Fill out an application for the project, we will contact you as soon as possible and answer all your questions.
Enterprise Electricity Consumption Forecasting System "EMAS.FORECAST"
"EMAS.FORECAST" is a universal forecasting system that predicts required parameters using machine learning models.
Input Data
Weather forecast
Weather archive
Data preparation
Missing value imputation
Correction of known false values
Feature generation affecting values
Time shifts
Day types
Feature selection and weighting
Selection of most relevant features
Determining weight for each feature
Model deployment
Machine learning execution
Model creation and deployment to production
Maximum achievable forecast accuracy
Evaluation
Monitoring
Retraining
Product Applications
- Consumption forecasting:
- for energy retail companies
- for large enterprises
-
Generation forecasting:
- for solar power plants
- for wind power plants
- Peak hour forecasting (for various organizations: both for consumers to minimize capacity payments and for generation companies)
- Nuclear power plant maximum generation forecasting
- Day-ahead market price forecasting, heating network parameters and other commercial indicators (as additional data source for EMAS.OPT)
Purpose
EMAS.FORECAST hourly electricity consumption forecasting program was developed for implementation in wholesale (DAM) and retail electricity market participants requiring improved consumption forecast accuracy and consequent reduction in electricity procurement costs.
The enterprise electricity cost optimization system includes implementation of software that enables:
- Collection and processing of all required input data for forecasting (generation facts, climate condition archives, climate forecasts);
- Data "preparation" including removal of known false values, feature generation and enrichment;
- Building and training forecasting models;
- Performing consumption/generation parameter forecasting based on models;
- Automatic model accuracy determination on current data and self-calibration (retraining) when needed.
Key Program Features
- Electricity consumption volume forecasting;
- System Operator peak hours forecasting;
- Electricity cost optimization;
- Preparation of highly accurate electricity consumption bids by large enterprises for energy suppliers;
- Cost reduction for electricity consumers.
Presentation
Development process includes:
- Collection and processing of input data (both statistical data from customers and open external data like weather data, ATS and System Operator websites etc.);
- Data preprocessing, outlier removal, new feature generation based on input data;
- Creating a pool of various models to select the most suitable for the task and choosing the best model by unified metrics;
- Establishing criteria for model retraining when forecast accuracy drops or after certain time intervals;
- Real-time testing;
- Product integration into customer infrastructure/provision as a service.
Advantages
- EMAS.Forecast module's microservice architecture easily integrates into business processes;
- Automated real-time data collection with guaranteed completeness from all sources required by both energy retailers and industrial enterprises;
- Clear forecast result analysis compared to actual data;
- Easy report generation by customer departments (no more difficult than working with Excel pivot tables);
- Integration with external information systems;
- High variability in analytical information presentation methods;;
- Information availability anytime anywhere - "thin client" technology (web portal) enables process control from tablets/mobile phones;
- EMAS.FORECAST module's mathematical algorithms use advanced data processing technologies and mathematical modeling methods in Python;
- EMAS.FORECAST production planning program achieves maximum forecast accuracy, improving electricity market efficiency and maximizing company marginal profit.
Interface Examples
