Loading ...
Electricity load & price forecasting software solutions & services
- Single AI forecasting software for price, load & solar generation
- Comprehensive AI, Deep Learning & Machine Learning algorithms for energy data
- Feature Extraction, no coding AI, model optimization and forecast accuracy analysis
Electricity load & price forecasting software solutions & services
- Single AI forecasting software for price, load & solar generation
- Comprehensive AI, Deep Learning & Machine Learning algorithms for energy data
- Feature Extraction, no coding AI, model optimization and forecast accuracy analysis
QR AI Forecaster is a universal AI energy forecasting platform designed for all the needs of electricity markets. It offers integrated Data Processing and AI machines.
Common Features
QR AI Forecaster can be used for all your forecast needs in electricity markets: load, energy and ancillary prices, offer stack, and solar generation. Libraries of premade models are available.
QR AI Forecaster is a no-coding AI platform. Users can drag-and-drop and combine various Deep Learning and Machine Learning methods, in a user-friendly dashboard, to create custom forecast models.
Analytics Governance is built in: each AI model and its parameters are visible and auditable. Each execution of an AI run is sand-boxed and parameters, input data and output results are saved individually.
The data frame of predictors, and the parameters of each AI model are configured on the screen in a user-friendly dashboard, without scripting or coding. No prior knowledge of AI, statistics and programming are needed to use QR AI Forecaster platform.
When forecasting load or price, many other time series can be added to the AI model as predictors. These can be multiple weather indicators, demand and supply side data, generation by fuel type including renewables, outages, ISO published hour and day ahead forecasts, consumer load or SCADA. Our AI models make a judicious use of predictors depending on the forecast type. Post-training, the system ranks the predictors by relevance, to help optimize the model.
A different calendar can be assigned to each AI forecast model to replace intra-week holidays’ data with nearest Sunday’s data, both for training and prediction.
You can create AI forecasting models for short term, intraday, up to 7 day ahead, and long term forecast. Forecast resolutions are 5 to 60 minutes. The frequency of short term forecast and publication is every 10 minutes, 24/7.
Each forecast can have upper and lower uncertainty bands computed via quantiles or standard deviation.
You can set the date range for model training, and the forecast horizon.
QR Data Processor & Feature Extractor
Data Processing
This no-coding platform allows users to drag and drop various data (time series), apply the desired mathematical operations and prepare and aggregate the outputs in a “Data Frame” to be used as the predictor input for the AI model.
Multiple time series with different timescales can be used (e.g., weather can be hourly and prices 15 minutes). The Data Processor will load data with the finest timescale and use empty fillers for missing data, awaiting the selected gap-filling method for each data type.
To enact data processing you can drag and drop from the menu a mathematical method to apply to a predictor. These are: simple functions (e.g., logarithm, exponential, averages, etc.), gap-filling, outlier detection and replacement, scaling, quantization, encoding methods. The parameters of each method appear in a popup window for editing upon clicking on the method icon.
You can set the date range for model training, and the forecast horizon.
Feature Extraction
5 categories of predictors are available:
- Internal: a shifted version of the main time series, e.g., previous period/hour/day same-period.
- External: multiple external data sets can be loaded as predictors to improve an AI forecast, by using the hidden relationships across the data sets. These can be external forecasts provided by other sources, e.g., weather, outages, hour/day-ahead forecasts from the ISO, supply side information, fuel type prices and generation. They can also be forecasts generated by our system. E.g., our own ISO system load forecast can be used as a predictor in nodal price forecasts.
- Formula: a mathematical expression to combine different predictors to create a new one.
- Built-in: derived from time index of time series, e.g. hour of day, day of week.
- 30+ Technical indicators from stock market. We often use MA & EMA to smooth excessive noise.
AI Models
QR AI Forcaster Dashboard
- QR AI Forcaster offers a comprehensive set of Deep Learning and Machine Learning methods powered by Python libraries such as “TensorFlow”, “Keras”, “Scikit-learn”. These are preconfigured for electricity load and price data to perform regression for forecasting, or classification (used e.g., in spike predictions, or DA reserve pricing).
- No coding is required, the models and their parameters are accessible in an intuitive web dashboard.
AI & Machine Learning Models
- Deep Learning: QR AI Forecaster offers a modular deep learning layer architecture allowing users to add/remove different types of deep layers via drag-and-drop in an intuitive web-dashboard. Users can then design a wide range of deep models by connecting the layers sequentially or in parallel. QR deep models support multiple data frames as input. The following deep models from Keras library are available: Long Short Term Memory Networks (LSTM), 1D/2D Convolutional Neural Networks (CNN), Recurrent Neural Networks (RNN), and Gated Recurrent Unit (GRU).
- Ensemble Models: QR AI Forecaster offers the following set of ensemble tree-based machine learning models. These distributed gradient boosting libraries are highly efficient, flexible and portable. The models used are: XGBoost, Random Forest, Catboost, ADABoost and LightGBM.
- Single Models: QR AI Forecaster offers some classical models, e.g., decision tree, SVM, KNN, ElasticNet, MLP (neural networks), and Huber Regressor.
AI Model Optimization
Model Splitting
- Electricity data has clear daily profiles e.g., peak, off, super peak, and similarly intra-week blocks such as weekdays and weekend, this is in total 3 x 2 or 6 profiles. QR AI machine allows to define splitters in custom blocks or profiles. Each block is automatically assigned to its own AI model, and its training data is obtained by splitting the main timeseries at run-time according to the selected profiles.
- Model splitting amounts to the automatic creation of a series of AI models that are more robust and accurate since they are adapted to different regimes of data.
Train & Export AI Models
QR AI platform allows you to train an AI model on schedule, e.g., every night at 1.00 am, and save the trained model for execution the rest of the day, e.g., every 15 minutes, while using the latest updated predictors as new data arrives.
Data Dashboard
- Since bad input data negatively impact an AI model, it is crucial to be able to inspect the data of each AI model, namely the target, predictors and forecast data, before and after every stage of data processing.
- All above data are gathered with plots in different sections of the dashboards for quick and easy visualization and inspection by data scientists, saving them the efforts to write code or run scripts.
Forecast Accuracy Analysis
- We offer powerful and flexible no-coding tools to create reports and dashboards, and to visualize data.
- Reports and dashboards can combine multiple data sources, e.g., actual raw data and forecast data, leading to accuracy analysis with any desired error indicator, e.g., MAPE, SMAPE, MAE.
- Dashboards can be created to display side by side the errors of multiple forecasts jobs with different warning thresholds for error reporting, e.g., green for good to red for high error.
QR AI Model Fine-tuner
AI model optimization is a key step in ensuring the accuracy and the stability of forecasts. AI models can have dozens of parameters, and identifying the optimal combination that produces the best result (in terms of accuracy, shape and behavior of forecasts) is very time consuming and difficult. This toolbox automates the complex task of identifying the optimal set of parameters by bootstrapping a genetic computing process around a limited number of forecasts. The optimal set of parameters are those that minimize the forecast error (MAEP) over a date range. This algorithm replaces the equivalent of hundred thousands of manual trial and errors. This tool operates in two different modes, back-test and forward-test.
- In back-test mode, the fine-tuner finds the best set of parameters over a past date range, e.g., one or more days.
- To further improve the results of back-testing, the forward-test mode takes the optimal back-test parameters and estimates the best set of parameters that could have predicted the actual data that has become available over a range of dates.
Features & Predictors Importance Analysis
- Identifying the optimal set of predictors or features is key in optimizing AI models. During training, QR AI Forecaster performs an analysis of the predictors and ranks them by importance or effectiveness, and displays the results in a bar chart plot.
- Users can iteratively add predictors, retrain the AI model and retain the best model.
Explore a full range of price and load Forecast as a Service solutions
AI Load Forecast Case Studies
Our Clients Say
Lets Talk About Your Needs
Our Contact Details:
