1 Introduction

Modular multilevel converters (MMCs) have been utilized in solid-state transformers (SSTs) taking the superiorities of modular structure, waveform quality improvement, and easy implementation into account [1, 2]. The commercially existing two package technologies are bond-wire and press-pack structure. The package structure of bond-wire module is shown in Fig. 1a [3]. The insulation system of bond-wire IGBT module mainly comprises of the silicone gel and baseplate. Moreover, the packaging structure of press-pack IGBT from WESTCODE company is shown in Fig. 1b [4,5,6,7]. The poly-ether-ether-ketone (PEEK) frame is utilized in the submodule to withstand the high voltage.

Fig. 1
figure 1

Diagram of package structure a bond-wire IGBT; b press-pack IGBT

The high-voltage IGBT module withstands DC voltage in off-state. The insulation and discharge characteristics of packaging insulation materials under DC voltage differ from those under AC voltage. In-depth understanding of the electric field distribution and partial discharge characteristics of the power module under DC voltage is very important for the advanced high-voltage package technology. To date, two existing approaches to apply the high voltage on the bond-wire power module are available [8]. According to the regulation of IEC 61287-1, collector, emitter and gate terminals of power module are connected in series, and the power frequency alternative current (AC) voltage is applied between the short-circuit three terminals and the substrate of the power module. The partial discharge signal of power module can then be tested and measured. Another method is to turn off the device by applying a negative voltage on the gate–emitter terminals or short-circuit the gate and emitter terminals. Then, the tested high voltage is applied on the collector and emitter terminals when the module is working under cut-off condition. The above-mentioned approaches can be utilized in the field of partial discharge measurement for bond-wire IGBT modules. However, only the latter one is suitable for the press-pack IGBT due to the specific package structure.

Lebey et al. studied the partial discharge characteristics of bond-wire IGBT module with rated voltage of 3.3 kV [9]. The phase-resolved partial discharge (PRPD) spectrums and discharge quantities of the test power module at various voltage levels were obtained. Three classic models, i.e., space discharge, corona discharge and surface discharge, were utilized to analyze the discharge mode of the bond-wire IGBT power module. Arumugam et al. [10] tested the partial discharge of bond-wire IGBT modules with diverse aging degrees of silicone gel. The derived PRPD spectrum information and discharge quantity of the tested samples were utilized to assess the state deterioration. In addition, Romano et al. [11] measured the partial discharge of the module under AC voltage and PWM voltage. In the same paper, it is found that the type of applied voltage has a great impact on the partial discharge measurement results.

The packaging structure of bond-wire IGBT is different from that of press-pack IGBT. One of the main insulation structures inside press-pack IGBT is the PEEK frame. Fu Pengyu et al. studied the insulation characteristics of press-pack IGBT in three aspects, i.e., insulation test, physical analysis and reliable design, which provides quantitative indicators for the reliable insulation design of press-pack IGBT. However, simulation and experimental validations of the insulation design methods are not conducted [12]. Furthermore, the internal electric field distribution of the submodule inside press-pack IGBT is predicted by constructing the finite element model under direct current (DC) voltage [13]. It is revealed that the area where electric field is concentrated locates at the end of the silver layer and the surface of the PEEK frame inside the submodule. In addition, partial discharge tests were carried out on press-pack IGBT submodule. The PRPD spectrum characteristics of various package structure, e.g., chip layer, PEEK framework and PEEK and silver layer combination, were obtained under AC and DC superimposed voltage. The influence of the inert gas pressures was also analyzed [14]. However, the influences of voltage polarity on partial discharge are not investigated, and the partial discharge characteristics are not fully analyzed.

In summary, the surface partial discharge characteristics of PEEK is significant of the high-voltage package technology of power semiconductors. Thus, in this paper, a three-dimensional FEM simulation model of press-pack IGBT submodule is built to analyze the electric field distribution. Then, the partial discharge test bench is built to investigate the surface partial discharge characteristics of PEEK under various both positive and negative high voltage. The remainder of this paper is organized as follows. Section 2 presents the FEM model of PPI submodule and the electric field distribution. Section 3 introduces the test bench of the surface partial discharge. Section 4 introduces the surface partial discharge characteristics of the PEEK. The last part draws the conclusion of this paper.

2 Detailed formats of manuscript electronic field analysis of press-pack IGBT sub-module

2.1 FEM model of press-pack IGBT sub-module

Since some selected structure, e.g., gate pins and gate print circuit broad (PCB) have little influence on the electric field distribution, these parts are neglected in the simulation model to mitigate the computation burden. The simplified schematic of the simulation model is shown in Fig. 2. The sizes of each layer are shown in Table 1. Notably, a protective gas layer with a thickness of 50 μm is added to the inner layer of the PEEK framework to simulate the unavoidable gap between PEEK framework and inner structures.

Fig. 2
figure 2

Simplified schematic of the simulation model

Table 1 Size of each layer in the simulation model

The meshing result of the 3D model is shown in Fig. 3. The partial division method is adopted to reduce the number of meshes of the overall model and improves the calculation efficiency. The overall number of elements is 8,748,047, the minimum element mass is 0.1586, and the average element mass is 0.6685, which indicates that the mesh quality of the overall meshing is acceptable. The meshing of the nitrogen environment is hidden, and the size of the nitrogen environment is about 10 times the size of the submodule.

Fig. 3
figure 3

Meshing results

Based on Maxwell’s equations, the governing equations of the DC electric field are:

$$ \left\{ \begin{gathered} \nabla \times E = 0 \hfill \\ \nabla \cdot J = 0 \hfill \\ J = \gamma E \hfill \\ \end{gathered} \right. $$

where E is electric field intensity, J is electric current density, and γ is conductivity. At the interface of different materials, the following continuity equations can be applied.

$$ \left\{ \begin{gathered} n \times (E_{2} - E_{1} ) = 0 \hfill \\ n \cdot (J_{2} - J_{1} ) = 0 \hfill \\ \end{gathered} \right. $$

where n is the normal direction of the interface. They can also be written as the function of electric potential:

$$ \left\{ \begin{gathered} \varphi_{1} = \varphi_{2} \hfill \\ \gamma_{2} \frac{{\partial \varphi_{2} }}{{\partial n_{2} }} = \gamma_{1} \frac{{\partial \varphi_{1} }}{{\partial n_{1} }} \hfill \\ \end{gathered} \right. $$

Press-pack IGBT can withstand forward DC voltages in the off state, but cannot block reverse high voltages due to the presence of diodes in the IGBT package. In the electric field simulation model, the external surface of collector and emitter surfaces are set to 3.3 kV and ground, respectively. In addition, the setting of the other surfaces is set to insulation surface to simulate the interface with the N2. Conductivities of the utilized materials are listed in Table 2.

Table 2 Conductivities of the utilized materials

2.2 Simulation results

The voltage distribution and electric field distribution results derived from the developed electric field simulation model is shown in Fig. 4. It shows that the voltage distribution decrease from the collector to emitter, and the PEEK framework withstands approximately all the voltage drops, which causes the critical electric field concentration inside the submodule.

Fig. 4
figure 4

Voltage and electric field distribution simulation results

3 Test bench of surface partial discharge

3.1 DC high-voltage test bench

The established partial discharge test bench under DC high-voltage is shown in Fig. 5. The output voltage of DC source can be adjusted by the voltage regulator. C is the filter capacitor with a capacity of 10nF and a rated voltage of 50 kV, which can reduce the ripple of output voltage. Rp is a protective resistor with a resistance of 3 MΩ, which used as current limiting protection. R1 and R2 form a resistance divider with a total resistance of 200 MΩ. Ck is the coupling capacitance with a capacitance of 600pF, which forms the of partial discharge signal loop with the device under test (DUT).

Fig. 5
figure 5

Partial discharge test bench under DC high voltage

Rogowski coil is placed between the coupling capacitor and the test object. Through the ratio frequency (RF) line, the partial discharge signal can be obtained by the combination of the signal acquisition system and the computer. The bandwidth of the high-frequency current transformer (HFCT) acquisition system used in this test platform is 3–50 MHz, and the voltage current ratio is 37 mV: 1 mA. The DC voltage generated by the test bench comprises ripple, which has a great impact on the partial discharge test system. Excessive DC ripple coefficient will affect the accuracy of measured partial discharge characteristics. The ripple coefficient of the test platform is less than 1%, which meet the standard and requirements of partial discharge experiment. The background noise of the platform is no higher than 6pC, which is also agreed with the experimental requirements.

In the light of the above discussed simulation results, electric field concentrates on the surface of PEEK framework. Thus, a PEEK sample, which is a square with a thickness of 3 mm and a length and width of 70 mm shown in Fig. 6, is utilized to investigate the surface partial discharge characteristics. In addition, a cylindrical electrode and plate electrode are utilized as electrodes to apply the voltage, and the PEEK sample is placed at the bottom of supported layer. The diameter of the cylindrical electrode is 5 mm, the chamfer radius of the contact surface with the sample is 1 mm, the thickness of the plate electrode is 2 mm, the width is 90 mm, and the distance d between the electrodes is 3 mm. The cylindrical electrode is connected to the high voltage through the connecting rod, and the plate electrode is grounded. The surface discharge characteristics of PEEK are analyzed under different DC voltage. Before the test, the samples are cleaned and dried to prevent dust from affecting the surface discharge measurement. The diagram of the surface partial discharge model is shown in Fig. 7.

Fig. 6
figure 6

PEEK sample

Fig. 7
figure 7

Diagram of the surface partial discharge model

3.2 Voltage rise process

Stepped-stress test method is utilized here to measure the surface discharge initial voltage. According to the recommended standard, the applied DC voltage, whose waveform is shown in Fig. 8, rises slowly with a steady step rate until partial discharge occurs. The initial discharge voltage is defined as the applied voltage when at least one discharge pulse occurs within one minute. Under the same voltage, 10–20 discharge pulses are collected, and the average value is taken as the initial discharge quantity. After that, the voltage is reduced to 0 V until the accumulated charge on the PEEK material sample is fully dissipated, which will take at least 15 min. Each initial discharge voltage is repetitively measured three times taking the dispersion into account.

Fig. 8
figure 8

Waveform of the applied DC voltage

The magnitude of surface discharge is measured by using fast boost method. Since the DC voltage is not reversed, the surface of insulating medium will accumulate charges under the action of applied voltage, namely the accumulated charges will affect the measurement of partial discharge. During the test, it is necessary to replace the samples or wait for at least 15 min before conducting the next voltage level test. In addition, the partial discharge is measured after the voltage rise to the target voltage lasts for 5 min to reduce randomness.

4 PEEK surface discharge characteristics under DC voltage

The q − t spectrum is the discharge sequence spectrum of DC partial discharge, which reflects the change of discharge quantity over time. The Δti − Δti+1 spectrum shows the relationship between the previous discharge time interval and the following discharge time interval in every three discharges. The qi − qi+1 spectrum shows the relationship between discharge quantities in two adjacent discharges. The Δqi − Δqi+1 spectrum shows the relationship between the amplitude difference of the first two discharges quantities and amplitude difference of the last two discharges quantities in three consecutive discharges, which indicate the amplitude difference relationship of continuous discharges. The Δt − n spectrogram and q − n spectrogram reflect the repetition rate of a certain discharge time interval and discharge amplitude, respectively.

4.1 Influence of applied voltage on PEEK surface discharge

In this paper, the surface discharge quantity and discharge repetition rate n/t under different voltages are investigated. Since the model is asymmetric, both positive and negative DC voltage are utilized to analyze the surface discharge characteristics. The distance between two electrodes is set at 3 mm. Positive or negative voltage are connected with cylindrical electrode, and the plate electrode is grounded.

The measured initial surface discharge voltage is 2.6 kV under positive voltage, and -1.9 kV under negative voltage which indicate a polarity effect. The q − t spectrograms of surface partial discharge under 3 kV ~ 6 kV with the duration of 100 s are shown in Fig. 9. When the applied voltage is 3 kV, the discharge pulse amplitude is concentrated at 20–25pC; when the applied voltage is 4 kV, the discharge pulse amplitude is mainly concentrated in 20–40pC, and there are some discharge points about 60pC; when the applied voltage is 5 kV, the discharge pulse amplitude is mainly concentrated in 20–50pC, and some can reach to near 90pC; when the applied voltage is 6 kV, the discharge pulse amplitude is mainly concentrated in 20–160pC, and high amplitude discharge occurs near 280pC.

Fig. 9
figure 9

q − t spectrum under positive DC voltage a 3 kV; b 4 kV; c 5 kV; d 6 kV

The relationship between maximum surface discharge quantity and average surface discharge quantity at different magnitude of the positive DC voltages is shown in Fig. 10. With the increment in applied voltage, the maximum surface discharge quantity and average surface discharge quantity both gradually increase. In detail, the maximum discharge quantity increases from 150 to 340pC, and the growth rate is 190pC/kV. The average discharge quantity increases from 43 to 110pC, and the growth rate is 67pC/kV.

Fig. 10
figure 10

Discharge quantity over the magnitude of positive DC voltage

The relationship between applied voltage and discharge repetition rate is shown in Fig. 11. It shows that the discharge repetition rate increases with the voltage magnitude. The discharges repetition rate can be regarded as a linear increase when the voltage is less than 4 kV, and the growth rate is 5.5 times/kV. However, when the voltage is above 4 kV, the number of discharges increases rapidly, with a growth rate of 22 times/kV.

Fig. 11
figure 11

Discharge repetition rate over the magnitude of positive DC voltage

Considering the influence of voltage polarity on the discharge process, Fig. 12 shows the pulse sequence diagram of − 3 ~ − 6 kV partial discharge under negative voltage. Compared with the positive polarity, the discharge quantity is smaller under the negative polarity voltage, and the discharge amplitude increases slowly with the increase in voltage. Due to the polarity effect of the model, the diffusion of electrons in space is rapid, and the distortion of the electric field near the pole electrode under the positive polarity is more serious. When the applied voltage is − 3 kV, the discharge pulse amplitude is concentrated at 15–25pC; the discharge pulse amplitude is mainly concentrated in 20–40pC at − 4 kV, the discharge pulse amplitude is mainly concentrated in 20–60pC and some discharges of about 90pC occurs at − 5 kV, the discharge pulse amplitude is mainly concentrated in 20–120pC, and high amplitude discharge over 200pC occurs at − 6 kV.

Fig. 12
figure 12

q − t spectrum under negative DC voltage a − 3 kV; b − 4 kV; c − 5 kV; d − 6 kV

The relationship between maximum surface discharge quantity and average surface discharge quantity and applied voltage is shown in Fig. 13. In the voltage range of − 5 kV ~ − 6 kV, the discharge capacity increases rapidly, the average discharge capacity increases from 41 to 94pC, and the growth rate is 53pC/kV; the maximum discharge capacity increased from 100 to 283pC, and the growth rate is 183pC/kV.

Fig. 13
figure 13

Discharge quantity over the magnitude of negative DC voltage

The relationship between discharge repetition rate and voltage under negative voltage is shown in Fig. 14. The discharge repetition rate increases slowly from 3 to 5 times at 3–4 kV, and the growth rate is 2 times/kV. In 4–6 kV, the growth rate is significantly accelerated, the discharge repetition rate increases from 5 times to 48.8 times, and the growth rate is 21.9 times/kV.

Fig. 14
figure 14

Discharge repetition rate over the magnitude of negative DC voltage

The test results under the negative voltage are similar to those under the positive voltage. The discharge quantities and the discharge repetition rate have a rising trend with the increase in voltage and rise rapidly at 5–6 kV. The reason is under low voltage, the discharge is mainly concentrated near the column electrode where the electric field intensity is relatively concentrated, the discharge intensity is low, the discharge frequency is small, and the electron acquisition ability from the electric field is small, and the discharge capacity is low. With the increase in voltage, the energy obtained by the electrons near the column electrode from the electric field, the movement rate, the discharge intensity, the frequency of electron avalanches, the discharge repetition rate and the discharge amount increase. In addition, due to the increase in electric field strength in the area far from the column electrode, a small amount of discharge sequence with large discharge capacity but low discharge repetition rate will also appear in the external field area, and the discharge in the external field will become more intense with the increase in voltage.

4.2 Surface discharge characteristic spectrum of PEEK

The characteristic spectra of the partial discharge under both positive and negative polarity at the magnitude of 6 kV are analyzed, respectively. Figure 15 shows the spectrum of the time interval between two adjacent discharges at 6 kV and − 6 kV, and the whole spectrum is triangular. The triangle in the spectrogram indicates that the time interval between the previous and latter discharge will become longer if the time interval between the previous discharge is shorter in three consecutive discharges. This phenomenon explains that the discharges will affect each other. The whole part with small time interval takes on a rectangular shape, indicating that the influence between discharges with small time interval is relatively small, and the discharges are relatively independent.

Fig. 15
figure 15

Spectral diagram of adjacent discharge time interval: a 6 kV; b − 6 kV

The relationship between two adjacent discharge amplitudes at 6 kV and − 6 kV is shown in Fig. 16. When the voltage is positive, the discharge quantity is relatively concentrated within 60pC, and the spectrum is an equilateral triangle. When the voltage is negative, the discharge is concentrated within 40pC, and the spectrum is inverted triangle. Between the two adjacent discharges, if the discharge quantity of the previous discharge is large, the discharge quantity of the following discharge will be probably reduced. This phenomenon reveals that the discharge will affect each other, namely the discharge is not an independent event, which is consistent with the analysis results of adjacent time interval spectrograms. Moreover, according to the observation, it can be concluded that the density of discharge points in the rectangular box with low discharge amount (shown in the red box) is high, indicating that the discharge with small discharge quantity is relatively independent.

Fig. 16
figure 16

Relationship between two adjacent discharge amplitudes: a 6 kV; b − 6 kV

The relationship between the amplitude difference of two adjacent partial discharges under 6 kV and − 6 kV voltages is shown in Fig. 17. The discharge amplitude difference is mainly concentrated near the origin, the amplitude difference of adjacent discharges is small, and the spectrum is triangular. It shows that two discharges with large amplitude and one discharge with small amplitude will occur in three consecutive discharges.

Fig. 17
figure 17

Amplitude difference spectrum of two adjacent partial discharges: a 6 kV; b -6 kV

In light of the above analysis of adjacent discharge time interval spectrogram and adjacent discharge amplitude spectrogram, small amplitude discharge and discharge with lower time interval are relatively independent discharges. It shows that in the part close to the column electrode, due to small space and concentrated electric field, it is easy to have electron collision, which leads to partial discharge, and the discharge is relatively independent; at the place far from the column electrode, the electron free path is larger, and the discharge capacity increases with the increase. And because of the large amount of discharge, the space charge dissipates slowly, and the discharge will affect each other.

The spectrum of adjacent time intervals and discharge times under 6 kV and − 6 kV voltages is shown in Fig. 18. Under the negative polarity, the discharge is concentrated in the area with small time intervals. The longer the time interval is, the smaller the discharge times will be. The overall spectrum shows a triangle shape. Under positive polarity, the discharge is more concentrated in a smaller time interval.

Fig. 18
figure 18

Spectrum of adjacent discharge interval over discharge times: a 6 kV; b − 6 kV

The spectrum of discharge amplitude and discharge times under 6 kV and − 6 kV voltages is shown in Fig. 19. It can be seen from the figure that there are relatively more discharges with lower discharge amplitude under different polarity voltages, and the higher the amplitude is, the lower the discharge times will be. And for positive voltage, the discharge is concentrated below 150pC, the discharge spectrum is more irregular, and there is a rapid decrement.

Fig. 19
figure 19

Spectrum of discharge amplitude over discharge times: a 6 kV; b − 6 kV

4.3 Surface discharge waveform of PEEK

The waveform of PEEK surface discharge is measured which is shown in Fig. 20. It shows that the pulse peak value is about 45 mV, and the waveform is vibration attenuation. The wave head time is about 20 ns.

Fig. 20
figure 20

Pulse voltage waveform

To filter the noise in the nonstationary surface discharge signal, wavelet denoising method is utilized to process the discharge waveform. It decomposes the wavelets of the noisy signal to obtain the wavelet coefficient. Since the wavelet coefficient of the signal is larger than the wavelet coefficient of the noise, by selecting a suitable threshold, noise can be filtered by remove the wavelet coefficient below the threshold. The whole process can be divided into three parts: feature extraction, low-pass filter and signal rebuilt, which is shown in Fig. 21.

Fig. 21
figure 21

Process of wavelet denoising method

The filtered PEEK surface discharge waveform is shown in Fig. 22, which has better signal-to-noise ratio. The spectral characteristics shown in Fig. 23 indicate the main frequency range of PEEK surface discharge is 10–20 MHz.

Fig. 22
figure 22

Filtered partial discharge waveform

Fig. 23
figure 23

Spectrum distribution of partial discharge waveform

5 Conclusion

In this paper, the finite element simulation model of one submodule in press-pack IGBT is established, and the electric field distribution of the PPI submodule is obtained. Moreover, the test platform of surface discharge of PEEK sample is developed to investigate the surface partial discharge characteristics. The main conclusions of this paper are as follows:

  1. 1.

    The surface partial discharge of PEEK sample has polarity effect. Under positive and negative polarity voltage, the average discharge quantity, maximum discharge quantity and discharge repetition rate of the tested sample have positive correlation with the amplitude of the applied voltage.

  2. 2.

    The qi − qi + 1 spectrum of PEEK material is a right triangle whose two right-angled sides are parallel to the coordinate axis. In addition, Δqi − Δqi + 1 spectrum is also a triangle, while the center of the triangle is the origin point of the coordinate system.

  3. 3.

    The front time of surface partial discharge waveform of the tested PEEK sample is longer, which can reach 20 ns. Moreover, the frequency spectrum is concentrated in the range of 10–20 MHz and the overall waveform oscillation attenuation.