Design more robust and better performing equipment using experimental design and multivariate analysis
Reduce costly unscheduled maintenance and machinery downtime with powerful multivariate prediction models
Optimize equipment and machinery use by understanding process and environmental factors affecting them
Generate sustainable energy more efficiently and cost effectively
Advanced data analysis tools can help renewable energy companies operate more efficiently, from R&D – such as developing more efficient solar panels and more robust wind turbine blades – to solving common problems such as predicting turbine breakdowns and optimizing the use of equipment depending on local conditions and demand.
Multivariate data analysis is used extensively in industries such as pharmaceuticals, chemicals, biotechnology, semi-conductors and agriculture to understand large, complex data sets that traditional statistics are inefficient or unable to handle. These powerful but relatively easy to use tools are now being applied by the energy sector to help improve their product and process performance.
Camo Analytics’s solutions can be integrated with existing control systems, accepting proprietary data formats from sensors, industrial databases (ODBC and OSI PI) and process instruments, using protocols such as Modbus, Profibus, TCP/IP and now OPC (DA, HDA and UA capabilities). The graphical outputs make interpretation simple and can be customized for all user levels, from expert to operator.
REAL BUSINESS BENEFITS
- Use design of experiments and multivariate data analysis to design more efficient solar cells
- Develop more robust turbines or blades using advanced materials
- Avoid production shutdowns with early event detection using multivariate predictive models
- Implement preventive maintenance programmes based on powerful regression models
- Understand the impact of environmental conditions to increase the lifetime of equipment
- Get an overview if a process is in control or requires intervention
Case studies and example applications
Case study: Wind turbine monitoring
A major European renewable energy firm operates wind farms with several hundred turbines in each. In one particular farm of 226 turbines, 28 had faults resulting in breakdowns. The client wanted to be able to predict which turbines were likely to malfunction in the future.
Each turbine had a number of sensors mounted on them so they could measure vibrations, temperature, pressure, load etc. In addition, each turbine had associated historical data including the number of hours of operation, total amount of energy produced and the location.
To begin the project, all the data from the sensors over a certain period of time was collected together with the historic data and analyzed using multivariate methods, from which a multivariate predictive model was developed.
Statistical methods were then used to define when production was ‘normal’ or within the sweet spot of optimal operating parameters.
Next, new observations (sensor data) could be projected onto the multivariate model developed earlier to identify deviations from normal situations. For situations outside the defined limits of normality, operators were able to see which variable (parameter or load) had changed. The resulting model was able to identify 100% of the wind turbines with failures. This could be applied in real time to monitor turbines and even the entire wind farm.
Case study: Product development for wind turbine blades
A global wind turbine manufacturing client used multivariate data analysis and design of experiments (DoE) to develop turbine blades constructed from advanced materials. Using DoE they were able to determine the optimal design parameters to produce more reliable, durable blades while analyzing quality parameters with multivariate analysis.
Example application: Predictive maintenance in a hydro-electric facility
A major European hydro-electric company had persistent problems with a turbine breaking down resulting in long periods of downtime and expensive maintenance. Multivariate data analysis could be used to analyze historical data to determine the conditions leading to breakdown. After instrumenting the turbine, the multivariate model could be refined with additional data points from sensor readings over time to improve accuracy. This approach enables systems to be put in place to alert operators in real time when a breakdown is imminent, thereby allowing them to implement preventive maintenance before failure occurs.
DON’T WASTE YOUR VALUABLE DATA
Most energy manufacturers collect an enormous amount of data from sensors, yet the majority do not exploit its full potential due to the perceived difficulty and lack of statistical knowledge. However, today’s data mining and analytical tools are much simpler to use and even more powerful, enabling industry leaders to get valuable insights from their data which are driving significant business improvements.