Russian Scientists Develop Algorithm to Forecast Power Consumption at Pumping Stations
The system replaces manual calculations and can deliver significant cost savings.
Researchers at Perm Polytechnic University have developed an automated system for calculating electricity consumption at pumping stations, replacing manual methods, the university’s press service said.
In Russia, pumping stations typically operate under the third pricing category (PC-3), which applies a flat, round-the-clock electricity tariff. For consumers able to plan their energy use, a more flexible option exists—the fifth pricing category (PC-5), where electricity prices change hourly to reflect real-time generation costs. To switch to this more cost-effective scheme, consumers must submit an hourly energy consumption plan for the following day, with deviations not exceeding 3%.
Because traditional manual forecasting methods have a relatively high error rate of 6.2%, the researchers developed a machine-learning–based system that reduces forecasting error by a factor of 3.7—to just 1.67%.
Identifying Hidden Patterns for Accurate Forecasting
The researchers say they were the first to propose an innovative solution for accurately forecasting the load of pumping stations while accounting for their specific role in municipal utility services. Their intelligent forecasting system is based on machine learning and can simultaneously analyze dozens of interconnected parameters, including subtle historical dependencies, to identify patterns and generate highly accurate predictions.
The model is trained on an archive of hourly electricity consumption data from a single pumping station in Perm, collected over many years. Using this data, the algorithm can recommend when it is more cost-effective to use the flexible PC-5 tariff and when to stick with the standard PC-3. According to Mishurinskikh, the system could save more than 163,000 rubles (about $2,000) per year for a single station.
The tool can be integrated into existing water supply management systems used by dispatchers or engineers by uploading historical hourly electricity consumption data for a specific pumping station. The program will then automatically collect current consumption data, analyze it based on identified patterns and days of the week, and generate an accurate hourly load forecast for the next day to be submitted to the electricity supplier.
