ISO-NE Model Verification EE Team 2011 / CSE Team 10 EE: Cynthia - PowerPoint PPT Presentation

ISO-NE Model Verification EE Team 2011 / CSE Team 10 EE: Cynthia Bissereth, Jonathan Davis, Ethan McRae CSE: Andrew Collins, Eric Flower, Sam Markelon

Background Power system dynamic models to accurately represent BPS ● Current System: Automatic Power Plant Model Verification (APPMV) tool ● ○ Runs 24/7 Gets event trigger, verifies event, runs PPMV, sends out results in email ○ ○ Results are qualitative Courtesy of ISO-NE

The Problem Data not easily read ● Comparisons are only quantitative ● ○ Engineers receive only graphs Cannot do chronological comparison for a single plant ○ ○ Data is not well-organized No way to rank models ● Courtesy of ISO-NE

Requirements Augment current metrics ● Make more readable representations ○ Manage, analyze, and quantify results automatically ● Score and rank models ● Highlight bad models ○ Provide automated ranking from quantitative comparison ○ Discuss alternative scoring methods ○ Current scoring metric ■ Root mean square error (RMSE) ■ Normalized root mean square error (NRMSE) ■

Solution Approach ● ECE Team designs scoring engine ○ CSE Team develops UI, DB, and implements engine ○ Webex meetings weekly ● ○ Whole team meeting Present current work ○

Solution Approach ● ECE Team designs and ○ implements scoring engine CSE Team develops DB and ○ web GUI ECE CSE Team Team

Solution/experiments/simulations Scoring metric ● RMSE ● ● NRMSE

Scoring Algorithm Weighted average of multiple signal metrics ● Correlation coefficient ○ Similarity of phase magnitude ○ Similarity of phase angle ○ Implemented as a set of functions in MATLAB ● Proposed by Dr. Rezaei and Dr. Venkatasubramanian in Quantitative Indicators ● for Quality of Fit Assessment in Power System Model Validation Problems

Metrics Magnitude Correlation Angle

Magnitude Distance Semimetric 𝛽 : sensitivity

Magnitude Distance Semimetric

Angle Distance Semimetric 𝛾 : sensitivity

Angle Distance Semimetric

Semimetric Surface

Semimetric Tuning ● Sensitivity parameter β ○ Determines shape of semimetric ● Formula relating β to angle cutoff ● Allows for precise score

Semimetric Tuning Cutoff percentage:5% Cutoff angle: 180 deg

Score

Score

Future Work ● Further testing of MATLAB tfest() function ○ How accurate is the estimate? ○ Test with 2, 3, 4 poles on models and data ● Frequency range ○ Range for integrating ○ Paper outlines 0 Hz to 1 Hz

MATLAB Results

Root Mean Square Error

Normalized Root Mean Square Error

Preliminary Results

Constraints Unable to work in person, online/teleconference meetings only ●

Project management No budget necessary ● Everything online ○ Previous issue with file types resolved ● Updated Gantt Chart on following slides ●

Gantt Chart: Phase 1: Project Introduction

Gantt Chart: Phase 2: Project Design

Gantt Chart: Phase 3: Project Implementation

Gantt Chart: Phase 4: Project Verification

Demo Day Project needs ● Laptop/computer to show working formula ○ Go through the solutions and simulations ●