Slot skew is applied as a method to increase the armature winding voltage waveform quality of synchronous hydro generators. Skew that matches the region of one slot pitch can effectively damp stator slot harmonics. However, achieving this condition can be difficult in some manufacturing cases, especially for the machines with greater axial length. That is why other methods are commonly used to increase the voltage waveform quality of large hydro generators. One such method is based on the damper winding slot pitch choice which ensures reduction of stator slot harmonics from the main magnetic field. Appropriate placement of damper bars over the pole shoe does not represent a significant technological problem and is much simpler to manufacture in compare with the slot skew or the fractional armature winding methods. The downside of damper slot pitch adjustment method is the damper bar currents incensement in steady state condition of the generator, which increases damper winding losses and also the rotor temperature. In order to decrease damper winding current for long term operation and enable the generator pole shoe design with smaller cross section damper bars, a combination of damper slot pitch and partial slot skew can be utilized. This paper gives insight on consequences that can occur for voltage waveform if slot skew does not not fully match the stator slot pitch and the advantages of above mentioned combined method for the design optimization of salient pole synchronous generator.

The designs of salient pole generators may differ considerably from one hydroelectric plant to another. Automatic optimization procedure is highly desirable, because the designer may have little experience with a similar machine. The presented approach defines the design space by 12 variables which have the largest influence on the goal function. The design space is constrained by a number of linear and nonlinear constraints. The optimization process is based on successive linearizations of the goal function and the nonlinear constraints followed by a simplex procedure. The process is highly effective because the goal function is heavily constrained, so the optimum is virtually always on the boundary of the feasibility region. The procedure has been tested on a number of earlier designs. The goal function could have been reduced on average by some 8 percent, had this software been available at the time of the design of these machines.