KAStrion project description
The production energy of wind farm is becoming more and more important in several countries. Its impact on the electrical grid grows every day, at the country and continental grid scale. The economic value of the wind farm does not only reside in the amount of energy produced but also in the capacity to forecast this energy. The forecast is built on production capacity available and wind prediction. The availability of the equipment and therefore the capacity to plan the maintenance based on a reliable condition monitoring of the equipment are crucial challenges.
KAStrion is an extension of a first technology transfer project AStrion. This project aims at integrating AStrion with the product VIBex, a vibration monitoring system already used in over 100 installations on wind turbines.
KAStrion vision and challenge
In that context, KAStrion aims at improving the production capacity forecast with a set of innovative technical solutions that would improve existing ones or eventually replace them.
KAStrion deliverables is a solution made of two modules.
- The first module is embedded in the wind turbine and does advanced automatic analysis of polyphase electric and vibration signals to monitor its state of health.
- The second module centralizes all the wind turbine analysis results in order to display the results to experts for final decisions and actions.
During the project journey, our goal is to regularly field test uncompleted versions of the solution to rapidly improve our experience.
KAStrion main differentiators
To achieve the highest market value, KAStrion is built from the gate go with the best end-user implication one could have, as end-users are partners of the project (VALEMO, MECAL). The strength of the proposed system is the design of a fully automatic system based on innovative approaches developed by the partners, with 3 differentiating factors:
- First with a specific data validation
In a data preanalysis, KAStrion will sort measures in order to consider only those without sensor problems or with less wind variations.
If an incorrect time signal is accepted by the system, the numerical outcome of the signal feature extraction is unpredictable, and the chances for data error detection are relatively low. Actually, the afterward incorrectness detection depends on the ratio of the number of signals permanently stored to the number of all registered signals, and generally is less than 1 %. Consequently, a certain trend value calculated from an invalid signal may ruin the system analysis, e.g. it may yield extreme false alarm levels.
KAStrion will integrate recent signal processing approaches developed by two partners of the consortium.
Specific time descriptors have been developed for wind turbines in order to detect invalid signals due to acquisition problems. The assessment of data acquisition, namely the correctness of signals recording process, is required for a relevant evaluation of any signal features as well as for legitimate storage of raw waveforms in the system database. The latter aspect is especially important in terms of gigabytes of disk space frequently wasted for corrupted data.
Moreover in wind turbine analysis, the input of the system, the wind, is completely non-stationary. Strong non-stationarities will also ruin the system analysis. This last point is fundamental in wind turbine analysis.
For detecting non-stationarities, KAStrion will consider a non-stationary index which has been defined in order to control the invariance of the time-frequency statistics in order to select the level of non-stationary in the measures whatever the nature of the measure (vibration or electrical one). This test is computed on a spectrogram of the measures. In addition, methods based on time rupture detection will be applied on a time gliding power and a time gliding Kurtosis in order to reinforce the time index.
The data validation in KAStrion will be completely automatic.
- Second with an automatic peak detection and classification
Classical vibration analysis is based on thresholding spectra. The choice of the thresholds and of the frequency bands on which are applied these thresholds are critical for an earlier detection.
KAStrion will propose a completely different approach illustrated in Figure 1 where all peaks of the spectra are automatically detected based on advanced signal processing methods. KAStrion will groups the detected peaks in harmonic families and side band modulations in order to highlight failure signatures. This method is completely automatic and does not need a priori parameter settings.
Given that the structuration of the peaks in the spectrum is considered instead of only amplitude levels in some frequency bandwidths, earlier detection is possible. Moreover the results do not depend on data base.
- Third with a polyphase electrical analysis
Some failures are visible in current signals easily accessible compared to vibration one. In addition with the previous analysis, KAStrion will propose a completely new approach.
Most sensors in such system provide three-dimensional physical quantities, such as three-phase electrical measurements or three-axis mechanical measurements. Classical approaches rely on the analysis of such quantities without taking into account their three-dimensional nature, and a crucial part of the diagnostic information is thus ignored. KAStrion will integrate new processing and analysis methods dedicated to 3D quantities in order to highlight the geometric characteristics of such data.
Finally, in KAStrion, each result will be coupled with the kinematic of the machine in order to indicate the system parts in failure. Indicators will be also provided to quantify the severity of the failure.
A final module will merge the information coming from the different sensors (vibrations and currents) and all the previous approaches in order to automatically provide to the user a comprehensive report about the failures of the system.
Each partner of the KAStrion project provides a strong expertise that supports the construction of the solution. From academic signal processing and reliability expertise, to wind farm day to day operational management, KAStrion is well setup to deliver its vision.
A strong link is established with the turbines manufacturers through VALEMO which is working with 7 different turbine manufacturers.
The GoToMarket for the wind turbine/farm application will be handled by EC-SYSTEMS. The GoToMarket for any other application of the solution will be discussed by the partners. Eventually, each partner will also developed its own GoToMarket strategy for the technical bricks that it owns.
All this will be detailed in the consortium document with the associated IP.
Standardization is key to wind farm operations. KAStrion will address this with the support of CETIM and VALEMO. CETIM is highly involved on the subject participating to several committees. It is involved in ISO TC 108 (machines condition monitoring), and following the work of the recently created WG 16 (condition monitoring in wind turbines).
Ensure proper operation lifelong of wind turbines is one of the aim of today. Standards are mainly focused on people safety but it is also necessary to ensure material prevention to ensure maximum reliability of the machine. VALEMO will consult insurer companies, operators and manufacturers, such that KAStrion system will be developed to satisfy those requisite. It is important not to forget that insurance is the previous step of bankability. VALEMO participates to the cluster Aquitaine which has set up a working group on this issue, bringing together developers, operators, manufacturers and insurance companies (Filhet Allard).
Based on academic research labs, technical centers, industrial companies and production plants, the originality of KAStrion project lies in the crossing of high competencies in the full value chain of the condition monitoring of a mechatronic system from the data captures to the final decision-making in order to improve the production goals. KAStrion gathers know-how
- in mechanical and electrical engineering for the physical understanding of the system, in signal processing for extracting information from the measured data,
- in reliability engineering and system safety for predicting the life time of the system,
- in computer science and control system for designing and developing an embedded prototype and a remote diagnostic center,
- in renewable energy production for testing and validating the system in operating wind farms.
KAStrion will provide a condition monitoring solution for both the electric generator and all the rotating parts inside the nacelle (the different gears, bearings and shafts).
KAStrion work plan overview
Following the development of KAStrion, the test and validation will have three phases:
- The proposed prototype will be first tested and validated using a test bench reproducing the wind-turbine kinematic and enabling non-stationary load condition. The test plan will be defined in order to match the market needs specified with VALEMO, MECAL and CETIM.
- Then, two onshore wind turbines of different manufacturers will be instrumented to evaluate the prototype performance in real condition in order to assess the turbine components wear (gears, rolling bearings,..).
- The equipment of a group of several turbines from different manufacturers will be setup for a six months test period to validate the Condition Monitoring System on a larger scale. Each test result will be confronted to experts in the field. Finally, the consortium partners will do their best to have the opportunity to extend the prototype test to offshore turbines. Results will also be validated by experts in the field.
In phase 2 will be tested the system operation, its performance in a test context on turbines devoted to this context and allowing some actions on the operating conditions to test the different part of the system. The scale of the test in phase 3 will be in a larger scale and in continuous time. In phase 3 the system will be installed on operating turbines and used by the regular teams in order to test its use in a real-world context and then to validate its ability to detect evolution of the state of the turbine.