolbox is based on Bayesian Hierarchical Models (BHM) which is trained on existing distribution network databases to develop a probabilistic model and can be applied to any geographical location worldwide, leveraging transfer learning. Thanks to the BHM, the tool is capable of generating multiple instances of the distribution system for a same region. The generated networks contain three-phase phase-consistent unbalanced networks, radial topology and information on the nodal demand distributions. The generated network also contain the critical reliability indices, specifically the interruption duration and frequency of failure for individual grid components, allowing its application in reliability-related studies. The tool is demonstrated for different case studies generating synthetic network datasets for different geographical regions around the world. The framework allows saving the generated networks into open-source platforms: PandaPower and OpenDSS. We also present an application for computation of probabilistic hosting capacity using the synthetic networks.

bayesgrid: An Open-Source Python Tool for Generating Probabilistic Synthetic Transmission-Distribution Grids Using Bayesian Hierarchical Models

Abstract

In this work, we present bayesgrid, an open-source python toolbox for generating synthetic power transmission-distribution systems for any geographical location worldwide, using the publicly available data from OpenStreetMap (OSM). The toolbox is based on Bayesian Hierarchical Models (BHM) which is trained on existing distribution network databases to develop a probabilistic model and can be applied to any geographical location worldwide, leveraging transfer learning. Thanks to the BHM, the tool is capable of generating multiple instances of the distribution system for a same region. The generated networks contain three-phase phase-consistent unbalanced networks, radial topology and information on the nodal demand distributions. The generated network also contain the critical reliability indices, specifically the interruption duration and frequency of failure for individual grid components, allowing its application in reliability-related studies. The tool is demonstrated for different case studies generating synthetic network datasets for different geographical regions around the world. The framework allows saving the generated networks into open-source platforms: PandaPower and OpenDSS. We also present an application for computation of probabilistic hosting capacity using the synthetic networks.
Paper Structure (21 sections, 13 figures, 3 tables)

This paper contains 21 sections, 13 figures, 3 tables.

Table of Contents

  1. Introduction
  2. Motivation
  3. Literature Review of Existing Tools
  4. Main Contributions
  5. Presenting bayesgrid
  6. Input Data
  7. Topology Estimation
  8. BHM Models
  9. Probabilistic Network Generation
  10. Main Modules and User Usage
  11. Module 1: Core Bayesian Models
  12. Module 2: Learning from data
  13. Module 3: Topology Generation
  14. Module 4: Transfer Learning and Synthesis
  15. Module 5: Saving into Standard Formats
  16. ...and 6 more sections

Figures (13)

  • Figure 1: Overview of the bayesgrid framework. The workflow proceeds from data acquisition to the training of Bayesian models, the generation of network topology via OpenStreetMap, and finally the synthesis of grid ensembles through transfer learning.
  • Figure 2: Overview of the main modules presented in bayesgrid tool.
  • Figure 3: Topology of the synthetic network generated from OpenStreetMap data for a 6 km radius around coordinates $(-23.649, -46.702)$. (a) The estimated grid topology, where black dots denote primary substations and distinct node colors define the service area supplied by each substation. (b) Distribution level transformers and voltage levels. (c) Transmission level integrated with high-voltage network.
  • Figure 4: Three distinct samples of phase allocation, all using the starting topology shown in Figure \ref{['fig:topology']}.
  • Figure 5: Posterior distribution of the total active power demand across all synthetic samples.
  • ...and 8 more figures