In order to adapt to different on-site monitoring conditions, make full use of multimodal monitoring information, and improve the flexibility and accuracy of transformer fault diagnosis methods, a fault diagnosis method of power transformer based on dynamic multimodal fusion was proposed in this paper. The method introduced a dynamic fusion strategy, which firstly constructed a layer of modal selection network that can autonomously screen the input monitoring information and dynamically select the diagnostic modes to adapt to the different monitoring conditions on site. Secondly, it established corresponding diagnostic models for different input modes, and adopted the corresponding fusion method to diagnose under the non-single modal conditions for achieving the full utilization of the monitoring information of each modality. Finally, actual cases collected from multiple municipal bureaus were used for verification. The results show that the method proposed in this paper can effectively improve the flexibility and accuracy of transformer fault diagnosis results, and can be adapted to different monitoring conditions on site. Compared with other methods, the recognition accuracy of this method is higher, up to 97.33%, and the false alarm rate and missed alarm rate are the lowest.

The crystallinity of chopped fibers has an important impact on the mechanical strength, thermal shrinkage resistance, toughness, and other properties of aramid paper. To investigate the influence and mechanism of chopped fibers crystallinity on the properties of meta-aramid paper, meta-aramid papers with different crystallinity of chopped fibers were prepared by controlled variable method, and their mechanical properties, electrical properties, and thermodynamic properties were tested. The results show that the electric strength of meta-aramid papers is not significantly affected by the crystallinity of chopped fibers, it is mainly determined by precipitated fibers. The tensile index of meta-aramid papers is almost not affected by the crystallinity of chopped fibers before hot-pressing, but increases with the increase of the crystallinity of chopped fibers after hot-pressing. When the crystallinity of chopped fibers increases from 27.19% to 47.91%, the tensile index of papers enhances from 46.2 N·m/g to 58.0 N·m/g. In addition, the thermal shrinkage resistance of papers is closely related to the crystallinity of chopped fibers. When the crystallinity of chopped fibers increases from 27.19% to 47.91%, the thermal shrinkage rate of papers decreases from 4.8% to 1.3%.

The differential bipolar repetitive impulsive voltage was used to simulate the impulse width modulated electrical stress environment of inverter-fed motor. The effects of impulse width, dead zoon time, and impulse frequency on the partial discharge inception voltage (PDIV) test results of enameled twisted pair wires and random wound inverter-fed motor stator winding were studied. Based on the electric field distribution of insulation under bipolar impulsive voltage and gas discharge theory, the influence mechanism of impulse voltage parameters on PDIV was analyzed. The results show that the impulse width is correlated with the generation of initial electrons, leading to a negative correlation between PDIV test results and impulse width. The changes in residual charges and spatial electric field distribution on the insulation surface caused by short dead time may be a reason for the positive correlation between PDIV and dead time. Due to the weak influence of impulse frequency on initial electron generation and electric field distribution, the correlation between PDIV and impulse frequency is low.

To achieve the healability and reprocessability of silicone rubber insulating materials, it is feasible to introduce healable functional groups into the structural network of silicone rubber to extend its service life and reduce the safety hazards of power system operation. This paper used silane coupling agents to introduce amino groups into the main chain of silicone rubber, and then introduced dynamic hindered urea bonds into the silicone rubber network through the reversible reaction of isocyanates and hindered amines. A healable silicone rubber material was prepared, and the mechanical, electrical, healable, and reprocessable properties of silicone rubber were studied. The results show that the insulation performance of silicone rubber containing dynamically hindered urea bonds meets the application standards. Under certain temperature and pressure, silicone rubber samples can heal mechanical damage and electrical breakdown damage with high healing efficiency. The sample can be reprocessed through the cutting-hot pressing process, and the reprocessed sample can still retain certain mechanical and electrical insulation properties.

Cellulose insulation paper affects the reliability of transformer operation due to the long-term action of temperature and electric fields. The existing research mainly focuses on the modification effect of nano-particles on insulation paper under a single physical field, while the modification research under the combined effect of electric-thermal ageing is not yet perfect. This research prepared cellulose insulation papers modified by nano-TiO₂ particle with mass fractions of 0, 1%, 3%, 5%, and 7%, and established corresponding molecular models based on molecular dynamics simulation. Combined with macroscopic electric-thermal combined ageing tests, the influence of electric field and temperature field on the electrical properties of modified cellulose insulation papers was studied, and the optimal modification content of nano-TiO₂ was explored. The results show that the electrical properties of cellulose insulation paper are mainly affected by temperature, and the introduction of electric field accelerates the ageing of cellulose. After adding nano-TiO₂ particles, the number of hydrogen bonds in the cellulose molecular model increases, the relative dielectric constant of the modified cellulose insulation paper decreases, and the electric strength increases. There is a threshold for the improvement of cellulose modification effect by the content of nano-TiO₂, and the best modification effect on cellulose samples is achieved when the mass fraction of nano-TiO₂ is 5%.

Basin insulators in GIS are prone to surface electric field distortion due to charge accumulation, which can lead to surface flashover. To improve the surface voltage resistance performance of insulators, this paper refers to the design concept of surface functional gradient materials and utilizes the uniform treatment characteristics of titanium deposition by atmospheric pressure plasma jet (APPJ). By controlling the treatment time at different radial distances, a basin insulator with continuous conductivity gradient were prepared. At the same time, a fast electric field calculation method based on surface conductivity was established by coupling DC electric field simulation calculations with an attention mechanism-enhanced long short-term memory network (LSTM). The surface conductivity of the basin insulator was continuously gradient optimized using the artificial bee colony (ABC) algorithm. And the electrical performance of the insulator before and after optimization was evaluated and compared using simulation models and flashover tests. The result show that the density and bonding structure of the plasma titanium deposition layer are related to the treatment time, and the optimal treatment time is 3 minutes, at which the surface conductivity of the insulator is the highest. The surface functional gradient modification significantly improves the uniformity of the electric field distribution on the insulator surface. The maximum field strength on the lower surface of the insulator decreases by 70.3%, the electric field distortion rate decreases by 24.69%, and the flashover voltage is 31.34 kV, which is 14.3% higher than that of unmodified insulators.

The traditional time-frequency domain reflection method (TFDR) is difficult to accurately obtain the time-frequency distribution of the acquired signal, which seriously reduces the location accuracy of cable defects. Therefore, a cable defect localization method based on Hilbert transform (HT) and TFDR was proposed in this paper. Firstly, the Gaussian frequency domain window function was introduced by Fourier digital filtering method to suppress the interference of the acquired signal. Then, the time-frequency distribution of incident traveling wave signal and reflect traveling wave signals were obtained by HT method. Finally, the instantaneous power spectrum of time-frequency distribution was used to characterize the time-frequency energy focusing characteristics of the defect reflected traveling wave signal, and the defect position in the cable was located through the peak value of the final curve obtained. The results show that, compared with the traditional Wigner distribution method and synchronous compression transformation method, the traveling wave signal obtained by this method has higher time-frequency resolution and no cross term interference, which can locate local defects in cables more accurately.

Electrical laminated wood is widely used in the field of transformer insulation due to its high mechanical strength, low dielectric loss factor, and dielectric constant close to transformer oil. However, its insulation performance is lower than that of laminated paperboard, and there is a significant risk of insulation accidents when it is used in high field strength area of transformer. In this paper, the raw material selection, processing technology, and current problems of electrical laminated wood were discussed and analyzed, and the influencing factors of its insulation performance were analyzed, providing references for improving the product quality of electrical laminated wood.

To solve the problem of poor thermal conductivity of polyimide (PI), PI was doped with hexagonal boron nitride (h-BN) modified by a surface wetting agent (F068) through in-situ polymerization, and then a series of h-BN/PI high thermal conductivity composites were prepared. The structure of h-BN/PI composites was characterized by Fourier transform infrared spectroscopy and scanning electron microscopy. The thermal conductivity mechanism of h-BN/PI composites was explored by introducing the Y.Agari model, and the influence of modified h-BN filler doping content on the thermal conductivity, mechanical, heat resistance, and insulating properties of the composites was analyzed. The results show that with the increase of modified h-BN doping content, the thermal conductivity and thermal stability of h-BN/PI composites increase, while their tensile strength, elongation at break, and electric strength exhibit a significant downward trend. When the doping amount of modified h-BN is less than 30%, the influence of polymer crystal size factor on the thermal conductivity of the composites is dominant. When the doping amount of modified h-BN is higher than 30%, the influence of the free factor of the thermal chain formed by thermal conductive particles on the thermal conductivity of the composites is dominant. When the doping amount of modified h-BN is 50%, the thermal conductivity of h-BN/PI composite reaches 0.83 W/(m·K), the electric strength is 198.6 kV/mm, the 5% thermal decomposition temperature is 595℃, the tensile strength is 64.8 MPa, and the elongation at break is 10.4%.

There are many types of insulating oils and insulating papers with different compositions, which will have different moisture equilibrium curves when they are used in conjunction. To obtain the moisture equilibrium curves of different oil-paper insulation systems, this research constructed a moisture equilibrium curve model applicable to new oil-paper insulation systems based on the relative moisture content relationship between oil and paper at moisture equilibrium, and verified the model by combining with the measured results of moisture equilibrium experiments. The results show that under typical operating temperature of transformer, the prediction accuracy of the moisture equilibrium curve model for different oil-paper insulation systems is higher than 90%, confirming the effectiveness of the model.