Energy Data Analysis and Machine Learning: Real-World Examples
Approved for 1 AIA LU
Energy usage data is critical to assessing and improving building performance. A variety of tools and techniques are used to work with energy data. State-of-the-art software tools and software algorithms can significantly streamline the process of working with data and uncover relationships and insights that are very hard to do manually. Machine Learning (ML) is one of the techniques seeing increased use and a fair amount of hype. While ML is not a silver bullet for all energy data analysis applications, ML is much more than a buzzword. There are numerous scenarios where you can start using ML right now to benefit from deeper, more efficient energy data analysis. In this webinar, we will see real examples of ML for evaluating energy consumption (with hourly, daily, weekly or monthly aggregation) using regression models in a single tool that allows users to focus on the data trends and relationships without tedious manual processing, such as manually exporting data to spreadsheets and using chart trend lines. One of the most significant advantages of an ML-enabled approach is avoiding potential erroneous data transfer, resulting in more efficient energy analysis. It also allows you to get results in near real-time, quickly identify outliers in your data, and choose the best model depending on metrics such as R2, RMSE, or model shape validation. Challenges arising from the analysis of a large buildings’ portfolio will be addressed in the webinar, as well as real-world examples.
- Energy consumption regression models – model types, purposes, and limitations
- Major difference between daily and hourly aggregation in energy data analysis
- Importance of automation of energy data preprocessing and models management
- De-mystifying ML through specific use case examples
SkyFoundry‘s SkySpark analytics platform automatically analyzes data from building automation systems, metering systems and other smart devices to identify issues, patterns, deviations, faults and opportunities for operational improvements and cost reduction. SkySpark is an open platform enabling data from a wide range of sources to be continuously analyzed, helping building owners, operators and engineers “find what matters” in the vast amount of data produced by their equipment systems. SkySpark is an ideal fit for commissioning and M&V applications.
Energocentrum Plus focuses on research and development in the field of energy and HVAC. It develops its own cloud system Mervis SCADA and PLC runtime Mervis RT/IDE. Its research focuses on machine learning and fault detection. Machine learning methods for the energy analysis are implemented in the Energy Twin product
Jan ŠirokýEnergocentrum Plus, s.r.o.
Jan’s doctoral studies were devoted to fault detection and predictive control. His current research focuses on machine learning for energy and HVAC data analysis. He is the head of the R&D team that develops the Energy Twin tool – an energy data analysis tool. He is involved in a number of practical applications of machine learning in the energy sector.
John PetzeCo-Founder and COO, SkyFoundry
John Petze is a partner and Co-Founder at SkyFoundry, the developers of SkySpark™, an analytics platform for smart device and equipment data. John has over 30 years of experience in automation, energy management and M2M/IoT, having served in senior level positions for manufacturers of hardware and software products . He is the Executive Director of Project-Haystack.org.