{
  "fetched_at": "2026-10-19T21:04:28Z",
  "source_url": "https://www.kaggle.com/competitions/energy-forecasting-data-challenge-public/data",
  "pages": [
    {
      "name": "data-description",
      "content": "## Files\n\n*   **train.csv** - the training set. Use this to develop and/or train your algorithm. \n*   **test.csv** - the test set. Use this dataset to make predictions that you will upload as submissions. The `test.csv` includes most of the parameters from `train.csv` but misses `P`, `load`, and `residual_load`. You are asked to forecast the `residual_load`. This can either be done directly **or** by forecasting `P` and `load` individually and calculating `residual_load` (`residual_load`= `load`-`P`). Your submission should only contain information about the `residual_load` ( see `sample_submission.csv`). \n*   **sample_submission.csv** - a sample submission file in the correct format. Your submission should look like this. Except that the predictions may be better 😉.\n\n## Parameters\n\n|    Name      |       Contained in Set (Train, Test)     | Description   | Unit  | \n|----------|------------------|------|------|\n| Time       | Train; Test |  Timestamp    | YYYY-MM-DD HH:MM:SS  | \n| P       | Train | Amount of power generated by the photovoltaic system.   | Watt [W]  | \n| Gb(i)       | Train; Test | Direct in-plane irradiance. It is the fraction of the solar radiation that directly reaches the ground. It is only available if clouds do not block the sun.  | Watt per square meter [W/m^2] | \n| Gd(i)       | Train; Test | Diffuse in-plane irradiance. It is the fraction of solar radiation that reaches the ground after being reflected or scattered by the atmosphere. Its a also available if clouds block the sun. | Watt per square meter [W/m^2] | \n| H_sun       | Train; Test | Sun height (elevation)   | Angle [°]  | \n| T2m       | Train; Test | Air temperature at 2m\t  |  [°C]  | \n| WS10m       | Train; Test | Wind speed at 10m\t  | [m/s]\t | \n| load       | Train | Amount of power required by the plant at this timestep. | Watt [W] | \n| residual_load       | Train | `residual_load` = `load`-`P`. If this value is positive: The amount of power that has to be taken from the grid. If this value is negative: The amount of energy which is fed into the electricity grid.  | Watt [W] | \n| dataset_id       | Train; Test | Train and Test data with the same `dataset_id` belong together  | id | \n\n## Train Test Splits\n\nTo make sure that your algorithm is working well and has not just randomly produced one good prediction, we created multiple train test splits. There are 11 disjunct train test splits in total. To make it easier for you we have marked related Train Test data with a common `dataset_id`.\n\n## Task\n\nCan you predict the `residual_load` for the times contained in the `test.csv` file?\n",
      "isPublished": false,
      "hasIsPublished": false,
      "postTitle": "",
      "hasPostTitle": false,
      "mimeType": "",
      "hasMimeType": false
    },
    {
      "name": "Description",
      "content": "# 👋 Introduction \n\n![Alt Text](https://media.giphy.com/media/nRk3z3sg6Ano6Y7gX2/giphy.gif)\n\nThe decentralized expansion of renewable energies is an important pillar on the way to the energy transition. Self-generation systems on industrial and commercial sites are a major field of application. Primarily, customers consume the electricity they generate themselves and purchase electricity if the quantities generated are insufficient to completely cover their energy load. The resulting residual load (residual load = energy load - self generated energy) must in any case be provided by the energy supplier (aka. the grid). Note, that the residual load can also be negative if the amount of self generated energy is greater than the load. In this case, the excess energy is fed into the electricity grid.\n\nIn order to ensure a stable energy supply, the energy suppliers rely on forecasts of these residual loads. In the past, those residual loads could be forecast based on many years of experience. Now, the share of solar energy systems and therefore the amount of energy self-generation increases more and more. This makes forecasting the residual loads more and more complex as there are dependencies on external factors like the weather. However, the forecast is necessary to maintain a proper supply operation. For this reason, it is essential to have a forecast of the residual load that is as accurate as possible.\n\n# 🚀Problem Statement\n\nParticipants are asked to develop an algorithm that forecasts the residual loads. You can find more technical problem descriptions and the data in the [data tab](https://www.kaggle.com/competitions/energy-forecasting-data-challenge/data).\n\n# Organizers\n\n|![](https://www.googleapis.com/download/storage/v1/b/kaggle-user-content/o/inbox%2F1811734%2F9253fffaf8bf37bba5b85c3e93f813b6%2Fhessian_ai.png?generation=1669728657947391&alt=media)   | ![](https://www.googleapis.com/download/storage/v1/b/kaggle-user-content/o/inbox%2F1811734%2F48e1cafdec6ab6371820f7e7650de5db%2Fhoe.jpg?generation=1669728658280042&alt=media)   | ![](https://www.googleapis.com/download/storage/v1/b/kaggle-user-content/o/inbox%2F1811734%2F8761c0dff398f0b9859f2b027216e1fd%2Faisr.png?generation=1669728658607042&alt=media)   | ![](https://www.googleapis.com/download/storage/v1/b/kaggle-user-content/o/inbox%2F1811734%2F7e8796d39ca2b60ab0a798ff6a3483d3%2FSTW_AG_4c.jpg?generation=1669728658510958&alt=media)  |\n|:-------------:|:-------------:|:-------------:|:-------------:|\n| [hessian.AI](https://hessian.ai/)| [House of Energy](https://www.house-of-energy.org/) | [AI Startup Rising](https://hessian.ai/entrepreneurship/)| [Städtische Werke AG, Kassel](https://www.sw-kassel.de/privatkunden/startseite/)|)",
      "isPublished": false,
      "hasIsPublished": false,
      "postTitle": "",
      "hasPostTitle": false,
      "mimeType": "",
      "hasMimeType": false
    },
    {
      "name": "Evaluation",
      "content": "The participants are asked to develop an algorithm that forecasts the data from the `test.csv` file. The forecast error which is used to determine the accuracy of the model is the **Root Mean Square Error (RMSE)**. The RMSE is also used to determine the scores in the leaderboard as well as to assess the accuracy of your solution in the end. Note that only parts of the data from your submission are used to score your submission in the leaderboard. Once the challenge is over we will use all the data from your submission to determine the best-performing algorithm. So don't waste your time trying to overfit the leaderboard ;-)",
      "isPublished": false,
      "hasIsPublished": false,
      "postTitle": "",
      "hasPostTitle": false,
      "mimeType": "",
      "hasMimeType": false
    },
    {
      "name": "rules",
      "content": "Don't cheat, be fair, be friendly, \n\nand most importantly have:\n\n![Alt Text](https://media.giphy.com/media/lmv5aDvwOgTmby3a13/giphy.gif)",
      "isPublished": false,
      "hasIsPublished": false,
      "postTitle": "",
      "hasPostTitle": false,
      "mimeType": "",
      "hasMimeType": false
    },
    {
      "name": "Frequently Asked Questions",
      "content": "",
      "isPublished": false,
      "hasIsPublished": false,
      "postTitle": "",
      "hasPostTitle": false,
      "mimeType": "",
      "hasMimeType": false
    },
    {
      "name": "Contact",
      "content": "",
      "isPublished": false,
      "hasIsPublished": false,
      "postTitle": "",
      "hasPostTitle": false,
      "mimeType": "",
      "hasMimeType": false
    }
  ]
}