長葆工業有限公司目前可以替國內風電產業服務的能量如下:
1.風資源評估,陸上與海上,機艙測風雷達與功率曲線量測,尾流與風場優化
2.全釩液流電池儲能
3.GWO BST and BTT虛;實培訓系統
4.ONYX風機預測性維運CMS,Pitch Monitor system , fleetMonitor, AI hub
5.丹麥Jomitek風機雷擊強度記錄裝置與葉片避雷效能偵測
6.DMFC直接甲醇燃料電池、高續航力氫燃料電池無人機在風機的巡檢應用,歡迎業界先進多多指教 !
1. If you had to define this service in one sentence
“Fleet monitor for SCADA data”
2. What is the roll out time/turnaround time?
Two weeks after we receive all the SCADA data. The aim is to go down to 3 days.
3. What is the amount of data we need? How many months’ worth to create the first report?
We will need the data of the entire site and minimum 10 turbines (all turbines so that our statistical models which compare turbines have enough samples), and at least 1 year so the data is representing all the different varieties in which the turbines operate.
In terms of type of data:
- SCADA 10 minutes data,
- SCADA event data (nice to have)
- Metadata (latitude and longitude), turbine make and model
On section 5.4 we’ll find data requirements on the proposal document
4. Minimum number of turbines needed
More than 10 would be ideal and would be ok. Less than 10 statistical models would not work as well and would be problematic. Statistic models relay on average data. Fewer turbines (6 or less) would require individual and manual examinations.
5. What is the data needed for the second report?
If it is periodic, we will set up a data connection to it, so the data is automatically sent. About 2 weeks before the report is due the data will be dumped, and a report will be generated and sent to the person in charge of data analysis before sending the final report to the customer.
6. Can this be used for direct drive turbines?
This would absolutely work for direct drive turbines, our statistical models work on any models as they are not built for any specific make/models.
7. What is the customer pain point?
We can spend more time on the models and address issues the customer is not addressing, like pitch accumulators. Sophisticated models must be built to guess what is causing a problem.
Secondly, we try to assign lost energy to all these issues, building a priority list through the ones that are causing biggest energy lost, and continuing with a list of issues that are not yet causing energy lost but will do in the future.
8. Key differential with AIHub?
On AiHub we offer all the data in one place including Case Management, plus the expertise of vibration.
9. Any product that does vibration and Scada in one tool?
No tool yet to date. We are focused on predictive maintenance; we are not trying to go to second SCADA data. In this space, there is no tool that does vibration and SCADA together.
Greenbyte system: compared to them we can say why is the issue being caused and what to do to solve it.
10. When is the AI kicking in? It is all historic data, right?
It is all historic, we are working on a time machine for the future. The recommendations we say now are usually along the lines we have noticed this is getting bad, we recommend this action. We are not yet able to predict remaining useful life but is in scope for the upcoming months.
11. When is the machine learning coming in?
The current analysis is based on a machine learning model which predicts, for example, what the temperature of the bearing should be, to then compare it and detect any overtemperatures. Then alarms are established automatically for that data exceeding the thresholds.
1.風資源評估,陸上與海上,機艙測風雷達與功率曲線量測,尾流與風場優化
2.全釩液流電池儲能
3.GWO BST and BTT虛;實培訓系統
4.ONYX風機預測性維運CMS,Pitch Monitor system , fleetMonitor, AI hub
5.丹麥Jomitek風機雷擊強度記錄裝置與葉片避雷效能偵測
6.DMFC直接甲醇燃料電池、高續航力氫燃料電池無人機在風機的巡檢應用,歡迎業界先進多多指教 !
1. If you had to define this service in one sentence
“Fleet monitor for SCADA data”
2. What is the roll out time/turnaround time?
Two weeks after we receive all the SCADA data. The aim is to go down to 3 days.
3. What is the amount of data we need? How many months’ worth to create the first report?
We will need the data of the entire site and minimum 10 turbines (all turbines so that our statistical models which compare turbines have enough samples), and at least 1 year so the data is representing all the different varieties in which the turbines operate.
In terms of type of data:
- SCADA 10 minutes data,
- SCADA event data (nice to have)
- Metadata (latitude and longitude), turbine make and model
On section 5.4 we’ll find data requirements on the proposal document
4. Minimum number of turbines needed
More than 10 would be ideal and would be ok. Less than 10 statistical models would not work as well and would be problematic. Statistic models relay on average data. Fewer turbines (6 or less) would require individual and manual examinations.
5. What is the data needed for the second report?
If it is periodic, we will set up a data connection to it, so the data is automatically sent. About 2 weeks before the report is due the data will be dumped, and a report will be generated and sent to the person in charge of data analysis before sending the final report to the customer.
6. Can this be used for direct drive turbines?
This would absolutely work for direct drive turbines, our statistical models work on any models as they are not built for any specific make/models.
7. What is the customer pain point?
We can spend more time on the models and address issues the customer is not addressing, like pitch accumulators. Sophisticated models must be built to guess what is causing a problem.
Secondly, we try to assign lost energy to all these issues, building a priority list through the ones that are causing biggest energy lost, and continuing with a list of issues that are not yet causing energy lost but will do in the future.
8. Key differential with AIHub?
On AiHub we offer all the data in one place including Case Management, plus the expertise of vibration.
9. Any product that does vibration and Scada in one tool?
No tool yet to date. We are focused on predictive maintenance; we are not trying to go to second SCADA data. In this space, there is no tool that does vibration and SCADA together.
Greenbyte system: compared to them we can say why is the issue being caused and what to do to solve it.
10. When is the AI kicking in? It is all historic data, right?
It is all historic, we are working on a time machine for the future. The recommendations we say now are usually along the lines we have noticed this is getting bad, we recommend this action. We are not yet able to predict remaining useful life but is in scope for the upcoming months.
11. When is the machine learning coming in?
The current analysis is based on a machine learning model which predicts, for example, what the temperature of the bearing should be, to then compare it and detect any overtemperatures. Then alarms are established automatically for that data exceeding the thresholds.