Global Tis 25 Key Generator
The entire SGen-1000A series is optimized for world-class efficiency of up to 98.9% and stability during load changes and flexible operation. They are designed for 10,000 start/stop cycles during their operational life, making them a cost-effective alternative to hydrogen- or water-cooled generators for industrial applications up to 370 MVA.
Global Tis 25 Key Generator
Our experts have designed a product line that meets all major market and application requirements. By applying our modular design philosophy and using pre-designed options, we can customize your generator the way you need it with minimum lead time.
There are more than 1,000 air-cooled generators in operation. Our design, engineering, sales, project management, manufacturing and testing experts work closely together in one location in Erfurt, Germany. By using the latest manufacturing technology and process automation we ensure outstanding quality.
Our generators are designed using our building-block system with optimized diameter and length sizing to achieve perfect alignment with your plant and the grid requirements. Grid studies can be performed. Many more options are available including variants of the cooling system and the position of the generator leads. All parts and connections are readily accessible and the enclosure can be easily removed for installation or maintenance.
Our modular design philosophy, using pre-designed options, ensures that we can customize your generator the way you need it. The generator matches your turbine's performance and provides margin for operational flexibility.
Our proven installation and maintenance concept sets a benchmark in reliability and availability. Professional installation of the generators ensures world-class availability and reliability for the entire lifecycle of the generator. For minimized downtime, we align the maintenance intervals of the generator with those of the turbine.
Oil and gas power applications, especially offshore applications, require low-weight generators with explosion protection solutions (e.g. pressurized or non-sparking) and entail stringent electrical standards. We can supply solutions for a broad variety of applications.
The existing foundation and electrical connections as well as the axial shaft center height can be re-used. When an aging water- or hydrogen-cooled generator is replaced, existing high-maintenance auxiliaries will be reduced. Operating and maintenance costs can be reduced and the potential service life will be extended.
A synchronous condenser solution comprises a generator connected to the high voltage network through a transformer. It is started by either an electric motor or a static frequency converter. Once synchronized with the network, the generator behaves like a synchronous motor without load, compensating reactive power and stabilizing the system, in case of a fault or imbalance events.
You can benefit from our fleet experience using natural sources for energy production. Adapted for steam turbines, our generators are used for biomass, waste-to-energy, concentrated solar power (CSP) and geothermal power plants.
Combusting biomass feedstock to create electricity does not contribute to global warming and helps to avoid the release of other harmful emissions. Also, viable fuels are produced in many industries with no additional processing costs involved, further improving the financial feasibility of new plants on favorable sites.
The hyperparameters contain: bath size (4 images for CFD, 1 image for AigleRN), optimizer (adam), learning rate (0.001), min-learning rate (0.000001), learning rate scheduler (plateau), patience (10), factor (0.95) with two functions (torch.optim.lr_scheduler.ReduceLROnPlateau and torch.optim.Adam based on Pytorch library [75]). These parameters are intrinsic parameter during training the neural network, such as learning rate. When we train the CFD, 4 images are input the neural network once time; When we train the AigleRN, 1 image is input the neural network once time. This setting can enable crack detection to obtain global optimum in the segmentation performance. We fix the parameters setting for these two databases during training neural network.
Meanwhile, it is clear that FFA can detect thicker crack than our proposed, and cannot extract the crack skeleton, which can cause the low precision rate, as is shown in Table 6. The method proposed is able to extract the crack skeleton. Secondly, it is observed that the method can obtain much more number of false positive than false negative, which lead to the higher recall rate than precision rate. Then, the 2-pixel distance can also help to improve the precision rate. Finally, the average vales based on test database can improve the global precision rate.