Great Courses:
- Technological Forecasting too Decision-Making - 22165 - EMGT 6910 - H80
- EMGT6965 Energy Analytics
Dissertation
https://repository.lib.ncsu.edu/bitstream/handle/1840.16/6457/etd.pdf?sequence=2&isAllowed=y
Npower Forecast
https://www.npowerjobs.com/Resources/npm12932MM18978ForecastingChallengeEbooklet08-16.pdf
Modeling too Forecasting Electricity Loads too Prices: Influenza A virus subtype H5N1 Statistical Approach
http://onlinelibrary.wiley.com/book/10.1002/9781118673362http://www.ioz.pwr.wroc.pl/pracownicy/weron/MFE.htm
source: https://branchesofmechanicalengineering.blogspot.com//search?q=leaderboard-for-gefcom2017-qualifying-match
Monday, March 6, 2017
Leaderboard for GEFCom2017 Qualifying Match!!!
[Update 5/18/2017]: ISO NE simply released the Apr charge information ii days ago. Jingrui too I induce got updated the leaderboard for the qualifying match. Please banking company check the rankings too allow us know yesteryear 5/26/2017 if at that spot is whatsoever issue.
The vi rounds of GEFCom2017 qualifying gibe simply ended lastly week. I'm for certain that the contestants are anxiously waiting for the leaderboard. Here is a brief report. I'll update this post equally ISO New England releases its recent charge data.
Stay tuned :)
The vi rounds of GEFCom2017 qualifying gibe simply ended lastly week. I'm for certain that the contestants are anxiously waiting for the leaderboard. Here is a brief report. I'll update this post equally ISO New England releases its recent charge data.
Out of 177 registered teams, 73 induce got submitted entries to the defined track, too 26 to the opened upward track. After vi rounds, 53 teams completed the defined rails alongside at to the lowest degree iv submissions, patch xx completed the opened upward track.
The due appointment of study too code is on March 10th, 2017. Please shipping them to hong.bigdeal@gmail.com. Follow the same protocol equally the forecast submissions. Please follow THIS GUIDE to laid upward the report.
Jingrui Xie created ii benchmarks:
- Vanilla Benchmark, which has been used to calculate the scores of the teams inward each round. See Q7 of THIS FAQ for to a greater extent than information.
- Rain Benchmark, which volition live used to pick out the teams beingness advanced to the lastly match.
(As an organizer of GEFCom2017, Jingrui Xie is non eligible for the prize.)
The spreadsheet alongside detailed scores tin live accessed HERE. The higher the score is, the higher the rank is.
Stay tuned :)
Source: https://branchesofmechanicalengineering.blogspot.com//search?q=leaderboard-for-gefcom2017-qualifying-match
Documentation inward Load Forecasting: iv Reasons too 8 Elements
Monday, Nov 24, 2014
Documentation inward Load Forecasting: iv Reasons too 8 Elements
In charge forecasting, particularly long term charge forecasting, documentation is belike the most of import task. The ultimate exam of documentation character is whether the forecasting organisation has been described inward particular therefore that other people alongside relevant educational activity background too sense tin reproduce the forecasts.
While forecasting is similar an adventure, exploring an unknown trail, documentation is similar walking the same trail i time again too i time again to tape what happened inward detail. Documentation oftentimes requires meaning amount of efforts, sometimes to a greater extent than than forecasting itself.
Rather than having the forecaster explicate the forecast inward somebody every time, documentation helps simplify the communication.
2. Further improvement
As the forecasting procedure gets to a greater extent than too to a greater extent than complicated, it's impossible for a forecaster to memorize everything s/he has done inward the past. Documentation serves equally a reminder when the forecaster is trying to cook improvement based on a formally developed model.
3. Business continuity
If the forecaster left the job, someone else alongside similar background tin convey over the work.
4. Auditing too defense
Industry rule requires utilities to defend their forecasts equally component of the charge per unit of measurement case. The defence is mainly based on documentation.
Responsibilities of everybody getting involved inward the entire charge forecasting process, such equally the ones who provide the raw data, draw the information from database, railroad train the models, convert the forecasts to the presentable format, review the forecast, suggest revision too approve the forecast.
2. Data
All the information sources involved inward the forecasting process, such equally charge data, atmospheric condition data, economic scheme data, demographic data, outage data, etc. If the information needs modification, all major versions of the information should live stored alongside notes indicating why too how the information is beingness revised.
3. Forecasting methodology
Methodologies for atmospheric condition station selection, outlier detection too information cleansing, model selection, hierarchical forecasting, forecast combination, etc.
4. Models too forecasts
The techniques (such equally multiple linear regression, autoregression integrated moving average, etc.) beingness used, the resulting models too forecasts. If the models are beingness updated or revised inward the forecasting process, all major revisions too versions of the models should live documented.
5. Error analysis
How accurate the forecasts are, including comprehensive measures on key conclusion periods, such equally MAPE (Mean Absolute Percentage Error) of monthly peak demand, RMSE (Root Mean Square Error) yesteryear hr of a day, etc.
6. Judgmental changes
If the forecasts are beingness revised based on personal sense and/or trouble organisation sense, details should live provided inward documentation, such equally who made the revision, how the forecasts are beingness changed, why too what the novel forecasts human face like.
7. Strength too limitation
Pros too cons of the model(s), forecast(s) too the forecasting process. For instance, is the model doing good during regular days but bear poorly during holidays?
8. Directions for futurity improvement
How the forecasting procedure tin live improved inward the adjacent forecasting cycle.
The Global Energy Forecasting Competition 2014 is going on correct now, alongside less than 3 weeks towards the end. The rattling lastly project later on the marathon-like 15-week contender is documentation. The contestants are required to submit their code too a study illustrating how they did inward the competition. Although nosotros are non providing whatsoever formal template for the report, this spider web log post volition server equally a full general guideline for the GEFCom2014 contestants inward all tracks. In the lastly report, nosotros would human face at to the lowest degree a summary of your scores too rankings (based on Inside Leaderboard), methodologies you've been using, a log of the 12 evaluation weeks showing evolution of your model(s) too key references that spark your bright ideas.
While forecasting is similar an adventure, exploring an unknown trail, documentation is similar walking the same trail i time again too i time again to tape what happened inward detail. Documentation oftentimes requires meaning amount of efforts, sometimes to a greater extent than than forecasting itself.
Why documentation?1. Communication
Rather than having the forecaster explicate the forecast inward somebody every time, documentation helps simplify the communication.
2. Further improvement
As the forecasting procedure gets to a greater extent than too to a greater extent than complicated, it's impossible for a forecaster to memorize everything s/he has done inward the past. Documentation serves equally a reminder when the forecaster is trying to cook improvement based on a formally developed model.
3. Business continuity
If the forecaster left the job, someone else alongside similar background tin convey over the work.
4. Auditing too defense
Industry rule requires utilities to defend their forecasts equally component of the charge per unit of measurement case. The defence is mainly based on documentation.
What to document?1. Who
Responsibilities of everybody getting involved inward the entire charge forecasting process, such equally the ones who provide the raw data, draw the information from database, railroad train the models, convert the forecasts to the presentable format, review the forecast, suggest revision too approve the forecast.
2. Data
All the information sources involved inward the forecasting process, such equally charge data, atmospheric condition data, economic scheme data, demographic data, outage data, etc. If the information needs modification, all major versions of the information should live stored alongside notes indicating why too how the information is beingness revised.
3. Forecasting methodology
Methodologies for atmospheric condition station selection, outlier detection too information cleansing, model selection, hierarchical forecasting, forecast combination, etc.
4. Models too forecasts
The techniques (such equally multiple linear regression, autoregression integrated moving average, etc.) beingness used, the resulting models too forecasts. If the models are beingness updated or revised inward the forecasting process, all major revisions too versions of the models should live documented.
5. Error analysis
How accurate the forecasts are, including comprehensive measures on key conclusion periods, such equally MAPE (Mean Absolute Percentage Error) of monthly peak demand, RMSE (Root Mean Square Error) yesteryear hr of a day, etc.
6. Judgmental changes
If the forecasts are beingness revised based on personal sense and/or trouble organisation sense, details should live provided inward documentation, such equally who made the revision, how the forecasts are beingness changed, why too what the novel forecasts human face like.
7. Strength too limitation
Pros too cons of the model(s), forecast(s) too the forecasting process. For instance, is the model doing good during regular days but bear poorly during holidays?
8. Directions for futurity improvement
How the forecasting procedure tin live improved inward the adjacent forecasting cycle.
The Global Energy Forecasting Competition 2014 is going on correct now, alongside less than 3 weeks towards the end. The rattling lastly project later on the marathon-like 15-week contender is documentation. The contestants are required to submit their code too a study illustrating how they did inward the competition. Although nosotros are non providing whatsoever formal template for the report, this spider web log post volition server equally a full general guideline for the GEFCom2014 contestants inward all tracks. In the lastly report, nosotros would human face at to the lowest degree a summary of your scores too rankings (based on Inside Leaderboard), methodologies you've been using, a log of the 12 evaluation weeks showing evolution of your model(s) too key references that spark your bright ideas.
Source: https://branchesofmechanicalengineering.blogspot.com//search?q=leaderboard-for-gefcom2017-qualifying-match
Saturday, July 25, 2015
Combining Load Forecasts from Independent Experts: Experience at NPower Forecasting Challenge 2015
Forecast combination is regarded equally i of the best practices of forecasting. I mean value it is a straightforward too practical approach to improving existing forecasts. This newspaper describes the method my students took inward the NPower Forecasting Challenge 2015. We volition introduce the newspaper at the 47th North America Power Symposium.
Citation
Jingrui Xie, Bidong Liu, Xiaoqian Lyu, Tao Hong, too David Basterfield, "Combining charge forecasts from independent experts: sense at NPower forecasting challenge 2015", the 47th North American Power Symposium (NAPS2015), Oct iv - 6, 2015
Abstract
The NPower Forecasting Challenge 2015 invited students too professionals worldwide to predict daily release energy usage of a grouping of customers. The BigDEAL squad from the Big Data Energy Analytics Laboratory landed a transcend 3 house inward the lastly leaderboard. This newspaper presents a refined methodology based on the implementation during the competition. We start build the private forecasts using several forecast techniques, such equally Multiple Linear Regression (MLR), Autoregressive Integrated Moving Average (ARIMA), Artificial Neural Network (ANN) too Random Forests (RF). We too then pick out a subset of the private forecasts based on their performance on a validation period, a.k.a. post-sample. Finally nosotros obtain the lastly forecast yesteryear averaging the selected private forecasts. The forecast combination on average yields a meliorate final result than the forecast from a unmarried technique.
Citation
Jingrui Xie, Bidong Liu, Xiaoqian Lyu, Tao Hong, too David Basterfield, "Combining charge forecasts from independent experts: sense at NPower forecasting challenge 2015", the 47th North American Power Symposium (NAPS2015), Oct iv - 6, 2015
Combining Load Forecasts from Independent Experts
Experience at NPower Forecasting Challenge 2015
Jingrui Xie, Bidong Liu, Xiaoqian Lyu, Tao Hong, too David Basterfield
Abstract
The NPower Forecasting Challenge 2015 invited students too professionals worldwide to predict daily release energy usage of a grouping of customers. The BigDEAL squad from the Big Data Energy Analytics Laboratory landed a transcend 3 house inward the lastly leaderboard. This newspaper presents a refined methodology based on the implementation during the competition. We start build the private forecasts using several forecast techniques, such equally Multiple Linear Regression (MLR), Autoregressive Integrated Moving Average (ARIMA), Artificial Neural Network (ANN) too Random Forests (RF). We too then pick out a subset of the private forecasts based on their performance on a validation period, a.k.a. post-sample. Finally nosotros obtain the lastly forecast yesteryear averaging the selected private forecasts. The forecast combination on average yields a meliorate final result than the forecast from a unmarried technique.
Improving Gas Load Forecasts With Big Data
http://onlinelibrary.wiley.com/doi/10.1002/gas.21905/abstract
Probabilistic charge forecasting via Quantile Regression Averaging on sis forecasts
Source: https://branchesofmechanicalengineering.blogspot.com//search?q=leaderboard-for-gefcom2017-qualifying-match
Thursday, May 14, 2015
Probabilistic Load Forecasting via Quantile Regression Averaging on Sister Forecasts
Although the probabilistic charge forecasting literature tin live traced dorsum to 1970s, the importance of the discipline was non good recognized until recent years. There are several approaches to producing probabilistic charge forecasts, such equally generating atmospheric condition scenarios to feed to betoken forecasting models, applying probabilistic modeling too forecasting techniques, too identifying the density business office of residuals. This newspaper starts a whole novel category for probabilistic charge forecasting methods - combining betoken charge forecasts.
The proposed methodology includes ii parts:
The proposed methodology includes ii parts:
- Generating sis betoken forecasts. The concept of sis models/forecasts is a cook novel inward this field. The models were start developed from my large information paper. This newspaper farther materializes the concept too applies it to probabilistic charge forecasting.
- Combining forecasts alongside quantile regression. While quantile regression is non a rare technique, at that spot has non been many studies that apply quantile regression. In this paper, nosotros role quantile regression to combining the betoken forecasts. The combination is dominantly meliorate than the probabilistic charge forecasts from private models.
The newspaper was accepted yesteryear IEEE Transactions on Smart Grid this week. The working newspaper is available HERE. I volition update the citation i time the newspaper is on IEEE Xplore.
Citation
Bidong Liu, Jakub Nowotarski, Tao Hong too Rafal Weron, "Probabilistic Load Forecasting via Quantile Regression Averaging on Sister Forecasts", IEEE Transactions on Smart Grid, accepted, working newspaper available from http://www.drhongtao.com/articles.
Probabilistic Load Forecasting via Quantile Regression Averaging on Sister Forecasts
Bidong Liu, Jakub Nowotarski, Tao Hong too Rafał Weron
Abstract
Majority of the charge forecasting literature has been on betoken forecasting, which provides the expected value for each footstep throughout the forecast horizon. In the smart grid era, the electricity demand is to a greater extent than active too less predictable than e'er before. As a result, probabilistic charge forecasting, which provides additional information on the variability too incertitude of futurity charge values, is becoming of groovy importance to powerfulness systems planning too operations. This newspaper proposes a practical methodology to generate probabilistic charge forecasts yesteryear performing Quantile Regression Averaging (QRA) on a laid of sis betoken forecasts. There are ii major benefits of the proposed approach: 1) it tin leverage the evolution inward the betoken charge forecasting literature over the yesteryear several decades; too 2) it does non rely therefore much on high character skilful forecasts, which are rarely achievable inward charge forecasting practice. To demonstrate the effectiveness of the proposed approach too cook the results reproducible to the charge forecasting community, nosotros build a illustration study using the publicly available information from the Global Energy Forecasting Competition 2014. Comparing alongside several benchmark methods, the proposed approach leads to dominantly meliorate performance equally measured yesteryear the pinball loss business office too the Winkler score.
You powerfulness also like:
- Improving Short Term Load Forecast Accuracy via Combining Sister Forecasts
- Probabilistic Electric Load Forecasting: Influenza A virus subtype H5N1 Tutorial Review
- Combining Load Forecasts from Independent Experts: Experience at NPower Forecasting Challenge 2015
- On Normality Assumption inward Residual Simulation for Probabilistic Load Forecasting
Tuesday, Dec 20, 2016
Winning Methods from npower Forecasting Challenge 2016
RWE npower released the lastly leaderboard for its forecasting challenge 2016. I took a covert shot of the transcend teams. Interestingly, the international teams (colored inward red) took over all of the transcend 6 places. Unfortunately, to a greater extent than or less of those top-notch Great Britain charge forecasters did non bring together the competition. I'm hoping that they tin present upward at the game to defend the country's legacy:)
In each of the previous ii npower competitions, I asked my BigDEAL students to bring together the contender equally a team. In both competitions, they were ranked transcend too beating all Great Britain teams (see the spider web log posts HERE and HERE). We also published our winning methods for electricity demand forecasting and gas demand forecasting.
This year, instead of forming a BigDEAL team, I sent the students inward my HERE.)
OK, plenty bragging...
I asked the transcend teams portion their methodologies alongside the audience of my spider web log equally what nosotros did in BFCom2016s. Here they are:
======================
Source: https://branchesofmechanicalengineering.blogspot.com//search?q=leaderboard-for-gefcom2017-qualifying-match
 |
| RWE npower Forecasting Challenge 2016 Final Leaderboard (top 12 places) |
In each of the previous ii npower competitions, I asked my BigDEAL students to bring together the contender equally a team. In both competitions, they were ranked transcend too beating all Great Britain teams (see the spider web log posts HERE and HERE). We also published our winning methods for electricity demand forecasting and gas demand forecasting.
This year, instead of forming a BigDEAL team, I sent the students inward my HERE.)
OK, plenty bragging...
I asked the transcend teams portion their methodologies alongside the audience of my spider web log equally what nosotros did in BFCom2016s. Here they are:
======================
Source: https://branchesofmechanicalengineering.blogspot.com//search?q=leaderboard-for-gefcom2017-qualifying-match
Friday, Apr 22, 2016
BigDEAL Students Receiving Promotions
As a professor, I let out nix meliorate than hearing the success stories of my students. Currently I induce got ii PhD students, Jingrui (Rain) Xie too Jon Black. Both of them are also working total fourth dimension inward the industry. This is the flavor of advertisement announcements inward many companies. Rain was promoted from Sr. Associate Research Statistician Developer to Research Statistician Developer, patch Jon was promoted from Lead Engineer to Manager. Here I'm rattling pleased to characteristic their curt biographies alongside the novel trouble organisation titles. For to a greater extent than details almost their profiles, delight banking company check out the BigDEAL electrical flow students page.
Congratulations, Rain too Jon, for the well-deserved promotions!
 |
| Jingrui Xie |
Jingrui (Rain) Xie, Research Statistician Developer, Forecasting R&D, SAS Institute Inc.
Jingrui (Rain) is pursuing her Ph.D. grade at UNC Charlotte where her inquiry focuses on probabilistic charge forecasting. Meanwhile, she also industrial plant full-time equally a Research Statistician Developer at SAS Forecasting R&D. At SAS, she industrial plant on the evolution of SAS forecasting components too solutions, too leads the release energy forecasting research. Prior to joining SAS Forecasting R&D, Rain was an analytical consultant at SAS alongside expertise inward statistical analysis too forecasting particularly on release energy forecasting. She was the Pb statistician developer for SAS Energy Forecasting solution too delivered consulting services to several utilities on charge forecasting for their organisation operations, planning too release energy trading.
Rain has extensive sense inward release energy forecasting including exploratory information analysis, pick of atmospheric condition stations, outlier detection too information cleansing, hierarchical charge forecasting, model evaluation too selection, forecast combination, atmospheric condition normalization too probabilistic charge forecasting. She also has extensive cognition too working sense alongside a broad laid of SAS products.
 |
| Jonathan D. Black |
Jonathan D. Black, Manager of Load Forecasting, System Planning, ISO New England Inc.
Jon is currently Manager of Load Forecasting at ISO New England, where he provides technical direction for release energy analytics too both short-term too long-term forecasting of load, distributed photovoltaic (PV) resources, too release energy efficiency. For the yesteryear 3 years he has led ISO-NE’s long-term PV forecasting for the vi New England states based on a diversity of province policy back upward mechanisms, too provided technical guidance for the modeling of PV inward organisation planning studies. Jon is directing ISO-NE’s efforts to railroad train enhanced short-term charge forecast tools that comprise the effects of behind-the-meter distributed PV, too has developed methods of estimating distributed PV fleet production profiles using express historical data, equally good equally simulating high penetration PV scenarios to seat futurity internet charge characteristics. Jon participates inward industry-leading inquiry on forecasting too integrating large-scale renewable release energy resources, too has served equally a Technical Review Committee fellow member on several multi-year Department of Energy studies. Upon joining ISO-NE inward 2010, Jon assisted alongside the New England Wind Integration Study too the blueprint of air current found information requirements for centralized air current powerfulness forecasting.
Mr. Black is currently a PhD educatee researching advanced forecasting techniques inside the Infrastructure too Environmental Systems programme at the University of North Carolina at Charlotte. He received his MS grade inward Mechanical Engineering from the University of Massachusetts at Amherst, where his inquiry at the UMass Wind Energy Center explored the effects of varying atmospheric condition on regional electricity demand too renewable resources availability. He is an active fellow member of both the Institute of Electrical too Electronics Engineers (IEEE) too the Utility Variable Generation Integration Group (UVIG).