The production process of bio-insulating liquids starts with the harvesting of palm nuts and castor seeds. Crude palm kernel oil was obtained by grinding the kernels contained in the palm nut. The oil extraction was done mechanically using a hydraulic press. In the case of castor oil, after the harvest, the seeds were dried, and then crushed to extract the crude oil. Several extraction methods are possible: traditional extraction, expeller pressing, hydraulic pressing and solvent pressing. The expeller press provides the best extraction rate; however, due to its high cost, we decided to use a hydraulic press for the extraction of crude castor oil. The extracted oils were then purified through degumming followed by refining. The purpose of degumming is to remove the substances contained in the oil that may become insoluble by hydration, including phospholipids and gums. To perform degumming, demineralized water is heated and added to the oil. The oil and water are gently mixed and the degummed oil is recovered by decantation. This operation is repeated several times to extract as much gum as possible. Neutralization consists of mixing the crude oil with an alkaline solution to neutralize the free fatty acids. Sodium hydroxide (NaOH), commonly called caustic soda, is the main alkaline compound used to reduce the acidity of crude vegetable oils. Equation (1) shows the neutralization reaction. In general, the concentration of the NaOH solution increases with an increase in the phospholipid content of the oil to be treated. Careful selection of the concentration of caustic soda and the percentage of NaOH per mass of oil will achieve the desired refining results and avoid excessive oil loss excessive [25] [26] [27].

The final step of the development of the methyl esters is transesterification. It leads to the formation of alkyl esters of fatty acids and glycerol, as shown in Figure 1. The main benefit of transesterification is a reduction of the viscosity of the insulating liquid, thereby increasing its ability to dissipate the heat generated in the transformer [28]. Methanol is the most commonly used alcohol in vegetable oil transesterification reactions because it has the best conversion rates. A basic catalyst, such as potassium hydroxide, was used in this process, since it has the advantage of being less corrosive and faster than acidic homogeneous catalysts such as sulfuric acid. After transesterification, it is essential to wash the neutralized oil with demineralized water to reduce as much as possible the soap traces still present in the oil. Then, the oil was introduced into a mixture of silica gel and activated clay to remove the oil pigments, primary and secondary products of oxidation, metals and soaps by an absorption process. The next step was a vacuum filtration (filter porosity 10 µm) of the oil in order to extract activated clay and silica gel residues. Finally, the samples were dried at 80˚C for 24 hours, then removed from the oven and sealed [29] [30].

R − COOH FreeFattyAcids + NaOH CausticSoda → R − COO − Na Sodiumsoap + H 2 O water (1)

Prior to thermal aging, the moisture content of the insulating liquids and the paper dielectric were reduced, to obtain dielectric samples with moisture content in accordance with the standards [31]. To reduce the moisture content, the two insulating liquids and the dielectric paper were heated at a temperature of 80˚C for 24 hours. The characteristics of the insulating liquids after the dehumidification process are presented in Table 4. Palm kernel oil methyl ester was developed and characterized by Mengata et al. in references [32] [33]. Although the castor plant (Ricinus communis) is available in Africa, Latin America and Asia, the castor plant we used is specific to the climate and agricultural environment of Central Africa. It was developed and characterized by Tchamdjio et al. [30]. The thickness of the insulation papers used in the accelerated thermal aging tests was 0.3 mm.

At the end of the dehumidification process, the samples were placed in closed flasks and the thermal aging tests started. Samples were exposed to an accelerated aging temperature of 110˚C according to the standard IEC 60216 [17] during 96h. This duration was chosen to observe the initial period of thermal aging. The aging protocol applied is the one developed by Tenbohlen and Koch [22] [34]. The authors performed thermal aging in both closed and open flasks. In our case, we performed thermal aging in closed jars, in order to represent the real operating conditions of a transformer filled with natural ester. In the same protocol, the authors proposed to insert metallic substances. We introduced pieces of copper in the jars in order to observe the catalytic effect of the metallic particles used in the design process of a transformer. The proportion of liquid insulating to paper and metallic particles is 20:1:0.875, a proportion found in power transformers [4] [35]. The thermal oven used is presented in Figure 2. It is a forced convection oven, controlled in temperature by a Proportional Integral Derivative (PID) controller. The temperature range covered by the controller is (0˚C - 400˚C) ± 2˚C.

Samples of dielectric paper and cooling liquids were taken after 48, 72 and 96 hours of thermal aging, and cooled in the dark for 24 h before performing the various physicochemical tests. Samples of dielectric paper were also taken to study the evolution of the breakdown voltage, based on the standard ASTM D149-09 [36]. The electrodes used have the disc shape, with a diameter of 25 mm and a height of 40 mm both for high voltage and grounded electrodes. The experimental platform used to measure paper BDV is presented in Figure 3. The insulating medium used was methyl ester of palm kernel oil, filling the requirements of the standards IEEE C57.147 [37] and ASTM D6871 [38]. For each sample of paper, the BDV measurement has been done 6 times, and the average value was plot. The analyses performed on the insulating liquids included acidity tests using ASTM D974 [39] and the determination of the quantity of dissolved degradation products using UV spectroscopy in accordance with ASTM D6802

[40]. The range of wavelength analyzed to determine the DDP is 360 nm - 600 nm. The UV spectrometer used is a Spectroquant® Pharo 300.

The evolution of the paper BDV with aging is presented in Figure 4. The paper BDV increases accordingly with the dielectric strength of oil. The papers immersed in MEPKO have a higher BDV than the similar papers immersed in MO. Dry paper possesses airspaces which will be filled by the insulating oil during impregnation; the paper BDV thus increases with respect to the BDV of the dielectric liquid. As the breakdown voltage of natural and methyl esters is greater than the breakdown voltage of mineral oil, the paper BDV dipped in MEPKO is greater than the BDV of paper dipped in MO. During the aging of the oil/paper insulation, moisture migrations will take place between oil and paper; hence, paper BDV will depend on the moisture level in oil and type of the oil. This explains why the BDV of paper dipped in MEPKO remains the highest. Hence, the type of the oil has a significant impact on the BDV of paper. The comparison between the two types of papers used during the thermal aging tests reveals that the BDV of NP-910 is higher than the BDV of TUK. This is due to the fact that the NP-910 has a higher thermal class than TUK. Hence TUK has a less stability

and is more degraded than NP-910. Therefore, in a power transformer, NP-910 can be used in areas where the insulation is subjected to high thermal and electrical stresses, for example the hotspot points, while TUK is used in the other areas of the transformer. In our study, during initial thermal aging, the paper BDV increased with the duration of impregnation. Previous authors who also measured the BDV of insulating paper have also reported that for temperatures less than 115˚C, the paper BDV increases with the duration of impregnation. This observation can be explained by the fact that at 110˚C, penetration of oil into the paper is the main process taking place between oil and paper. During penetration, the airspaces are filled with oil thus increasing the dielectric strength of the solid insulation. The values of BDV of TUK measured were close to the results found by previous authors. In the case of NP-910/MEPKO, the results are higher than those found by previous authors. These high values can be attributed to the fact that MEPKO has a high BDV than the esters used by previous authors. In addition, the samples were sufficiently dehumidified before the beginning of the thermal aging tests. The drying process impacted positively on the samples, as a reduction of the free water molecules of an oil implies the increase of its BDV [16] [17].

The total acid number evolution for the different combination oils and paper is presented in Figure 5. Globally the acidity of methyl esters of vegetable oils is greater than the acidity of mineral oils. However, the measured values of acidity of esters did not exceed the limits prescribed by the standard IEEE C57.147 [37]. Also, the high values of acidity measured with methyl esters do not imply a severe deterioration of solid insulation, as methyl esters produce mainly high molecular acids which react with paper in a transesterification reaction. This reaction enhances the paper resistance against deterioration. In addition, the acids produced can serve as catalysts for a hydrolysis of the methyl esters. As methyl esters have a high saturation limit, they can keep the paper drier than mineral oils.

A comparison of the acidity produced the papers aging reveals that the acidity of Nomex-910 samples in MO is smaller than the acidity of TUK samples in MO. However, the opposite tendency is observed when using MEPKO: the acidity of Nomex-910 samples in MEPKO is greater than the acidity of TUK samples in MEPKO. These contradictory tendencies could indicate that Nomex-910 performs better in MO than MEPKO, while TUK gives better results when immersed in MEPKO than when it is immersed in MO. The results of acidity studies presented in our studies are in accordance with previous results published in the literature showing that at temperatures less than 120˚C, the acidity level of MO/Nomex-910 samples is less than the acidity of MO/TUK samples [12]. Other authors also reported that the stability against oxidation of natural ester/Nomex-910 samples is less than the stability of natural ester/TUK samples [17]. These results reinforce the conclusions drawn after the BDV tests, stating that Nomex papers should be used in transformers only in areas where the solid insulation is subjected to high thermal and electrical stresses, and TUK should be used in the other areas of the transformer.

The aging of insulating oils is closely related to the quantity of decay products generated. New unused oils have a relatively low amount of decay products. As the oils aging increases, the quantity of decay products will also increase. Figure 6(a) and Figure 6(b) present the absorbance spectra of MO and MEPKO at 0 hours and 96 hours respectively, for the solid dielectrics used. The curves shift towards longer wavelength, indicating an increase in the relative DDP. Two absorption peaks are observed with mineral oil, around 380 nm and 450 nm. The same peaks have been also found by previous authors [41] [42]. The integration of the area under the curves gives the decay dissolved products (DDP). Figure 7 presents the variation of the DDP for the 4 types of mixed insulations studied. The DDP has a strong relationship with the amount of dissolved decay content in oil. The DDP of MO/Nomex-910 samples is lower than the DDP of MO/TUK samples, while the DDP of MEPKO/Nomex-910 samples is higher than the DDP

of MEPKO/TUK samples. Hence, Nomex-910 significantly improves the oxidation stability of the MO. However, Nomex-910 reduces the oxidation stability of methyl esters. Even though the thermal class of Nomex paper in methyl ester oil is more when compared with TUK paper, Nomex-910 paper is giving more dissolved decay products in ester oil. An increase in the quantity of decay products implies an increase of the oil viscosity and consequently a reduction of the cooling performance of the oil.

The results of UV-vis spectroscopy confirm the results of acidity studies presented above. Also, these results are in accordance with previous comparative studies which show that Nomex-910 has a higher thermal class than TUK. This high thermal class permits Nomex-910 to achieve a high breakdown voltage than TUK. Also, Nomex-910 improves the oxidation stability of mineral oil, but reduces the oxidation stability of natural and methyl esters. In addition, during the first hours of thermal aging for temperatures in the range (0˚C - 115˚C), the degradation rate of natural esters is greater than the degradation rate of mineral oils. The results obtained in our study are in perfect accordance with the results found previously by other authors who investigated thermal aging effects on the oil/paper mixed insulation [12] [17] [43]. Moreover, the acidity of the oils at the end of the aging period did not exceed the values prescribed by the standards ASTMD3487 [44] for the mineral oils and IEEE C57-147-2018 for natural esters [37] [45]. It would very instructive to perform accelerated aging studies reaching the loss of life point of the oil/paper mixed insulation [23] [46] [47] [48] [49] [50]. The results presented in this work show that MEPKO has good insulating properties and can therefore replace MO in distribution and power transformers. Also, MEPKO performs well when associated with TUK. However, Nomex-910 has a higher thermal class than TUK. Consequently, we propose that Nomex-910 should be used in power transformers mainly in areas where a high thermal stress is experienced. In the other areas, where solid insulation is needed, TUK can be used. Such appropriate use of dielectric paper could help to save money and better preserve the insulating liquid viscosity.