Forecast hourly energy consumption for a certain period of time

État: Closed
Prix: €1000
Propositions reçues: 7
Gagnant: rahulpattar

Winner

Résumé du concours

We are a software company that delivers niche software for the energy field, especially to energy suppliers.
Our customers (energy suppliers) are buying energy in advance for their customers (energy consumers), based on energy consumption forecast. They need to have as small errors as possible because errors mean financial loss so that is our contest about.

We need a solution for forecasting the hourly energy consumption based on historical hourly consumption.

Things you should know and take into account:
1. The hourly consumption is stored in a Microsoft SQL database (version 2012 or 2014);
2. Energy consumers are different from each other and they have different consumption patterns (for example one consumer might be a office building and another might be a grocery store or a industrial facility);
3. Energy consumption differs from one month to another. For example some consumers might consume more energy on winter and less on summer, or vice versa;
4. Energy consumption might differ from working day to non working day or legal holidays. For example, an office building or a production facility consumes more energy on Monday - Friday than on Saturday and Sunday. We have legal holiday information (schedule) into the database;
5. Energy consumption might differ for some consumers depending on more factors, such as: outside temperature (office buildings), solar radiation (office buildings, solar energy production facilities), precipitation (water pumping stations), wind (wind energy production facilities), etc.. Some energy consumers might depend on some or all of these factors (more or less) and others might not depend on any of those factors. We currently do not have this information (nor historical data, nor forecasts).
6. Forecasts should be made for a certain period of time, from one day to 3 months.

What are we looking for?
We need a solution for the things mentioned earlier, solution that we can use integrated with our software, for forecasting energy consumption based on historical consumption data. This solution should be preferably coded into the database, but that is not a must. We can accept any solution that solves the problem in a elegant efficient way, with as small errors as possible.

What do we expect from you?
First of all, please describe the path that you think is the best. Please describe the advantages and disadvantages of your proposed solution.

Who should apply to this contest?
We need someone that has experience with forecasting; someone who previously did something similar.

We will not take into account the contest solutions that do not contain at least:
1. Solution full description containing advantages and disadvantages;
2. Example of generated forecast using your solution;
3. From what you will write we will determine if you fully understand the request. If we can not determine from your reply that you understand our request, we will not consider your solution. So, your solution should prove you understood the request;

Deliverables:
1. Solution source code;
2. Solution example code for forecasting;
3. Solution compile example for testing purpose.

Also, a bid should be placed for integrating your solution into our platform (our platform is made in C# visual interface with Microsoft SQL database). The consumption is stored into the database as seen in the attached file.

Attached there is a consumption historical data for one consumption point.

Good luck!

Compétences recommandées

Winner

Publiez un concours comme celui-ci

Commentaire de l'employeur

“Great job!”

scringhelteamsrl, Romania.

Meilleures propositions de ce concours

rahulpattar India

1
vinaygupta88 India

0
jabolsoft Philippines

0
sunny2dutta India

0
kmsndp India

0
girofreelancer Italy

0
thampl Croatia

0

Afficher plus de propositions

Tableau de clarification publique

kmsndp
- il y a 7 ans
Respected Sir,
I have commented on my entry #29
Request you to plz check and reply me..
Thanks in advance,
SANDEEP.
- il y a 7 ans
Kunhahammad
- il y a 7 ans
Can you give me chance,I will find out a solution.........
- il y a 7 ans
antonatkachev
- il y a 7 ans
I am now ready with my entry but I can't submit because the contest is closed. If you are still interested I could email it to you for your consideration. In case you've already chosen the winner I do not pretend to get any prize. Just I did the great work to prepare and I feel upset that it can't be used for anything. Please contact me if you are still interested.
- il y a 7 ans
1. TheGuyver2040
  - il y a 7 ans
  Rather than exchange contact information and risk account suspension/termination, Scringhelteamsrl can simply extend the contest deadline. Problem solved.
  - il y a 7 ans
antonatkachev
- il y a 7 ans
SSA is novel model-free approach to time series analysis. This technique has impressive potential for forecasting of stationary processes such as hourly energy consumption in particular. It is widely used for stock prices analysis and forecasting. Also the method is successfully applied to climatic processes, e.g. geological activity and volcanoes' eruptions forecasting.

As my professional interest lies in the field of maths I have developed SSA toolkit for my work purposes and tested it on multiple problems.

Right now I am working on your data. I'll submit the results shortly. Thank you for the opportunity to participate in the contest.
- il y a 7 ans
Voir 1 message de plus
1. Titulaire du concours
  - il y a 7 ans
  You did not applied after all?
  - il y a 7 ans
2. antonatkachev
  - il y a 7 ans
  Dear Sir
  - il y a 7 ans
anru
- il y a 7 ans
This might be a stupid question, but... excel contains data for a single consumption point, correct? And it's one of 2000 in total? I think I read that somewhere in the comments, could be wrong, so, can you please verify?
- il y a 7 ans
1. Titulaire du concours
  - il y a 7 ans
  The excel is indeed for just 1 consumption point and yes in the database might be even more than 2000 consumption points.
  - il y a 7 ans
Titulaire du concours
- il y a 7 ans
To all: Hourly Consumption history is into the SQL database and it gets updated on monthly basis. Also, the hourly forecast should end up into the database as well.
- il y a 7 ans
Titulaire du concours
- il y a 7 ans
To all: please note that the forecast should take into account the winter - summer (last Sunday of October => the day has 25 hours) and summer - winter (last Sunday of March => the day has 23 hours) hour change. This should be made for EET this means that the hour change in March is hour 3 becomes hour 4 (so the day has 23 hours) and in October is hour 4 becomes hour 3 (so the day has 25 hours). For March, in the last Sunday of the month, for hour 03-04 the forecast should be 0. For October, in the last Sunday of the month, there should be 25 hours of forecast, of 24 hours but the hour 03-04 should be doubled.
- il y a 7 ans
1. Titulaire du concours
  - il y a 7 ans
  "of 24 hours but the hour 03-04 should be doubled." = or 24 hours but the hour 03-04 should be doubled.
  - il y a 7 ans
Titulaire du concours
- il y a 7 ans
To all: Please note that the forecast should be made hourly, as the consumption history data is. So, for example if we make a forecast for 02.02.2015, the forecast should be made for hours 1-24 for 02.2015.
- il y a 7 ans
Titulaire du concours
- il y a 7 ans
#6 Please write something. The picture you shared is to small to distinguish something. Please describe your forecasting logic.
- il y a 7 ans
Titulaire du concours
- il y a 7 ans
#8 Please use english. I do not understand the language you used.
- il y a 7 ans
kmsndp
- il y a 7 ans
Request you to kindly comment on my entry, so that I can move forward for the final submission.. Thanks in advance, #7
- il y a 7 ans
1. Titulaire du concours
  - il y a 7 ans
  hi. Thx for your proposal. Lest assume we are making a forecast for the period 02.02.2015 - 09.02.2015. Will the forecast for 02.02.2015 and 09.02.2015 for the Hour 0-1 be the same? If yes, this is our current forecast mechanism and is ok but it is not good enough - it has many forecasts errors. If you go for this path you should take into account other factors as well such as daily consumption distribution in a month and other factors.
  - il y a 7 ans
vinaygupta88
- il y a 7 ans
I would integrate the excel with SQL database (using ODB data connection) to get the data cuts as we require and then provide a forecast which client can view in excel itself. I will do some cosmetic changes to excel file to give a finished product look like a dashboard. Excel will be completely automated and with a press of a button client should be able to view forecast.

If client want the output on the GUI built on C# platform only then only we would need to either integrating the excel with C# or write a separate code in C# without using excel. I think best is to align with client the results first using excel and then we can use that algorithm to build in C#
- il y a 7 ans
1. Titulaire du concours
  - il y a 7 ans
  We intend to use this algorithm to forecast consumption either for one consumption point of for several (or all) consumption points individually. By using a intermediary excel file means that for each consumption point we should populate the excel file first. This would be a slow process since forecasts should be made for 1 - 2000 consumption points in a single click by the application user. This will mean that we should do a loop that populates the excel for each of the 2000 consumption points from the database and then get the result from the excel and insert it into the database. This will be a very time consuming operation. Can you make your transport your mechanism from excel to SQL?
  - il y a 7 ans
freelanmohan7
- il y a 7 ans
What would you like to forecast? daily or hourly energy consumption? 90 days could be long for certain algorithms. What are the minimum days you would like to forecast.
- il y a 7 ans
1. Titulaire du concours
  - il y a 7 ans
  The forecast should be hourly, as the consumption information is. The algorithm should have a minimum of 1 day forecast and a maximum of 90 days. If you feel that your algorithm does not work for 90, we can accept a forecast algorithm that works for a period of maximum 1 month.
  - il y a 7 ans
rahulpattar
- il y a 7 ans
For your point #2 : Where is that information stored? The type of consumer and their consumption patterns. The excel you have attached is only consumption and not the consumer and consumption type. I would write a stored procedure which accepts the number of days of forecast and return the forecast based on the data available. All the business logic will be in db.
- il y a 7 ans
1. Titulaire du concours
  - il y a 7 ans
  Hi. All information is stored into the database. True, I gave only the consumption history. As new consumption is added our application creates monthly consumption profiles for each consumer. Hope this answers your question.
  - il y a 7 ans
Titulaire du concours
- il y a 7 ans
#4 - Please specify your solution with explanation. I cannot see anything in the pictures provided.
- il y a 7 ans
Titulaire du concours
- il y a 7 ans
#2 - Please specify how you think we could use your Excel mechanism integrated into C# taking into account that all of our historical data is stored into a MS SQL Database.
- il y a 7 ans
jeehosong
- il y a 7 ans
What kinds of energy do you prefer?
- il y a 7 ans
Voir 2 messages supplémentaires
1. jeehosong
  - il y a 7 ans
  Thanks. Then should I implement the six problems?
  - il y a 7 ans
2. Titulaire du concours
  - il y a 7 ans
  yes.
  - il y a 7 ans
Dovvosoft
- il y a 7 ans
I am participating, will submit my application architecture by EOD tomorrow.
- il y a 7 ans
antonatkachev
- il y a 7 ans
I am participating in the contest
- il y a 7 ans
cugamelover
- il y a 7 ans
I will work on my entry and submit it later
- il y a 7 ans