Entry Period

December 20, 2018 - February 08, 2019

Posted by

Tiny_logo

3428

Viewers

Published date

December 20, 2018

PV in HOKKAIDO

About this project

■Updates
2019.02.05
・"FAQ" has been added.
2019.02.18
・"FAQ" has been updated.

Uploads_2fc5ad25cf813a1f530321ecb6b8e5601f_2fmain-01

Uploads_2fc5ad25cf813a1f530321ecb6b8e5601f_2f2018_title

[Purpose of the Contest]
We, electric power suppliers, are now endeavoring to expand the introduction of renewable energy in the form of electric power of secured quality.
The representative renewable energy, solar power generation, is a fluctuating-type power source that depends on weather conditions. For this reason, the precise prediction of power output of solar power plants is critical to the effective use of solar power generation as a power supply source.
We have developed and introduced unique methods to predict power output of solar power plants based on weather data, and utilized them in our actual works. However, the prediction accuracy can be degraded by local weather changes, recent climatic changes, or impacts due to snow/accumulated snow on solar panels during the winter season. Further efforts are required to improve the prediction accuracy under these conditions.

Tokyo Electric Power Company Holdings (TEPCO) and Hokkaido Electric Power Company (HEPCO) have therefore decided to jointly host a "PV in HOKKAIDO - Contest Seeks Best Models for Predicting Power Output of Solar Power Plants in Hokkaido, Japan". The aim of the contest is to discover innovative methods and new approaches to solar power generation prediction by inviting contest Participants to submit technologies to predict power output of solar power plants in the Hokkaido area.

Contest Participants will be asked to predict the sum of power output from a set of specified solar power plants connected to power systems in the Hokkaido area. We, the Contest Hosts, will determine the winners by comprehensively considering the proposed prediction models, their practicality and expandability, and the accuracy of the predicted results.

We intend to apply the most excellent of the proposed methods from the contest in our actual operations with the aim of expanding the introduction of solar power generation and improving the stability of power supply by predicting power output of solar power plants more accurately.

Uploads_2f24fcac5df73c39fa9257858a43473772_2f2018_r_summary

●Topic
Prediction of the total power output from specified solar power plants in the Hokkaido area.

●Available data (Participants can freely select the data)
・Historical data set for training (provided in advance by the Contest Hosts)
(1) Actual power output data from the past
(2) Rating output, locations, azimuth angle, and inclination angle
(3) Independently measured amount of global solar radiation and the measurement locations

・Other external data (Participants obtain the data on their own)

●Items to be submitted
(All Participants)
(1) Predicted power output of solar power plants
(2) Explanatory report describing the method (summary)

(Participants Selected for the Final Review)
(1) Explanatory report describing the method (details)
(2) Presentation video

●Selection review / Evaluation
(First Selection)
- Prediction accuracy (quantitative evaluation)
- Explanatory report describing the method (qualitative evaluation)
(Final selection)
- Contents of the First Selection
- Explanatory report describing the method (qualitative evaluation)
- Contents of presentation (qualitative evaluation)

●Prize/Prize money
Grand Prize (1 team) Prize money: JPY 700,000
Runner-up (1 team) Prize money: JPY 300,000

■Award for Accuracy (top 3 teams)
1st prize Prize money: JPY 500,000
2nd prize Prize money: JPY 400,000
3rd prize Prize money: JPY 300,000
■Award for Method (several teams)
Total prize money: JPY 600,000
■Special Award (several teams)
Total prize money: JPY 200,000

* The Judging Committee decides the Grand Prize, Runner-up, award for the method, and special award, based on a review.
* The award for accuracy is decided based on prediction errors during the entire period (RMSE).

Uploads_2f24fcac5df73c39fa9257858a43473772_2f2018_r_application_2520guideline

Uploads_2f24fcac5df73c39fa9257858a43473772_2f2018_r_task

[Question setting]
Utilizing the data provided by the Contest Hosts in advance or external data collected by each Participant, Participants are required to predict the sum of power output from solar power plants during the target period.

◆ Given
Participants are to propose method to predict power output of solar power plants of the following day (DAY:d) using his/her own selected data sets which can be obtained by 6:00 pm on the day before the targeted prediction day(DAY:d-1) .

Uploads_2f1e56e132c8455695fea815ca4b649bf8_2f2018_given

◆ The prediction targets
The prediction targets specified by the Contest Hosts are 15 solar power plants connected to power systems in the Hokkaido area (407.3 MW capacity in total).
The targeted solar power plants are selected on the basis of the relatively large scale of their operations and the periods they are to operate within the prediction target period.

Uploads_2f783d0a0894229a752df8382518c83a54_2f07_target

Uploads_2f24fcac5df73c39fa9257858a43473772_2f2018_r_submission

[Items to be submitted]
(Participants)
(1) Predicted power output of solar power plants for the target term
Use the method proposed by the Participant to predict the sum of power output(MW)* of solar power plants during the target period (from 0:00 am, January 1, 2018 to 11:30 pm, January 31, 2019) for successive 30-minute periods. Submit the results in CSV file format specified by the Operations Secretariat. A sample CSV file will be sent to the Participant together with the information that can be used for prediction.
* The method should predict the sum of power output(MW) of specified solar power plants. Please note that the prediction target is not power output of individual solar power plants.

(2) Explanatory report describing the method (summary)
 Write a summary of the proposed prediction method, the data type to be used for the prediction, and any information and details in support of specially designed points, etc.

(Participants Selected for the Final Review)
(1) Explanatory report describing the method (details)
 Write out the details of the proposed prediction method and discuss what can be found from comparison between the prediction results and the correct answer data separately provided.
(2) Presentation video
 Create a video to describe the proposal and its advantages. The video is to be within 5 minutes. Please note that, because the review time is limited, the Contest Hosts will not review any portion of a presentation video running beyond 5 minutes.

(How to submit)
The Operations Secretariat will separately give the eligible Participants instructions on how to format and submit the various items to be submitted.

Uploads_2f24fcac5df73c39fa9257858a43473772_2f2018_r_evaluation

[Method of evaluation]
The submitted items will be evaluated in the following stages. For details, please check the link below. (The link below is a google document page. When you click, a new tab will open. If you cannot access the page, please change your browser setting or contact Contest Secretariat.)
 → See “Selection Criteria”


(First Selection)
(1) Accuracy of prediction (quantitative evaluation based on an evaluation index of prediction accuracy)
(2) Contents of the explanatory report describing the method (summary) (practicality/expandability of the prediction method)

(Final Selection)
The prediction accuracy and practicality/expandability of the prediction methods are comprehensively determined.

Uploads_2f24fcac5df73c39fa9257858a43473772_2f2018_r_data

[Overview of available data]
 Participants freely select the data to be used for prediction. Please note that there are some the limitations applied to the external data that the Participants may obtain and use. For details, please check the link below. (The link below is a google document page. When you click, a new tab will open. If you cannot access the page, please change your browser setting or contact Contest Secretariat.)
 → See "Rules"

If a prediction can be performed using as little less data as possible, the method will be favorably assessed for practicality.

◆ Historical data set for training (provided in advance by the Contest Hosts)
(1) Data on actual power output in the past
(2) Rating output, locations of installation, azimuth angle, and panel-inclination angle
(3) Independently measured amount of global solar radiation, and the measurement locations

◆ External data
Information freely obtained by the Participants

◆ Other
The Japan Meteorological Agency publishes the amount of global solar radiation at seven observatories in Hokkaido. Use them as reference together with the total solar radiation level provided as training data. http://www.data.jma.go.jp/gmd/risk/obsdl/index.php#

Uploads_2f24fcac5df73c39fa9257858a43473772_2f2018_r_schedule

[Schedule]
Following are the dates scheduled for the Contest.

Uploads_2f78b388ef4850644b8a4184096372c942_2f12_schedule_chart

Entry
 - December 20, 2018 (Thu) to February 8, 2019 (Fri)

Delivery of Submissions for the First Selection [Predicted PV power generation output + Explanatory report describing the method (summary)]
 - February 12, 2019 (Tue) to March 15, 2019 (Fri)

Deadline for the preliminary ranking
 - February 22, 2019 (Fri)

Announcement of the preliminary ranking (tentative)
 - March 1, 2019 (Fri) (subject to change)

Notification of the results of the First Selection (to Participants Selected for the Final Review)
 - April 15, 2019 (Mon) to April 26, 2019 (Fri)

Delivery of submissions for the Final Selection [Explanatory report describing the method (details) + Presentation video]
 - May 7, 2019 (Tue) to May 24, 2019 (Fri)

Final Selection
 - Early June, 2019 (planned)

Notifications/Announcement of Prize Winners
 - Late June, 2019 (planned)

Uploads_2f24fcac5df73c39fa9257858a43473772_2f2018_r_awards

[Each prize / award]
Prize Winners for each of the following prizes are chosen from among the Participants Selected for the Final Review* and prize money is awarded.
◆ Grand Prize, Runner-up (one team each)
 Objective: Two teams judged by the Judging Committee to have submitted excellent prediction methods with excellent prediction accuracy
 Prize money: Grand Prize JPY 700,000, Runner-up JPY 300,000

◆ Award for Accuracy (1st to 3rd prize)
 Objective: Top three teams in terms of prediction accuracy
 Prize money: 1st prize JPY 500,000, 2nd prize JPY 400,000, 3rd prize JPY 300,000

◆ Award for Method (several teams)
 Objective: Several teams judged by the Judging Committee to have submitted outstanding prediction methods in terms of practicality or expandability
 Prize money: A total amount of JPY 600,000 will be divided among the awarded teams

◆ Special Award
 Objective: Several teams judged by the Judging Committee to have performed outstandingly on any points other than the above
 Prize money: A total amount of JPY 200,000 will be divided among the awarded teams

* If none of the predictions meet certain criteria, the result for award may be "No Prize Winner."

Uploads_2f24fcac5df73c39fa9257858a43473772_2f2018_r_awards_2520announcement

[Selection results and prize money]
The results of the Final Selection will be published on the Contest Website and the websites of the Contest Hosts. The Contest Hosts will also notify to each Prize Winners separately and pay out the prize money.

Uploads_2f24fcac5df73c39fa9257858a43473772_2f2018_r_eligible

[Qualification for participation] (updated: 2019.02.05)
The qualifications/conditions for participation in the Contest are shown below.
1. No age limits (if a minor applies, approval of a legal representative is necessary).
2. Participation as a group is also possible.
3. If the same Participant uses different methods, multiple entries are acceptable.
4. Participants should be TEPCO CUUSOO account members (Participants must agree with all provisions TEPCO CUUSOO user policy).
5. Participants must agree with all of the provisions of the "Participation Rules."
6. Power industry personnel may participate in this contest. However, individuals working at any of the contest hosting companies maybe asked to refrain from participation.

Uploads_2f24fcac5df73c39fa9257858a43473772_2f2018_r_flow

[How to participate/Flow]
<Entry>
1. Register a TEPCO CUUSOO member account (if not registered)
2. Understand the purpose and rules of the Contest
3. Receive (download) the study data (partial), fill in the entry form, submit a pre-proposal to the Contest Website, and agree to the Participation Rules
4. Notification of entry acceptance*
 * A notification will be sent to the applicant within 7 business days of the submission of the pre-proposal. If the proposal is submitted from December 25, 2018 (Tue) to January 6, 2019 (Sun), the notification will be sent after January 7, 2019 (Mon).
 * If an entry has not been accepted, the applicant can resubmit the entry with modifications (please review and revise the contents of the pre-proposal).
5. The study data (complete version) are received (only by accepted Participants).

◯ At pre-proposal, you will be asked
(1) Technique, Data and Reason (about 50 words):The technique and data you plan to use and the reason for their usage
(2) Intention Plan (about 50 words):Your thoughts on how you plan to tackle the specific conditions (Nature of the area, greater predictions, effects of snow and snow cover, etc.)

In considering pre-proposal, please see the following data samples.
 - sample1_P201601.csv:Sample data on actual power output in the past in the past (from January 1, 2016 to January 31, 2016)
 - sample2_QT201601.csv:Sample data of dependently measured amount of global solar radiation, and the measurement locations (from January 1, 2016 to January 31, 2016)
 → Download the data

(Google drive is used to share files. When you click the link, you will be directed to an external site. If you cannot access the page, please change your browser setting or contact Contest Secretariat.

<When the entry is accepted>
1. Consider a prediction method and perform prediction works based on the training data set (complete version).
2. If prediction results on power output are submitted within the submission period, you can check your tentative ranking and prediction accuracy on the Contest Website.
* You can submit the reviewed method as many times as you want before the submission deadline (the latest submission is used for evaluation. The preliminary ranking will be published only once).
* Only prediction results submitted within the preliminary ranking limit are included.
* The preliminary ranking is optional. This will not affect the First Selection.
3. Create a report on the predicted power output of solar power plants with an explanation of the method (summary)
4. Submit your prediction results and an explanatory report describing the method (summary) by the deadline

<When you receive a notification for the Final Review>
1. Receive the correct answer data (actual values) on power output of solar power plants during the target prediction period
2. Create/submit an explanatory report describing the method (details), including the details of the prediction method, a discussion on a comparison between the prediction results and the correct answer data, and a presentation video of up to 5 minutes in duration based on the report

< When you receive a notification that you have won a prize>
1. Submit information to confirm your identity and bank account details for remittance of the prize money
2. Confirm/notify the receipt of the prize money
3. Discuss or make a contract for a proposal with the Contest Hosts (use of prediction methods, joint study, etc.) (relevant Participants only)

Uploads_2f24fcac5df73c39fa9257858a43473772_2f2018_r_judging_2520committee

[Judging Committee]
The Judging Committee to review and select the Contest Prize Winners will consist of employees of the Contest Hosts who have experience with the processes to predict power output or demand and supply operations in the Contest Host companies, in addition to the following external experts. (Final Selection)

(External judges: honorifics omitted)
◆ Chief Judge
・ Hiroyuki Kita, Dr.Eng, Professor, Hokkaido University
 - Graduate School of Information Science and Technology Division of Systems Science and Informatics
 - Research areas; Electrical and electronic engineering / Power engineering / Power conversion / Electric machinery

◆ Judges
・Masaki Yagami, Dr.Eng, Professor, Hokkaido University of Science
 - Dept. of Electrical and Electronic Engineering Faculty of Engineering
 - Research areas; Electrical machinery engineering / Power engineering

・Ryoichi Hara, Ph.D, Associate Professor, Hokkaido University
 - Graduate School of Information Science and Technology Division of Systems Science and Informatics
 - Research areas; Electrical and electronic engineering / Power engineering / Power conversion / Electric machinery

Uploads_2f24fcac5df73c39fa9257858a43473772_2f2018_r_inquiry

[Inquiries]
To inquire about the Contest, please email the Operations Secretariat.
Depending on the content of your inquiry, the Contest Hosts may contact you directly to respond. Please understand in advance.
Inquiries on the selection review will not be accepted.
Responses to inquiries emailed from December 25, 2018 (Tue) to January 6, 2019 (Sun) will be sent after January 7, 2019 (Mon).

Contact: contest@lib-ag.co.jp
(Operations Secretariat: CUUSOO SYSTEM co., ltd. / TEPCO Town Planning Co.,Ltd. / LiB Co.LTD)

Uploads_2f24fcac5df73c39fa9257858a43473772_2f2018_r_steering_2520committee

[Operational organization]
<Joint hosting>
・ Tokyo Electric Power Company Holdings, Inc.
・ Hokkaido Electric Power Co., Inc.

<Operations Secretariat>
・ CUUSOO SYSTEM co., ltd. / TEPCO Town Planning Co.,Ltd. / LiB Co.LTD

<Backup>
Technical Committee on Frontier Technology and Engineering, The Institute of Electrical Engineers of JAPAN

Uploads_2ffd8c3febb190295e30aae6681e64b0af_2fline

[Special Instructions]
●When submitting an application for entry, all applicants will be asked to accept Participation Rules on the Entry Form.
(The link is a google document page. When you click, a new tab will open. If you cannot access the page, please change your browser setting or contact Contest Secretariat.)

● Use of any furnished training data shall be limited to contest applications. The training data shall not be provided to non-applicants and its use is strictly prohibited for unauthorized activities. In addition, once the contest period ends, all furnished data shall be deleted or destroyed.

●Participants are asked to consult the Contest Secretariat in advance if they are considering using the method constructed for this contest or any results for a university graduation thesis, master’s thesis, doctoral dissertation or submission of a paper to an academic society or other such organization. Preparation of a manuscript without such permission is prohibited. Also, inclusion of any furnished data as-is in a manuscript is not allowed. Applicants and participants are asked to consent to these stipulations.

●Please note that the organizers, people that the organizers deem necessary for judging, people that the organizers deems necessary for considering, and the secretariat shall view and judge the submissions. There are the possibilities that submissions will be seen by members from Contest Hosts' group companies in order to consider the further utilization possibility on business. In such case, information will be handled properly.

●To obtain the right to win prizes in this Contest, Participants are required to hold talks with the Contest Hosts (use of prediction methods, joint study, etc.)


Uploads_2ffda7b2a51939484b6832d4485c818b82_2ffaq_faq

Q. Is the Prediction Target Term prediction of past period?
A. Because the target term is from January 1, 2018 to January 31, 2019, and submissions start from February 1, 2019, the prediction must be of the past.


Q. Why predict the past?
A. Since the actual values of power output of solar power plants are not open to the public, participants are to predict power output of solar power plants using meteorological actual data etc. In this contest, we are seeking methods to accurately predict the pawer output from the given sources.


Q. Is it possible to talk with Operations Secretariat on the phone?
A. In principle, we only accept inquiry via mails. We will correspond individually when there are unavoidable cases.(See Article 3: Contest)


Q. Is it possible to directly contact Operational Organization?
A. Any direct contact such as inquiries, requests, and solicitation is prohibited. (See Article 3: Contest)


Q. Is it possible to refuse prize payment?
A. You can refuse your prize payment after winning. Inform the Operations Secretariat of your intention to decline the prize money when you are contacted as a prize winner candidate.


Q. Can person who belongs to the organizers of Operational organization?
A. Power industry personnel may participate in this contest. However, individuals working at any of the contest hosting companies maybe asked to refrain from participation. Please ask the Operations Secretariat.


Q. Does the "Rating output" of solar power plants represent the output for panels, and not PCS?
A. The rating output data is based on open data of Hokkaido Regional Environment Office site.
In principle, all data used for this contest shall only be information that are open and publicly available. Since not all available sites make public of whether the Rating output is for output of panels or PCS, we are afraid that we are not at liberty to answer.


Q. Is data on actual power output the data on normal power output without considering power output?
A. Yes, that is correct.


Q. It is stated that "azimuth angle and panel-inclination angle" are average values of the solar panel installation area. Are these values simply derived by dividing sum of the solar panel’s azimuth/inclination angle by the number of solar power plants.
A. Yes, that is correct.


Q. Does "Independently measured amount of global solar radiation" represent the horizontal solar radiation?
A. Yes, that is correct.


Q. In respect to "%RMSE" in evaluation, does actual data(ai) represent data for only when electric power is generated? (Does it mean that it does not include data when power isn’t generated?)
A. No. Actual Data(ai) represents data for the whole target date.

When you propose your ideas, TEPCO CUUSOO account is needed.
* Please be aware of the submission deadline.

You can receive information on challenges related to your interest if you register and set your interest information.

Please come and join us!