Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Risk-Aware Planning of Power Distribution Systems Using Scalable Cloud Technologies

Shiva Poudel, Poorva Sharma, Abhineet Parchure, Daniel Olsen, Sayantan Bhowmik, Tonya Martin, Dylan Locsin, Andrew P. Reiman

TL;DR

This paper addresses uncertainty in distribution-grid planning caused by spatial-temporal patterns in DER adoption and weather. It proposes a scalable framework combining a Markov-based DER adoption model with GridLAB-D power-flow simulations and a cloud-based architecture, producing $n×m$ scenarios, each with $8,760$ hourly points. Key contributions include a modular AWS-based workflow (Lambda, Step Functions, S3) and GridAPPS-D integration enabling pre-processing, simulation, and post-processing at cloud scale. Demonstration on the IEEE 123-bus feeder shows that 15,000 power-flow simulations across $n×m$ scenarios can be completed in under an hour, revealing that planning for average conditions is insufficient and that extensive scenario analysis can expose costly worst-case risks.

Abstract

The uncertainty in distribution grid planning is driven by the unpredictable spatial and temporal patterns in adopting electric vehicles (EVs) and solar photovoltaic (PV) systems. This complexity, stemming from interactions among EVs, PV systems, customer behavior, and weather conditions, calls for a scalable framework to capture a full range of possible scenarios and analyze grid responses to factor in compound uncertainty. Although this process is challenging for many utilities today, the need to model numerous grid parameters as random variables and evaluate the impact on the system from many different perspectives will become increasingly essential to facilitate more strategic and well-informed planning investments. We present a scalable, stochastic-aware distribution system planning application that addresses these uncertainties by capturing spatial and temporal variability through a Markov model and conducting Monte Carlo simulations leveraging modular cloud-based architecture. The results demonstrate that 15,000 power flow scenarios generated from the Markov model are completed on the modified IEEE 123-bus test feeder, with each simulation representing an 8,760-hour time series run, all in under an hour. The grid impact extracted from this huge volume of simulated data provides insights into the spatial and temporal effects of adopted technology, highlighting that planning solely for average conditions is inadequate, while worst-case scenario planning may lead to prohibitive expenses.

Risk-Aware Planning of Power Distribution Systems Using Scalable Cloud Technologies

TL;DR

This paper addresses uncertainty in distribution-grid planning caused by spatial-temporal patterns in DER adoption and weather. It proposes a scalable framework combining a Markov-based DER adoption model with GridLAB-D power-flow simulations and a cloud-based architecture, producing scenarios, each with hourly points. Key contributions include a modular AWS-based workflow (Lambda, Step Functions, S3) and GridAPPS-D integration enabling pre-processing, simulation, and post-processing at cloud scale. Demonstration on the IEEE 123-bus feeder shows that 15,000 power-flow simulations across scenarios can be completed in under an hour, revealing that planning for average conditions is insufficient and that extensive scenario analysis can expose costly worst-case risks.

Abstract

The uncertainty in distribution grid planning is driven by the unpredictable spatial and temporal patterns in adopting electric vehicles (EVs) and solar photovoltaic (PV) systems. This complexity, stemming from interactions among EVs, PV systems, customer behavior, and weather conditions, calls for a scalable framework to capture a full range of possible scenarios and analyze grid responses to factor in compound uncertainty. Although this process is challenging for many utilities today, the need to model numerous grid parameters as random variables and evaluate the impact on the system from many different perspectives will become increasingly essential to facilitate more strategic and well-informed planning investments. We present a scalable, stochastic-aware distribution system planning application that addresses these uncertainties by capturing spatial and temporal variability through a Markov model and conducting Monte Carlo simulations leveraging modular cloud-based architecture. The results demonstrate that 15,000 power flow scenarios generated from the Markov model are completed on the modified IEEE 123-bus test feeder, with each simulation representing an 8,760-hour time series run, all in under an hour. The grid impact extracted from this huge volume of simulated data provides insights into the spatial and temporal effects of adopted technology, highlighting that planning solely for average conditions is inadequate, while worst-case scenario planning may lead to prohibitive expenses.

Paper Structure

This paper contains 12 sections, 2 equations, 7 figures, 1 table.

Table of Contents

  1. Introduction
  2. Methodology
  3. DER Adoption Model and Scenario Generation
  4. Grid Simulation and Impact Analysis
  5. Proposed Architectural Framework
  6. Pre-processing
  7. Simulation Execution
  8. Post-processing
  9. Demonstration
  10. Simulation Setup
  11. Results
  12. Conclusions

Figures (7)

  • Figure 1: IEEE 123-bus test feeder with adoption probability of customers. Each lumped load is disaggregated into a set of houses.
  • Figure 2: Proposed architecture showing modular components orchestrated as a workflow (left) and various distributed processing compute patterns (right).
  • Figure 3: Feeder PV/EV capacity for 500 $\times$ 30-year scenarios.
  • Figure 4: Performance comparison with increasing sophistication in simulation.
  • Figure 5: Impact visualization in service transformers due to potential EV/PV growth in 30 years.
  • ...and 2 more figures