Repowering and evaluation of new power of synchronous generators

Useful life of winding insulation is about 30 years. This may be reduced when it is overloaded or when it works in aggressive environments. When retrofitting is undertaken an increase to a higher insulation class is recommendable. So the generator's capacity should be increased, and this will not just more than fully compensate for the investment made it will also result in a more efficient use of the raw materials used and thus contribute to sustainable development. Our Brazilian experience shows that retrofitting with repowering is successful. The objective of this study is to present two cases of repowering, in which the old insulating materials were replaced by other, modern ones. So, eight SG of a Power Plant in Cubatão, S.P and two SG of CEMIG, MG had its power increased up to 40%. The discussion also deals with how to determine the new power obtained after the repowering, before the final load tests. In large machines, the load tests are critical. In order to calculate the field current of the full load SG to determine the temperature rise in field, indirect tests are undertaken. Which are the methods that present reliable results? This study compares three study methods, ASA, IEEE 115, and the General Method with their theoretical analyses. The ASA method is now regarded as normal in many countries. According to Brazilian Standards it is given under NBR-5052. In the IEEE 115, this method is named “Phasor Diagram Analysis—Salient-Pole Machines”. The General Method is an academic treatment with a better theoretical basis.

x l -Leakage reactance r -Stator resistance V f -Field voltage R F -Field resistance F A -Armature Magnetomotive force

1-Introduction
This study is based on two CEMIG SG repowered in 2007 according to Table 1. It is also based on eight 20-pole, 11-kV, EMAE SG, which were repowered from 1995 until 2007 to obtain new power with an increase of up to 40%.

2-Methods used to confirm new power
After repowering, the result must be confirmed with final tests with nominal load so as to guarantee the new projected values.
In large machines, the test with nominal load implies putting it in parallel with an electrical network system. Before the test, it is to be recommended that the new parameters be guaranteed as well so as to confirm the planned range of temperature. This is possible with normalized tests but not with loaded ones. The purpose of this study is to contribute to a discussion on better ways to determine the correct field current with no load tests. Methods are to be applied under steady state conditions, everything being modeled in frequency dominion.

2.2-Rotor
The determination of the rise in the temperature of the rotor may be calculated as follows: a) The field should be fed with a current equal to that necessary to keep the generator in a condition of rated load. This can be done with the SG excited on no load or in short circuit, but either of the two tests could cause damage to the generator. b) We can undertake extrapolation using the data obtained with no load or short circuit tests in order to obtain the rise of temperature at full load. The problem consists, therefore, in accurately determining the value of the full load excitation current. IEEE 115 presents two methods for determining load a In machines with water-cooled armature only the rise of the short circuit test must be considered, because the coolant practically eliminates the iron heating influence in the winding.
field current. ABNT-NBR5052 presents several methods, with emphasis on the ASA method. Finally, a discussion of the General Method which is understood as theoretically the correct one. The next item presents a discussion of the three above-mentioned methods.

2.2.1-EC Determination by Phasor Analysis Method for Salient Poles
This method corresponds to the reactance or flux superposition method. Magnetic and electrical parameters are decomposed into direct and quadrature axes. This description is similar to that given by IEEE-115 item 5.3.3: Suppose SG to have V, I and cos f. Current is to be decomposed into the d axis (I d) , and q axis (I q) . I forms an angle d+f with the q axis, and can be calculated as: Arc tang (d+f)= (Vsenf+Ix q ) /Vcosf (1) δ: the power angle (between E f and V): Thus, I q and I d current can be calculated: The no-load voltage to maintain V at the SG terminals can be calculated by: Error correction due to saturation: E fns is determined on the air gap linear curve, thus MC is supposed to be Linear. It is possible to obtain F fns and E fns from figure 2. The correction of the saturation effect is obtained by the voltage corresponding to resultant flux F R. This voltage, named E R, is determined as presented in (5): The correction is the difference between the air gap and the saturation curve in Figure 2. DF f +F fns =Ffs is the mmf field necessary to maintain the alternator under specified conditions. Ef no saturated is obtained from the saturation curve. x l is calculated from the project or experimentally determined for example by the Poitier Method which presents some errors but is acceptable for almost all Norms. This method presents reliable results, however it does not correspond to the physical reality since neither machine works on the air gap line nor does the correction obtained by the sum of DF f correspond to the physical reality. Another way to consider the saturation effect is to suppose linear MC in E R . Correction is made with K=E' R /E R. Figure 3: Linear MC at E R x d(s) at point E R is calculated by dividing x d(ns) by K, excluding x l . The reactance x q is not corrected because the larger part of quadrature flux passes through the air. If E' fns is the voltage in the (fictitious) Linear New Magnetic Circuit, expression (6) could be written like expression (4) but with x ds in the place of x dns E' fns was determined by mmf F f , which can be found on the abscissa axis from a perpendicular at E' fns. The E f real value can be found on the saturation curve. The analysis of figure 3 confirms the above explanation.

The Graph Method using Poitier (corresponding to the ASA Method)
IEEE 115 describes this method but limits it to SG nonsalient poles. ABNT NBR 5052 describes it as the ASA Method but not as limited to non-salient poles. This method adds the corresponding mmf value to the nominal voltage in the air gap line (F f0ns ) with mmf to keep the SG in short circuit with the nominal current (F A +F x ). If F f0ns is the reference, the angle of (F A +F x ) is (90+f). In order to obtain F f , DF f value must add in (F f0ns + (F A +F x )). That value is the difference between mmf corresponding to E R on the saturation curve and the air gap line. The E R value is calculated from the graph as shown in Figure 4. The values obtained by this method are reliable. However, there is no theoretical basis for the method because it neither ns Ef corresponds to physical reality nor to any mathematical model.

-General Method a-Non Salient Poles
In this method, Ff and F A act in the same MC; the sum of both is the resulting F R ; thus E R and V may be found. F A is calculated by the design or determined experimentally by the Poitier method.
To calculate the necessary F f to maintain SG in nominal conditions, the first step is to obtain F R. So it is necessary to know E R , which can be found from equation (5) or graphically as showed in Figure 5. F f is obtained from the expression F f allows E f to be found from a no load voltage curve.

b-General Method: Salient Poles
With a salient SG pole it is not possible to undertake the direct composition of F f and F A because they act in MCs with different permeances. Decomposition on two axes: direct and quadrature, always allows offering different but constant permeance to each if saturation is not considered.
F d is a component of F A on a direct axis, and it can be calculated as follows: and it can be added to F f because both are on the same axis. Meantime it is necessary to observe: F f has a rectangular spatial distribution, but spatial harmonics are not taken into consideration because of the pole design, which has variable value and increases from the center to the extremity, so permeance decreases, and it is possible to obtain a sinusoidal flux. But F A and consequently F d are practically sinusoidal, and when they are applied to a variable permeance the resultant flux will contain harmonics. Only the fundamental one is taken into consideration in this study. However, it is necessary to consider a correction for the Fd value, as proposed by J. A. Shouten.

Method of Calculation
To determine field current I f and corresponding E f , that makes it possible to work with nominal voltage and current as well as a determined power factor, the necessary steps are(r not considered): a) The voltage in the E R air gap can be determined either by expression (5) or graphically as presented in Figure 5. b) Value on direct axis is: It is possible to prove E R is on the quadrature axis, and F R is on the direct axis, and this value can be determined from a load curve. c) F Ad and F R (direct axis) composition permits the calculation of F f or I f . d) F f or I f on a no-load curve allows the determination of E f .

3-Results
Two successful cases are presented that justify repowering:  Case 1 emphasizes which parameters change and which do not.  Case 2 discusses three methods by which the field current in the load may be determined, and compares it with the real value. Class/Elevation (rotor) B/60º F/ 80º

-Considerations related to stator and rotor manufacture and insulation
Thick layers of cotton with asphalt were replaced by others with isolated bars including fine polyester or fiber glass layers with a material such as Nomex in the stator windings. Field winding conductors were also replaced by others isolated for F. Current density increase based on the rise in temperature and/or an increase in conductor section allowed a 31% enhancement of SG capacity. The SG changed its parameters as had been expected: all reactances increased about 30%, as presented in table 4. Time constant values behaved as expected.

-Case 2 SG n o 1 at Henry Borden Power Plant
A 33 MVA SG repowering permitted an increase of 30% of power as shown in Table 5.

-Considerations about Electrical Project
Repowered project analysis leads to interesting conclusions: a) In the stator, the conductor section was increased by 15%, from 822.31 mm 2 to 946.37mm 2 , and current density from 2.10A/mm 2 to 2.38A/mm 2 ( 13.3%). Bars were also replaced by the Robell type. Repowering was brought about by increases in the conductor section and density. The data analysis allows us to conclude that the increase in the density of the current (13.3%) is perfectly acceptable to the new temperature class. b) In the previous rotor, the conductor cross section was of 4.064x50.8=206.45mm 2 . After repowering, the conductor cross section changed to 5x50=250mm 2 representing an increase of 21% but the number of turns was reduced from 62.5 to 50. The reduction in the number of turns of field winding was necessary to limit the field voltage, because the excitation voltage must be limited to its earlier power. In order to assess the impact on the rotor it is necessary to calculate the projected current value at the new power. Tests are made with the ASA method adopted by NBR 5052. Before repowering the field current was calculated by the indirect method and confirmed by a load test, as shown in table 6. The value of the current found by the ASA method (indirect) was 398A, representing a level 3% lower than the experimental results. The load test results after repowering are shown in Table 7. In this study the theoretical methods of support discussed lead to the conclusion that the only correspondence close to physical reality is that obtained by the General Method.

-Considerations on stator rotor manufacture and insulation
The impregnation during coil manufacture has to be performed under vacuum pressure in order to guarantee the absence of air, combining synthetic epoxy-and polyesterbase resin with solid mica-base material insulation. The insulating tape is the essential part of the MICADUR technique, which consists of the application of a fine fiber glass layer that serves as a base for the mica paper. The two materials are joined with epoxy base resin. Anti-corona painting and application of low resistivity epoxy base varnish, allows a suitable basis for the insulation of the nucleus. At the extremities of the bars a high resistivity cover is used to reduce the superficial voltage gradient of the winding.

3.2.3-Considerations on the Mechanical Project
The increase in power from 33 to 43 MVA requires a corresponding mechanical increase in demand. The main concern is presented below:

-Axle
A horizontal axle unit engaged by two Pelton turbines is submitted to two combined torsion (resulting from binary pair application) and flexion (caused by the massive action of rotating parts, magnetic thrust of rotor in relation to stator and radial hydraulic thrust of turbine) efforts.