Cold Start of a 240-MVA GSU Transformer Filled With Natural Ester
Transcription
Cold Start of a 240-MVA GSU Transformer Filled With Natural Ester
256 IEEE TRANSACTIONS ON POWER DELIVERY, VOL. 30, NO. 1, FEBRUARY 2015 Cold Start of a 240-MVA Generator Step-Up Transformer Filled With Natural Ester Fluid Steven P. Moore, Senior Member, IEEE, William Wangard, Kevin J. Rapp, Member, IEEE, Deanna L. Woods, Member, IEEE, and Robert M. Del Vecchio, Member, IEEE Abstract—A 240-MVA, 165-GRDY/20-kV generator step-up transformer was insulated with Envirotemp FR3 fluid. The transformer was energized during a cold temperature period in mid-December 2010. This is believed to be the largest generator step-up transformer filled with a natural ester fluid that was energized at low temperatures. The lack of cold start experience of natural ester-filled power transformers created questions and at least some uncertainty as to the outcome. This paper describes the instrumentation used and data collected during the cold startup procedure. In addition, a computational fluid dynamics analysis of this transformer design was performed that provided calculated temperatures and elapsed times until steady-state temperatures were achieved. Comparisons were made between the actual and calculated values. The results indicate that the cold startup procedures for a power transformer are the same whether filled with mineral oil or natural ester fluid of similar characteristics as used for this paper. Index Terms—Cold temperature startup procedures, generator step-up transformer, hottest-spot winding temperature, natural ester dielectric fluid, top-oil temperature. I. INTRODUCTION T RANSFORMERS containing natural ester fluids are growing worldwide. The advantages of natural ester fluid in transformer technology are measured by significant improvements in fire safety, environmental performance, and sustainability [1]. However, the performance of the solid-liquid insulation at operating temperatures encountered during extremes in weather needs to be understood. Natural ester fluid technology has evolved to include power class transformers with ratings as high as 400 kV and 240 MVA, with some units operating since 2004. This paper documents the cold temperature startup of the new 240-MVA (GSU) transformer near Burlington, IA in mid-December 2010. The idle transformer Manuscript received January 10, 2014; revised May 01, 2014; accepted June 06, 2014. Date of publication July 02, 2014; date of current version January 21, 2015. Paper no. TPWRD-00030-2014. S. P. Moore is with the SPX Transformer Solutions, Waukesha, WI 53186 USA (e-mail: [email protected]). W. Wangard is with the Engrana LLC, Evanston, IL 60201 USA (e-mail: [email protected]). K. J. Rapp is with the Cargill Industrial Specialties, Brookfield, WI 53005 USA (e-mail: [email protected]). D. L. Woods is with the Alliant Energy, Muscatine, IA 52761 USA (e-mail: [email protected]). R. M. Del Vecchio is with SPX Transformer Solutions, Inc., Fremont, CA 94538 USA (e-mail: [email protected]). Color versions of one or more of the figures in this paper are available online at http://ieeexplore.ieee.org. Digital Object Identifier 10.1109/TPWRD.2014.2330514 and surrounding outdoor temperatures at the startup event were near the pour point of the fluid. High viscosity of fluids near their pour points causes reduced fluid flow, resulting in less heat transport. New transformers with natural ester fluid, especially the GSU type, were generally started during warm seasons or periods of warmer temperatures. Thus, no cold start experience and data were gathered that related to the actual temperatures inside a large power transformer. The main question is does the cellulose winding insulation in a fully loaded transformer with natural ester fluid at or near its pour point overheat? The insulation in power transformers is typically a combination of liquid and solid dielectric materials. The dielectric strength of natural ester fluid in comparison with other types of insulating fluids has been measured to 50 C and found to be equal or better. Previously published work showed that a single-phase 167-kVA distribution transformer, energized at full-rated load after equilibration at 30 C in a cold temperature chamber, maintained temperatures well within acceptance standards [2]. For the distribution transformer study, the insulation materials aged within the industry-accepted rules and the full unit life expected was maintained. The three-phase GSU transformer under the cold start conditions in this field study provided important data that were compared to output from computational fluid dynamics (CFD) analysis, using design data in a model of the transformer. Agreement between the temperature of the field GSU at full expected load and the CFD provided confidence in the accuracy of the model. II. PROPERTIES OF NATURAL ESTER FLUIDS AT COLD TEMPERATURES Natural ester insulating fluids for transformers are made from vegetable oils which are a combination of different triglyceride compounds. Each of the roughly eight major compounds has unique physical characteristics, such as melting and crystallization or freezing point temperatures. As the temperature of the fluid decreases, the individual compounds eventually reach their freezing point and crystallize. The freezing, gelling, crystallizing, and melting of natural ester fluids are processes that take place over an extended time period and temperature range. As the solid crystals form, the fluid becomes increasingly opaque or cloudy. The solid crystals build as more individual natural ester compounds in the combined mixture crystallize. The increasing numbers of crystals slowly increase the viscosity of the fluid because, as the crystals grow, they form larger networks until they impede the flow so much that the pour point of the fluid is 0885-8977 © 2014 IEEE. Personal use is permitted, but republication/redistribution requires IEEE permission. See http://www.ieee.org/publications_standards/publications/rights/index.html for more information. MOORE et al.: COLD START OF A 240-MVA GENERATOR STEP-UP TRANSFORMER FILLED WITH NATURAL ESTER FLUID reached. The fluid pour point is defined as 3 C above the temperature of fluid in a tube tilted at 90 where the fluid does not flow within 5 s of time [3]. The natural ester fluid used in the 240-MVA GSU transformer has a 21 C pour point. The fluid in the cooling ducts would only cease to flow if held at 21 C for greater than 72 h or in less time if the fluid gets colder. Thus, the fluid in this temperature range is in an opaque, highly viscous condition, but still flows slowly to transport heat away from the interior of the transformer. The transformer loses heat from the interior outward to the tank wall, so the fluid at the tank walls cools first. As the fluid cools, it slowly crystallizes and becomes more viscous and acts like a thermal insulator which helps to further reduce heat loss through the fluid layer at the tank wall. This maintains the interior temperature of a transformer. 257 TABLE I COMPARISON OF THE PROPERTIES OF THE NATURAL ESTER FLUID USED AND CONVENTIONAL TRANSFORMER MINERAL OIL. THE PROPERTIES ARE 25 C UNLESS SPECIFIED TABLE II STEADY-STATE TEMPERATURES IN C AT 200 MVA III. GSU TRANSFORMER DESIGN AND INITIAL STEADY-STATE COOLING ANALYSIS The GSU transformer has a three-phase ONAN rating of 145 MVA and an ONAF rating of 240 MVA. It has inner and outer HV windings connected in series and an LV winding sandwiched between them. Cooling of the windings is accomplished by means of directed oil flow using oil-flow washers. The flow is thermally driven by differences in oil buoyancy at different temperatures. The external cooling is by means of radiators with or without fans. In addition, there is some cooling through the tank walls. The cooling design for this transformer was based on an in-house cooling simulation program. This was used previously in a comparison study of several transformers where mineral oil and natural ester fluid were used in the same transformer [4]. In that study, the cooling program results were compared with test data and temperature differences between the two coolants were reasonably well predicted. The feature of the cooling program which makes this possible is that it uses the thermal parameters of the coolant as an input, including their temperature dependences. The parameter which differs most significantly between the two fluids is the viscosity. The model is based on fluid flow through ducts. The ducts involved are ducts in the coils through which the coolant is directed and ducts in the radiators. The bulk oil mixing in the transformer tank is treated by assuming a linear temperature distribution between the cold bottom oil out of the radiators to the hotter top oil out of the windings. Thus, the coolant properties of interest are the viscosity, density, thermal expansion coefficient, and thermal conductivity as well as their temperature dependences. These properties, in turn, affect the fluid friction through the ducts and the surface heat-transfer coefficient from the winding surfaces to the coolant. Fluid friction and the surface heat-transfer coefficient are determined by established correlations through unitless numbers, such as the Reynolds, Nusselt, and Prandtl [5]. The heat is generated in the windings by losses and eddy current losses. There are also eddy current losses in the structural components, such as the tank and winding clamping structure which must be considered. All of these losses as well as the coolant properties are temperature dependent. The heat is dissipated through the tank walls and radiators to the air, also using established correlations for the heat-transfer coefficients. Thus, an iterative solution method is used to arrive at a steady-state solution where the input losses are equal to the dissipated losses for a given power input to the transformer. The temperatures at this steady state in the windings and coolant can then be printed out. Thermal parameters of the natural ester fluid and mineral oil are given in Table I for comparison purposes. The natural ester fluid parameters were obtained from the manufacturer [6], [11], [12], and the mineral oil parameters are taken from [7]. The thermal expansion coefficient affects the density as a function of temperature T via (1) where is the density at temperature . This temperature dependence is important for buoyancy driven flow. The thermal conductivity and specific heat are fairly constant in the temperature range of interest. However, the temperature dependence of the viscosity is quite pronounced. For the natural ester fluid, the viscosity data were fit to (2) where is in and T is in . (2) At 10 C, the viscosity of the natural ester fluid is 4.6 times higher than that of mineral oil and is almost that much higher at more normal temperatures. The cooling program was run at the ONAN and ONAF ratings to check the adequacy of the cooling system at these ratings and to check that the guarantees would be met. However, for the cold start test conducted in this paper, the transformer was run under OA conditions (without fans) at a three-phase megavolt amperes of 200 and an ambient temperature of 3.0 C. The temperatures of interest are given in Table II. These steady-state results will be compared with the steadystate results achieved by the more detailed finite-element fluid flow analysis and with the test results described later. 258 IEEE TRANSACTIONS ON POWER DELIVERY, VOL. 30, NO. 1, FEBRUARY 2015 IV. COMPUTATIONAL FLUID DYNAMICS AND THERMAL ANALYSIS METHODOLOGY A. Overview A fluid dynamics study is conducted for the cold startup at a 200-MVA load for the transformer containing a natural ester coolant. The convective heat transfer and fluid motion of the coolant in the transformer is modeled using a commercial computational fluid dynamics (CFD) software tool. The geometry of the system is simplified to an approximate 2-D-axisymmetric domain, including detailed flow through windings and the external heat exchanger. The natural ester coolant fluid is modeled as a Newtonian fluid. The thermal energy sources in the transformer arise from and eddy current losses in the windings, core losses, and tank/ clamp losses. The sources are obtained from an electromagnetic (EM) model of the windings for the transformer operating at a reference condition. In the physical transformer operating at a power level different from the reference condition, scaling laws are used to compute the instantaneous thermal sources. Fig. 1. CFD model of the transformer. B. System Geometry and Computational Domain The transformer core and windings are confined inside a rectangular tank to which external heat exchanger units are attached. The heat exchangers or radiators consist of an array of plates that span vertically between inlet and outlet header pipes that emerge from the side of the tank. The fluid flow is driven by buoyancy effects and is not pumped. The heat exchanger has external fans to increase heat transfer to the environment but they are not used during startup. Therefore, the external heat transfer to the environment is driven by free convection, conduction, and, to some degree, by radiation. The external tank is made of mild steel and is 10 mm thick. The transformer has three legs, each with the same winding pattern. It is assumed that heat generation from each winding is identical. Thus, instead of modeling all three legs, only one leg is modeled, and the system volume is equal to one-third of the actual tank volume. The three-phase core is made of high permeability laminations. The windings are made of paper-covered copper wire. The winding disks are separated by pressboard spacers to enable horizontal oil flow for cooling. This flow is directed by suitably spaced washers to force a serpentine flow pattern. Between the coils are cylindrical pressboard barriers, which isolate flow between the various windings. A representative image of the cross-section of one of the transformer legs is shown in Fig. 1. This figure shows the external tank, the heat exchanger, yoke, core, press plates, and windings. It does not show the flow collars or winding barriers. C. Physical Heat Exchanger The physical heat exchanger (PHE) consists of a set of parallel plates all of the same length, fed by an inlet header which exits the top of the transformer tank. The plates are parallel and face each other with a gap separation that is fixed. Heat is transferred to the environment by air movement between the plates. For free convection, the flow is primarily vertical in nature. Radiative heat transfer is negligible. Due to high conductivity of TABLE III PARAMETERS OF THE PHYSICAL HEAT EXCHANGER the radiator plates and tank, conductive heat flow can be ignored. The heat exchanger consists of plates of dimensions , , and , where is the length of the plate between inlet and outlet headers, is the plate width, and is the hydraulic diameter of the cross section. Each plate has a lateral total area given by . The factor of two accounts for both plate sides. The dimensions are listed in Table III. D. Computational Heat Exchanger Our main problem is that we have to ensure that the 2-Daxisymmetric heat exchanger matches the flow characteristics of the physical experiment. The computational heat exchanger (CHE), shown in Fig. 2, is a series of concentric cylindrical annular channels with length having a uniform gap width and radial separation . To match the flow characteristics of the PHE to the CHE, the CHE must have matching hydraulic characteristics. is not equal to . must be determined so that the total surface area of the CHE matches the PHE. The length of the CHE plates is set equal to the length of the plates in the PHE, to match the pressure drop for a given length. Thus, . Similarly, for the friction factor to match, the hydraulic diameter of the model must match the physical experiment. Thus, . The gap between the model plates is . The value of must be greater than the plate thickness: in order to avoid meshing problems. MOORE et al.: COLD START OF A 240-MVA GENERATOR STEP-UP TRANSFORMER FILLED WITH NATURAL ESTER FLUID TABLE IV HEAT-SOURCE DATA. 259 145 MW has (radial) length and axial height . In the experiment, the radius of the yoke is . Thus, it is can be shown that an equivalent porosity of the model zone is (5) Using table values, the porosity of the yoke zone is Fig. 2. Model of the CHE. The total area of the CHE is calculated by summing up the exposed area of a set of circular annuli of incrementally increasing radii. For a set of plates, the total surface area is given by (3) where is the inner radius of the innermost channel and is yet unknown, but will be determined shortly. The total area of the PHE is used to transfer heat from all legs of the transformer. Since our CHE only models one leg, set . Substituting into (3) yields an equation in terms of and . To avoid meshing conflicts, impose the solution , where is the radius of the model tank. The resulting quadratic equation has the solution (4) brackets indicate that the solution is rounded where the down to the nearest integer. The model tank radius is calculated based on the total volume of the tank. From these values and the expressions developed in the previous section, . Solving for the number of channels and the innermost radius of the first channel yields 9 and 1.4639 m. E. Core Yoke The CFD model of the yoke must be made porous to accommodate axial flow of the coolant while simultaneously being able to capture thermal transients. The porosity of the CFD model is calculated so that the total mass of the solid phase of the yoke zone matches the experiment. In the model, the yoke 0.832. F. Model Equations The equations are modeled using a finite volume formulation using the commercial software ANSYS Fluent. The equations are solved using a second-order transient formulation that includes the effects of buoyancy using the Boussineq approximation. See [8] for details of the methods and equations. In the solid zones, there is only pure convection and heat generation. The heat generation terms are calculated from the EM losses in the windings and the core. For each component, the thermal source has been calculated for a reference operating condition. The formula for the thermal source (in watts) is given by (6) is the transformer where P is the transformer power and power at the reference operating condition. The load and “no load” subscripts refer to the components of power at the reference conditions. Table IV gives the load and “no load” thermal sources and zonal volumes for each region. The heat generation term in the model is the thermal source divided by the volume of the respective zone, and is assumed to be spatially uniform. G. Fluid Properties The density of the natural ester fluid is given by the linear relationship (7) where Equation (7) is a linearized version of (1), using slightly different notation and parameter values. 260 IEEE TRANSACTIONS ON POWER DELIVERY, VOL. 30, NO. 1, FEBRUARY 2015 TABLE V EFFECTIVE COPPER DENSITIES TABLE VI THERMAL PHYSICAL PROPERTIES OF THE COMPONENTS Fig. 3. Average temperature near Burlington, IA—NOAA Weather 2010. K. Computational Mesh The mesh consists of 310 421 quadrilateral cells. The collars and washers are modeled using impermeable walls. Nonconformal meshes are created between the winding zones to maintain mesh quality. The outer tank thickness is modeled as a solid zone. The viscosity of the natural ester fluid is given by (2). The heat capacity of the natural ester fluid is given by (8) where the temperature is in Kelvin. H. Windings and Other Solid Materials In the CFD model, the volume occupied by the copper windings includes the paper wrapping, and its effect on the total mass of the modeled zone cannot be neglected since the mass of the winding directly affects the thermal response. For each winding, we calculate the effective solid density as the mass of the copper divided by the total volume of the zone. Table V lists the values for the winding zones. The thermal conductivity of copper is , and its heat capacity is . Thermophysical properties of the other solid materials are shown in Table VI. L. Solution Methodology The CFD model is performed using ANSYS Fluent v14.0, which utilizes a finite volume approach to the solution of the conservation laws [9]. Time stepping is performed using a second-order algorithm. The convective discretization terms are second order for momentum and energy. The pressure discretization utilizes the body-force weighted method, and the pressure-velocity coupling uses the PISO method. The fluid is modeled as a Boussinesq fluid with the parameters indicated in the previous section. Time steps were limited, based upon stability, to about 0.25 s and the solution was run for about four days. Thus, a total of about 350 000 time steps were modeled. Each time step converged within 5–10 iterations. The default convergence criteria were used. V. COLD STARTUP PROCEDURE AND TIMELINE A. Weather Station Data for Burlington, IA, USA I. Boundary Conditions The bottom of the tank rests on a concrete pad. It is assumed that heat transfer to the ground is negligible. The heat transfer from the sides, top, and heat exchanger surfaces is a convective boundary condition. The local heat flux normal to the boundary is given by The outdoor temperature in December 2010 as measured at the NOAA weather station near Burlington, IA is shown in Fig. 3. The installed transformer cold startup was performed on December 13, 2010, which was the coldest average temperature for the entire month. Alliant Energy measured temperatures and transformer load during the installation and startup period as summarized in Table VII. (9) where the external convection coefficient is given by . The ambient temperature varies with time and its value is given by local meteorological conditions. J. Initial Conditions When the transformer is energized, the fluid velocity is zero and the initial temperature is equal to the ambient temperature. B. Transient Load and Experimental Temperature Rise of Transformer Components On December 12, 2010, the GSU at the Burlington Generating Station was energized after being assembled and vacuum filled about two weeks before. The unit was allowed to soak for about 17 h until 6:00 A.M. on December 13, 2010. The criteria was that the unit should soak until the top-oil temperature reached 5 C at which time, the load was ramped up slowly in MOORE et al.: COLD START OF A 240-MVA GENERATOR STEP-UP TRANSFORMER FILLED WITH NATURAL ESTER FLUID 261 TABLE VII TIMELINE AND TEMPERATURE READINGS Fig. 6. CFD results from the cold start case. Gross transformer load, ambient temperature, and calculated results of maximum component temperatures. VI. COMPUTATIONAL RESULTS COMPARED WITH TEST DATA A. Cold Startup Case Fig. 4. Transformer at its site. Fig. 5. Sample thermographic image. 30- to 50-MW increments until reaching the desired capacity of 200 MW. Had this GSU been filled with mineral oil, the startup procedure would have been the same except for the 5 C criteria. This additional criterion was added as a unit of this size filled with natural ester fluid had not been energized from a cold state before. Thermographic images were taken of the GSU during the startup procedure. Temperatures for the tank wall and outermost radiator panels were recorded and are shown in Table VII. A picture of the transformer is shown in Fig. 4 along with a sample thermographic image in Fig. 5. The CFD simulation was run for about four days. Time zero corresponds to core energization. Load power was applied at 0.693 days at which time the ambient temperature was 13.8 C. This was the start of the computational analysis. The ambient, maximum winding, and top-oil temperatures are shown in Fig. 6, along with the gross load. Steady state at 200 MVA was reached after 3.5 days. The low-voltage (LV) winding had the highest peak temperature of 80 C. The next highest of 75 C occurred in the outer high-voltage winding (HVO). The inner high-voltage winding (HVI) had a peak temperature of 60 C. The steady-state winding model predicted a peak temperature of 83.2 C for the LV winding, 76 C for the HVO winding, and 65.5 C for the HVI winding. In Fig. 7, the temperature of the top-oil sensor is compared to the CFD model. During the initial temperature rise, prior to the onset of circulating flow through the heat exchanger, the errors in the calculation are about 10 C. However, the onset of steady flow through the heat exchanger stabilizes the flow. After about 1.5 days, the error is a maximum of 4 C and drops to about 1 C at the end of three days. The top oil reaches 57 C in three days. The steady-state top-oil calculation is 61.6 C. The GSU described in this paper is equipped with four fiberoptic (FO) temperature sensors. Two are embedded in the LV winding and two in the HVO winding at the estimated hotspot locations. Unfortunately, data from these probes were not recorded at the time of the cold startup. Normally, the transformer is monitored hourly for top-oil temperature, and a simulated winding hot-spot temperature via a winding temperature indicator (WTI) was calibrated to indicate the LV hotspot in this case, and fiber-optic (FO) probes if available. Ideally, the CFD results should be compared with the FO data. However, the WTI data can serve as a substitute with some caveats: “It is shown that in case of rapid load change, the methods used by classic WTI can indicate a lower temperature by more than even if they are properly adjusted for steady state conditions.” [10]. 262 Fig. 7. Comparison of temperatures predicted at the top-oil location by the CFD simulation and the recorded measured temperature. Fig. 8. Average of the two LV winding fiber-optic probe temperatures minus the WTI temperature versus the rate of change of the WTI temperature. In April 2013, the GSU was de-energized for three weeks and then re-energized with a sudden load. For this run, data from the FO and WTI probes were recorded. The LV FO probe data minus the WTI data are plotted versus the rate of change of the WTI temperature in Fig. 8. This plot supports the observations made by [10]. It shows that the temperature of the WTI can be as much as 9 C below the FO temperature for the most rapid temperature change. This plot can be used to correct the WTI temperatures to indicate the true hot-spot temperatures when the temperatures are changing rapidly. The CFD model results for the LV hot-spot temperature are plotted along with the WTI temperature versus time in Fig. 9. The temperatures in Fig. 9 track each other fairly well through the dips and peaks of the power input but are a bit off as steady state approaches. This could be an artifact of the WTI indicator as mentioned previously, especially since the upper branch of the curve in Fig. 8 shows a discrepancy of about C as steady state is approached on the left. If we add 10 C to the WTI information shown in Fig. 9, the maximum winding temperature is still only 45 C for a fluctuating load and only 78 C for a steady-state load. So for this worst case assumption, temperatures are still well below the maximum nameplate operating temperature. Thus, the cold startup strategy currently employed is sufficient to avoid overheating the cellulose insulation and IEEE TRANSACTIONS ON POWER DELIVERY, VOL. 30, NO. 1, FEBRUARY 2015 Fig. 9. LV winding hot-spot temperature comparison of the CFD model calculation and WTI data. Fig. 10. CFD results from the very cold, sudden start case. Gross transformer load, ambient temperature, and calculated results of maximum temperatures for various components. natural ester fluid for the 105 C insulation system used per the standard [13]. During temperature rise tests at the factory, the calculated hottest spot rise for the HV winding was 65.6 C and the measured hottest spot temperature rise using fiber optics was 64.4 C. For the LV winding, the calculated value was 69.6 C and the measured one was 70.9 C. B. Very Cold Startup/Sudden Power CFD Calculation In this case, the initial transformer and ambient temperatures are fixed at 25 C. At time zero, the transformer power is raised suddenly to 200 MVA. Initially, the fluid viscosity is quite high and it is thought that this may inhibit the formation of natural convection currents to cool the transformer windings. Results of the simulations are shown in Figs. 10 and 11. In Fig. 10, the winding and top-oil temperatures and gross load are plotted versus time. The winding temperatures ramp up almost immediately and overshoot the steady-state temperatures. It is not until a significant amount of heating has occurred that there is enough density gradient in the fluid to push fluid through the heat exchanger and start the thermo-siphon. This occurs at about 6 h as shown in Fig. 11. MOORE et al.: COLD START OF A 240-MVA GENERATOR STEP-UP TRANSFORMER FILLED WITH NATURAL ESTER FLUID 263 [13] IEEE Standard for the Design, Testing, and Application of Liquid-Immersed Distribution, Power, and Regulating Transformers Using HighTemperature Insulation Systems and Operating at Elevated Temperatures, IEEE Standard C57.154–2012, 2012. Steve P. Moore (SM’90) received the B.S. Engineering Technology-Electrical degree from the Milwaukee School of Engineering, Milwaukee, WI, USA. He has been with SPX Transformer Solutions, Waukesha, WI, USA, and all of its previous company names. Much of his 38-year career was spent on marketing and sales of new power transformers, turn key projects, and field service. His current activities are in the areas of load-tap changers and the application of natural ester fluid in medium and large power transformers. Fig. 11. Heat exchanger mass flow rate for the very cold, sudden start case. The thermosiphon flow has become fully developed and reached steady state after 8 h. VII. CONCLUSION In conclusion, we found that the same cold startup procedures that are used for power transformers filled with mineral oil can be used when they are filled with natural ester fluid with similar physical characteristics, particularly viscosity, to the fluid used in this transformer. There were no dielectric issues observed during energization and startup. The calculated temperatures from the CFD model were close to the measured temperatures. Even if the load was ramped up much faster than normal, the hottest spot temperature is not expected to go above the 105 C rating of the insulation system. REFERENCES [1] D. Bingenheimer, L. Franchini, E. Del Fiacco, J. Mak, V. Vasconcellos, and K. Rapp, “Sustainable electrical energy using natural ester technology,” presented at the CIRED 21st Int. Conf. Elect. Distrib., Frankfurt, Germany, Jun. 6–9, 2011. [2] K. J. Rapp, G. A. Gauger, and J. Luksich, “Behavior of ester dielectric fluids near the pour point,” presented at the IEEE Conf. Elect. Insul. Dielectr. Phenomena, Austin, TX, USA, Oct. 17–20, 1999. [3] Standard Test Method for Pour Point of Petroleum Products, PA 5.01, ASTM Standard D97, ASTM International, West Conshohocken, PA, USA, 2012. [4] R. M. Del Vecchio and R. Ahuja, “Comparison of the thermal performance of FR3, a natural ester based coolant, with transformer oil,” in Proc. CIGRE, Brugge, Belgium, 2007. [5] R. M. Del Vecchio and P. Feghali, “Thermal model of a disk coil with directed oil flow,” in Proc. IEEE Transm. Distrib. Conf., New Orleans, LA, USA, Apr. 11–16, 1999, pp. 914–919. [6] K. J. Rapp and J. Luksich, “Transformer design and natural ester fluids,” in Proc. 12th Int. Elect. Insulation Conf., Birmingham, U.K., May 29–31, 2013, pp. 48–52. [7] F. Kreith and W. Z. Black, Basic Heat Transfer. New York, NY, USA: Harper & Row, 1980. [8] F. White, Fluid Mechanics. New York: McGraw-Hill, 1986. [9] “ Fluent 14.0 User Manual,” ANSYS, Inc., 2013. [10] J. N. Berbrue, B. L. Broweleit, and J. Aubin, “Optimum transformer cooling control with fiber optic temperature sensors,” 2007. [Online]. Available: www.neoptix.com/literature/v1165r04_Art_Optimum_Cooling.pdf [11] Cargill, Cargill Dielectric Fluids reference R2120, Cold Start Recommendations for Envirotemp™ FR3™ Fluid Filled Transformers, May 2013. [Online]. Available: www.environmentalfluids.com [12] Cargill, Cargill Envirotemp™ FR3™ Fluid reference R2020, Envirotemp FR3 Fluid Behavior in Cold Temperature Environments, May 2013. [Online]. Available: www.environmentalfluids.com William Wangard received the Ph.D. degree in mechanical engineering from Colorado State University, Fort Collins, CO, in 2001 He was a Senior Consulting Engineer with Ansys, Inc. (formerly Fluent, Inc.), Evanston, IL, USA. He founded Engrana LLC, Evanston, IL, USA, in 2009. Engrana provides thermal and fluid-flow simulation consulting services for clients spanning many industrial sectors. Kevin J. Rapp (M’10) received the B.S. degree in chemistry from the University of Wisconsin, Parkside, WI, USA, IN 2003 after completing undergraduate research in lipid-cellulose interactions chemistry. He began his career at the Thomas A. Edison Technical Center, Cooper Power Systems, Franksville, WI, USA, in 1976. In 2012, the Envirotemp FR3 Fluid, which Kevin co-invented, became part of a new dielectric fluids business unit at Cargill, Brookfield, WI, USA, where he is currently Senior Chemist. He is involved in standards work at ASTM, West Conshohocken, PA, USA, as Chairman of the D27.15 and D27.91 subcommittees, and is the Technical Advisor for the U.S. National Committee, and for IEC TC10 regarding insulating fluids for electrotechnical applications. He holds many U.S. and international patents and has published many papers. Deanna L. Woods (M’11) is a Substation Manager of Construction, Maintenance, and Engineering for Illinois Power, Decatur, IL, USA. Currently, she is a Senior Engineer with Alliant Energy, Madison, WI, USA, responsible for the substation predictive maintenance program and mentoring substation engineers. Robert M. Del Vecchio (M’78) received the Ph.D. degree in physics and the M.S. degree in electrical engineering from the University of Pittsburgh, Pittsburgh, PA, USA, in 1972 and 1978, respectively . After serving in various academic positions, he then joined the Westinghouse R&D Center, Pittsburgh, in 1978, where he worked on modeling magnetic materials and electrical devices. He joined North American Transformer (now SPX Transformer Solutions), Milpitas, CA, USA, in 1989, where he developed computer models and transformer design tools. Currently, he is a Consultant.