Discover our solutions for electricity trading companies.
We support trading companies with various algorithms. This ranges from clustering of consumers, prediction of energy consumption or production from RES, to analysis of data from an energy distribution system operator, from electricity consumption point and analysis of data from the electricity networks coordination plan and improvement of forecasts for the following day. We also support trading companies and their RES generators with technical and investment analyses of the siting of energy storage facilities at the RES source.
For a given PV or wind farm, we provide a prediction of energy production for the following day. We do this by calibrating historical weather forecasts against actual production in advance. This sets us apart from many online services.
We build clustering and machine learning algorithms to automatically categorise consumers into different types. This is how we create predictions of the PPE consumption profile for the next day and the entire annual profile.
We support trading companies and generators with a renewable energy source with analyses of the profitability of placing energy storage next to the source. Ultimately, we can control such storage with algorithms that determine the optimal strategy for each day.
We used neural networks to forecast the coordination plan (in particular the operation of coal units and RES sources on a national scale) more accurately than electricity networks.
We have developed algorithms that allow us to automatically cluster energy consumers into different categories using data remotely collected by OSD from meters at points of consumption. For each category, we define a representative weekly energy consumption profile, which is calibrated by the volume of consumption and the seasonality and repeatability of the data from week to week. This is the starting point for making a highly accurate prediction of energy consumption across the trading company's customer portfolio. This prediction can be provided both for the next trading day (for contracting in the energy market), as well as on a seasonal and annual basis (for balancing the annual profile of all energy deliveries and energy off-takes).
We have built a neural model that makes more accurate predictions of power by coal-fired generating units and the operation of RES sources in the Polish electricity system than PSE. The model periodically analyses forecast data from the coordination plan and determines corrected, more accurate predictions of individual parameters of the national power grid. The more accurate prediction allows the trading company to make its own algorithms to forecast prices in the energy market and make more profitable contracting.
In cooperation with trading companies, we provide technical and investment analyses for generators interested in siting energy storage at a RES source. We carry out financial analyses with a focus on arbitrage, the capacity market, and the sale of RES energy at high peak prices.
We show the expected revenue in three scenarios using different strategies to control the charging and discharging of the storage. As a software company, we are able to ensure that these strategies are realistically implemented on the target storage once it is built.
We trained neural network-based models that, based on information from the PSE coordination plan, made a price prediction for each hour of the next day in the Day-Ahead Market. Our most accurate neural models were able to achieve an accuracy of 78 per cent of the price prediction with an error of less than 7.20 PLN/MWh, while 92 per cent of the predictions were within the +-20 PLN/MWh range.
We built both very simple, polynomial regression models for initial price estimation or the hours of the morning and evening peak in energy prices, as well as trained neural models to take into account multiple factors.
We have developed regression, neural and equation-based algorithms for renewable energy production forecasting.
Reliable renewable energy production forecasting should consist of several elements: calibration of historical weather forecasts for a given location against actual observations (as weather forecasts contain not only obvious random error, but also systematic error - the systematic error should be fully zeroed out).
The second element of the model is site characterisation. It can be based on the technical data of the PV panels or wind turbine. Preferably, it is plotted from a point cloud derived from a SCADA system or similar - as the effect of cut-off on the inverters, e.g. as a result of a power guard, etc., can be observed. For interested trading companies or generators, we calibrate such models and provide forecasts every morning.
dr inż. Bartosz Górecki – Założyciel QuickerSim
Telefon: +48 503 444 886
E-mail: b(kropka)gorecki(malpa)quickersim.com