Let start-ups work with your data to unlock untapped potential!
Natural Gas Price Prediction Natural Gas Price Prediction REALIZED OUTCOME
Optimization of trading activities by reliable forecasting of the natural gas price for the day ahead trading leads to increased efficiency in storage operation, balancing group management and portfolio management.
Die Optimierung der Handelsaktivitäten durch eine zuverlässige Prognose des Erdgaspreises für den Handel am Vortag führt zu Effizienzsteigerungen im Speicherbetrieb, Bilanzkreismanagement und Portfoliomanagement.
Some municipal utility providers would like to optimize their trading activities. Here, we set our focus to one special contract of the natural gas market, the so-called day ahead. Intraday and short-term price prediction of this specific product can help municipal utility providers to optimize their trading activities with regard to storage operation, balancing group management and portfolio management. Can we design algorithms to forecast intraday and short-term price movements in this context?
Like any other municipal provider Kolumbus is trading natural gas in the above context. An accurate forecast of the expected intraday and short-time price movements directly impact trading decisions at Kolumbus. These trading activities are often based on the experience of experts. While algorithms for predicting mid- and long-term prices are studied in several articles, we set our focus to algorithms for intraday and short-term prediction. In order to find proper algorithms for this Kolumbus delivers a variety of data that influence gas prices.
While temperature strongly affects the upcoming demand in winter months, this effect is much weaker in summer months. This nonlinear relationship between temperature and demand is why they are typically modelled using sigmoid functions. Originating from network operators, these functions are also used in the SLP (Standard Last Profil, dt. Standard Demand Profile) to determine the upcoming demand for households and small businesses. Moreover, prices for electrical power and for emission allowances directly influence the profitability of gas-fired power plants. While experts see intraday- and short-term prices for electrical power and emission allowances at trading platforms, these indicators can be used as features in a machine learning model. Additionally, the expected temperature for the upcoming days can be used as a feature. Planned maintenances in the Norwegian gas production are published online and can be used as further indicators. The ultimate goal of this project is price prediction. As an intermediate step, the sigmoid function computed by Kolumbus can be used or a new model to forecast demand based on temperature can be developed. Can we build reliable algorithms to find relations between the mentioned indicators and gas prices? Can we predict prices?
A project duration of approximately 3 months is expected.
What you bring to the table In-depth knowledge of intraday price prediction of other commodities Experience in the market of natural gas Expertise in forecasting and time series analysis
● Prediction of two time frames where the natural gas intraday price is 0.5% higher respectively lower than the average intraday price.
● Prediction of tomorrow’s natural gas day ahead average price. (OPTIONAL) ● Usable model/machine to forecast intraday and short-term price movements. Milestones
The project can be divided into four milestones:
1. The first milestone is reached when the available data has been loaded and visualized. Basic relations between temperature and natural gas demand have been understood. Also, relations between the profitability of gas-fired power plants and the prices for electrical power and emission allowances have been understood as well. 2. The second milestone is reached when a model/machine has been developed that accurately predicts intraday price movements. 3. The third milestone is reached when a model/machine has been developed that accurately predicts tomorrow’s day ahead average price (OPTIONAL). 4. The fourth milestone is reached when the algorithm is outperforming experts. An analysis based on the data sample is presented to and accepted by Kolumbus. HOW TO APPLY?
Request your data sample to tackle the challenge
2. Get your application ready to upload
PDF with your problem solution company details contact person PLEASE NOTE: You cannot save your application during the process, so please have all information ready before you hit the apply button