USER SUCCESS STORY
How an expert consultancy built physically plausible digital twins for power plants more effectively.
Visplore allows us to quickly identify appropriate data packages in the plant history, eliminate outliers from the data, and effectively analyze the resulting clean data of the physical processes. – Performance Engineering Expert, Expert Consultancy Company
Since electric power cannot be stored in sufficient quantities, power plants must follow the demand of the grid under all ambient conditions. A purely data-driven digital twin based on historical records of the plant is therefore prone to failure, as the records do not cover all possible situations that may occur.
Therefore, an Austrian expert consultant company developed a new approach for digital twins based on physical process models that are refined utilizing historical data but safely allow for extrapolation into formerly ‘unknown’ regions. To adjust such models to the plant as built, representative operating points must be derived from the plant records for the tuning of the performance characteristics.
The engineers needed to explore and validate historical data of the power plant to find those time periods or situations where ‘good’ data for the adjustment of the models to the actual performance of the plant were available.
An important part of this screening was to identify implausible time ranges and to exclude them from model tuning. On the other hand, various sets of specific load conditions had to be selected to make sure that the physical model covers representative data over the entire operating range.
By choosing Visplore, the expert engineers were able to review hundreds of time series and were guided to time ranges or situations needed for model tuning. Despite facing thousands of data rows, they were able to interactively zoom into areas of interest and check them in further detail.
By visualizing correlations for many different operation parameters, the performance engineers could quickly check the plausibility of the data. Utilizing Visplore’s advanced display and drill-down features, they were able to identify data quality issues such as outliers or periods of transient operation, and tuning data sets could be cleaned for modeling right away.
Visplore’s ready-to-use cockpits reduced the time for exploring and cleaning the data set enormously. Due to the interactive nature of Visplore’s pre-configured cockpits, the engineers were able to gain significantly higher confidence in the data.
Eventually, this also allowed the company to save around $10,000 of cost per project. Moreover, detailed checks of the models’ physical plausibility can now be conducted much more regularly and very efficiently by comparing model predictions with current data from the plant. Being able to quickly identify deviations and to verify that these deviations are caused by performance changes and not data quality issues, the digital twin can be updated much more frequently to ensure high accuracy of the simulation results over time.