Research on an Electromagnetic Interference Test Method Based on Fast Fourier Transform and Dot Frequency Scanning for New Energy Vehicles under Dynamic Conditions

: In recent years, electromagnetic interference (EMI) of new energy vehicles, including di ﬀ erence mode symmetric interference and common mode asymmetry interference, has attracted the attention of many scholars. So far, EMI tests for new energy vehicles under steady conditions cannot reﬂect the actual EMI of the running vehicle. The results of EMI test methods based on fast Fourier transform (FFT) under dynamic conditions have worse frequency resolutions, and frequency / amplitude accuracy has low precision. Therefore, this paper proposes an EMI test method based on FFT and dot frequency scanning (DFS) for new energy vehicles under dynamic conditions. The identiﬁcation method for accelerating, sliding, and braking conditions is studied. A comprehensive EMI key evaluation index system for new energy vehicles is built, including characteristic points with maximum amplitude, area, ratio, and density coe ﬃ cients for high-amplitude characteristic points. Among them, the maximum amplitude is an index to evaluate extreme values. The ratio of high-amplitude characteristic points is a comprehensive index to evaluate the overall region. The density coe ﬃ cient is an index to evaluate the local region. Finally, this method is applied to three vehicles. With the same instruments, by reducing the FFT frequency span, the frequency resolution and frequency accuracy increase. The results indicate that the EMI of new energy vehicles can be tested under dynamic conditions with high accuracy according to the operable evaluation indexes.


Introduction
Nowadays, new energy vehicles are rapidly developed and have become an important means of transportation [1,2]. However, compared to other types of vehicles, new energy vehicles have more electronic components, which are also more complex. In charging and discharging states, serious electromagnetic interference (EMI) occurs, including difference mode symmetric interference and common mode asymmetry interference [3]. EMI testing is one of the most important tests for new energy vehicles. The current EMI test standards for new energy vehicles (e.g., CISPR 12:2009 [4], ECE 10.05 [5], SAE J551-5-2012 [6], GB/T 18387-2017 [7], etc.) only stipulate that the vehicle is tested under steady conditions [8] (e.g., SAE J551-5-2012: when under BRAKE APPLIED, CREEP, and CRUISE steady conditions, the vehicle conditions are constant). Under steady conditions, the EMI test cannot truly reflect the EMI of the running vehicle. SAE J551-5-2012 noticed this problem and indicated that the EMI test standard for dynamic conditions (where the driving conditions change) is still under study [9][10][11]. Therefore, it is very urgent to study EMI testing technology for new energy vehicles under dynamic conditions.   Figure 2a shows an analysis of the rough sweep with the FFT spectrum, and Figure 2b shows the precise sweep process with FFT and DFS. The test includes the rough sweep and precise sweep. The rough sweep analyzes various dynamic conditions of vehicle, including accelerating, sliding, and braking conditions, which can be seen in the upper part of Figure 2a. The abscissa is time, and the ordinate is speed. An analysis of the FFT spectrum of the magnetic field's radiation emission intensity is carried out to obtain the EMI intensity graph, which can be seen in the lower part of Figure 2a. The abscissa is time, and the ordinate is frequency. Different colors are used to distinguish the amplitude. The precise sweep also completes the same dynamic conditions of the vehicle as the rough sweep, which can be seen in the upper part of Figure 2b. The abscissa is time, and the ordinate is speed. Results of the DFS magnetic field's radiation emission is carried out to obtain the EMI amplitude graph, which can be seen in the lower part of Figure 2b. The abscissa is time, and the ordinate is amplitude.   Figure 2a shows an analysis of the rough sweep with the FFT spectrum, and Figure 2b shows the precise sweep process with FFT and DFS. The test includes the rough sweep and precise sweep. The rough sweep analyzes various dynamic conditions of vehicle, including accelerating, sliding, and braking conditions, which can be seen in the upper part of Figure 2a. The abscissa is time, and the ordinate is speed. An analysis of the FFT spectrum of the magnetic field's radiation emission intensity is carried out to obtain the EMI intensity graph, which can be seen in the lower part of Figure 2a. The abscissa is time, and the ordinate is frequency. Different colors are used to distinguish the amplitude. The precise sweep also completes the same dynamic conditions of the vehicle as the rough sweep, which can be seen in the upper part of Figure 2b. The abscissa is time, and the ordinate is speed. Results of the DFS magnetic field's radiation emission is carried out to obtain the EMI amplitude graph, which can be seen in the lower part of Figure 2b. The abscissa is time, and the ordinate is amplitude.
The rough sweep and precise sweep are two separate processes. Only when the driving conditions of rough sweep and precise sweep are completely consistent will the precise process be meaningful. In the current study, the pneumatic manipulator is controlled only by the air pressure. Control precision is not enough. Driving conditions during rough sweep and precise sweep are difficult to equally maintain. However, condition parameters such as speed and acceleration during the precise sweep process can be used to identify various conditions. This can make sure the precise sweep conditions correspond to those of the rough sweep. Then, DFS is performed on the corresponding conditions. At the same time, speed and acceleration constantly change when the vehicle accelerates, slides, and brakes. Every frame of the FFT spectrum corresponds to different speeds and accelerations. The EMI is much more complicated than the steady conditions. The point with the maximum subtraction value between the measured value and relevant standard limit in a single spectrogram is called the characteristic point. Ideally, each characteristic point needs DFS. In fact, both identification conditions and changing sweep frequency on the EMI receiver take some time. It is difficult to perform DFS on every characteristic point under dynamic conditions. Choosing the characteristic points for further sweeping from so many points is important. Therefore, in order to achieve an EMI test based on FFT and DFS for new energy vehicles under dynamic conditions, it is necessary to study the methods to identify dynamic conditions and determine characteristic points. The rough sweep and precise sweep are two separate processes. Only when the driving conditions of rough sweep and precise sweep are completely consistent will the precise process be meaningful. In the current study, the pneumatic manipulator is controlled only by the air pressure. Control precision is not enough. Driving conditions during rough sweep and precise sweep are difficult to equally maintain. However, condition parameters such as speed and acceleration during the precise sweep process can be used to identify various conditions. This can make sure the precise sweep conditions correspond to those of the rough sweep. Then, DFS is performed on the corresponding conditions. At the same time, speed and acceleration constantly change when the vehicle accelerates, slides, and brakes. Every frame of the FFT spectrum corresponds to different speeds and accelerations. The EMI is much more complicated than the steady conditions. The point with the maximum subtraction value between the measured value and relevant standard limit in a single spectrogram is called the characteristic point. Ideally, each characteristic point needs DFS. In fact, both identification conditions and changing sweep frequency on the EMI receiver take some time. It is difficult to perform DFS on every characteristic point under dynamic conditions. Choosing the characteristic points for further sweeping from so many points is important. Therefore, in order to achieve an EMI test based on FFT and DFS for new energy vehicles under dynamic conditions, it is necessary to study the methods to identify dynamic conditions and determine characteristic points.

Identification Method for Accelerating, Sliding, and Braking Conditions
The method to identify accelerating, sliding, and braking conditions is mainly based on different features of different dynamic conditions. Suppose there are n measured values of speed in a very short time, observe t .

Identification Method for Accelerating, Sliding, and Braking Conditions
The method to identify accelerating, sliding, and braking conditions is mainly based on different features of different dynamic conditions. Suppose there are n measured values of speed in a very short time, t observe . Then, we have the measured values of speed, V measured = [v 1 , v 2 , . . . , v i , . . . , v n ], and the calculated values of acceleration, A calculated =[a 1 , a 2 , . . . , a i , . . .]. The minimum speed in t observe is v min , v min = min{V measured }. The minimum acceleration is a min , a min = min{A calculated }. The maximum acceleration is a max , a max = max{A calculated }. Features of different dynamic conditions of new energy vehicles are shown in Table 1. The physical significance is as follows: (1) When the new energy vehicle is accelerating, a i > 0, and then a min > 0. To avoid misjudgments, a small, positive threshold, a T_acc , is introduced. Therefore, a i ≥ a min > a T_acc > 0. On the other hand, if the situation when a min > a T_acc is detected, it indicates that the new energy vehicle is accelerating. (2) When the new energy vehicle is sliding, v idling < v i and a i < 0, then v idling < v min ≤ v i and a max < 0. To avoid misjudgments, a small, negative threshold, −a T_slide , is introduced. Therefore, v idling < v min ≤ v i and a max < −a T_slide . On the other hand, if the situation when v i ≥ v min > v idling and a max < −a T_slide is detected, it indicates that the new energy vehicle is sliding. (3) When the new energy vehicle is braking, a i < a slide < 0, then a i ≤ a max < a slide < 0. To avoid misjudgments, a small, negative threshold, −a T_brake , is introduced. Therefore, a i ≤ a max < a slide − a T_brake < 0. On the other hand, if the situation when a i ≤ a max < a slide − a T_brake is detected, it indicates that the new energy vehicle is braking.

Acceleration Sliding Braking
Features of speed and acceleration Figure 3 shows the identification of results from dynamic conditions. The abscissa is time (s). The ordinate is speed (km/h) and acceleration (m/s 2 ). According to the features of speed and acceleration, the dynamic conditions of the vehicle during the period from 0 to 43.5 s can be identified, which consist of 0~5.6 s (accelerating), 5.6~8.0 s (sliding), 8 Figure 3 shows the identification of results from dynamic conditions. The abscissa is time (s). The ordinate is speed (km/h) and acceleration (m/s 2 ). According to the features of speed and acceleration, the dynamic conditions of the vehicle during the period from 0 to 43.5 s can be identified, which consist of 0~5.6 s (accelerating), 5.6~8.0 s (sliding), 8.0~9.2 s (braking), 9.2~12.3 s (sliding), 12.3~20.8 s (accelerating), 20.8~25.3 s (sliding), 25.3~27.0 s (braking), 27.3~30.5 s (sliding), 30.5~40.2s (accelerating), and 40.2~43.5 s (sliding). Based on the features of speed and acceleration, different dynamic conditions can be identified from any group of speed and acceleration curves. In steady conditions, there is no dynamic accelerating, sliding, and braking conditions but only idling, cruising, and steady braking conditions. EMI is more in line with actual driving conditions under dynamic conditions for changing speed.

Method to Determine Characteristic Points
Under steady conditions, const = v and 0 = a . According to a spectrogram of the frequency sweep with lower sweeping speed, a characteristic point corresponding to the maximum amplitude can be obtained. As multiple spectrograms with faster speeds were obtained from FFT spectrum analyses, each spectrogram is a repeated test. Every spectrogram has a characteristic point. Among all characteristic points, the characteristic point corresponding to the maximum amplitude is the final characteristic point.
A schematic of EMI test results based on FFT for new energy vehicles under dynamic conditions is shown in Figure 4. The upper graph shows the speed and acceleration curves, whose x coordinate is time and y coordinate is speed/acceleration. The lower graph shows the 3D diagram of the FFT spectrum analysis, whose x coordinate is time, y coordinate is amplitude, and z coordinate is frequency. It can be seen that the speed and acceleration under dynamic conditions change all the time. The speeds/accelerations of multiple spectrograms from FFT spectrum analyses, as well as the characteristic points, are different, which should be paid attention to under dynamic conditions. Therefore, the method to determine characteristic points under dynamic conditions is more complicated than the method under steady conditions with constant speed. Moreover, in applying

Method to Determine Characteristic Points
Under steady conditions, v = const and a = 0. According to a spectrogram of the frequency sweep with lower sweeping speed, a characteristic point corresponding to the maximum amplitude can be obtained. As multiple spectrograms with faster speeds were obtained from FFT spectrum analyses, each spectrogram is a repeated test. Every spectrogram has a characteristic point. Among all characteristic points, the characteristic point corresponding to the maximum amplitude is the final characteristic point.
A schematic of EMI test results based on FFT for new energy vehicles under dynamic conditions is shown in Figure 4. The upper graph shows the speed and acceleration curves, whose x coordinate is time and y coordinate is speed/acceleration. The lower graph shows the 3D diagram of the FFT spectrum analysis, whose x coordinate is time, y coordinate is amplitude, and z coordinate is frequency. It can be seen that the speed and acceleration under dynamic conditions change all the time. The speeds/accelerations of multiple spectrograms from FFT spectrum analyses, as well as the characteristic points, are different, which should be paid attention to under dynamic conditions. Therefore, the method to determine characteristic points under dynamic conditions is more complicated than the method under steady conditions with constant speed. Moreover, in applying FFT spectrum analyses to EMI, the frequency resolution accuracy is low, and the exact characteristic points under dynamic conditions need further research. FFT spectrum analyses to EMI, the frequency resolution accuracy is low, and the exact characteristic points under dynamic conditions need further research.
where suspect U is the set of amplitudes and suspect 1 , ..., , ..., . suspect F is the frequency set and A distribution diagram of characteristic points is shown in Figure 5 with the set of characteristic points. The abscissa is frequency, and the ordinate is amplitude.

Distribution Diagram of Characteristic Points
p suspect_ j U j , f suspect_j is the j-th characteristic point, where U j is the amplitude and f suspect_ j is the frequency of p suspect_ j U j , f suspect_ j . Suppose that there are z spectrograms in a group of FFT spectrum analysis results. The set of characteristic points is: where U suspect is the set of amplitudes and U suspect = U 1 , . . . , U j , . . . , U z . F suspect is the frequency set Suppose that p max (U max , f U_max ) is the characteristic point with maximum amplitude and U max = max U suspect . p min (U min , f U_min ) is the characteristic point with minimum amplitude and U min = min U suspect .
A distribution diagram of characteristic points is shown in Figure 5 with the set of characteristic points. The abscissa is frequency, and the ordinate is amplitude.

EMI Evaluation Indexes for New Energy Vehicles under Dynamic Conditions
According to Figure 5,

EMI Evaluation Indexes for New Energy Vehicles under Dynamic Conditions
According to Figure 5, p max (U max , f U_max ), area Π, point ratio η, and density coefficient ρ of high-amplitude characteristic points can be obtained. The physical significance is as follows: (1) The characteristic point has a maximum amplitude p max U max , f suspect_max and a set of characteristic points, U max = max U suspect , then f U_max can be found. U max is an index to evaluate extreme values, indicating the maximum EMI value under these dynamic conditions. The larger U max is, the larger the EMI value is, and the larger EMI that is generated by new energy vehicles.
(2) Area Π and point ratio η of high-amplitude characteristic points. Area Π of high-amplitude characteristic points is the area with the amplitude higher than the threshold, U T . The amplitude range of the characteristic points in area Π is [U T , U max ]. Suppose that there are N Π characteristic points in area Π. Then, the ratio between N Π and the total number of characteristic points, z, is named point ratio η of high-amplitude characteristic points. η is calculated as: When U T is constant, the larger the η is, the larger the EMI that is generated at multiple frequencies by new energy vehicles. Π and η are comprehensive indexes to evaluate the overall region.
(3) Density coefficient ρ of high-amplitude characteristic points. Set a rectangular shape area as a local observation area, Π f , with left frequency boundary f left , frequency width w, lower amplitude boundary U down , and amplitude length l (l = U max − U down ). The amplitude range of the characteristic points in area When U down = U T , suppose that there are N characteristic points in area Π f . Then, the ratio between N and wl is named the density coefficient ρ of high-amplitude characteristic points. ρ is calculated as: The total number of characteristic points z increases with time, resulting in N of the same condition and the same region also increases with time. ρ will be affected as different dynamic conditions have different durations. In order to compare the ρ among different dynamic conditions, condition duration t is introduced. The density per unit time ρ t is calculated as: When U down and w are constant (U down = U T ), the larger the ρ is, the more EMI that is generated by new energy vehicles, indicating a greater EMI in this area. ρ is an index to evaluate local regions.
According to the definitions of p max (U max , f U_max ), Π, η. and ρ, EMI evaluation indexes for new energy vehicle under dynamic conditions are shown in Figure 6.

Method to Determine Characteristic Points Based on EMI Evaluation Indexes
A distribution diagram of characteristic points and EMI evaluation indexes is illustrated above. Now, the problem is that the frequency resolution's accuracy from the FFT spectrum analysis is low. Searching all the characteristic points that effect EMI evaluation indexes the most in carrying out

Method to Determine Characteristic Points Based on EMI Evaluation Indexes
A distribution diagram of characteristic points and EMI evaluation indexes is illustrated above. Now, the problem is that the frequency resolution's accuracy from the FFT spectrum analysis is low. Searching all the characteristic points that effect EMI evaluation indexes the most in carrying out small frequency range FFT and DFS helps to improve accuracy of the frequency resolution.
U max is an index to evaluate extreme values. Π and η are comprehensive indexes to evaluate the overall region. ρ is an index to evaluate the local region. p max (U max , f U_max ) and ρ are EMI evaluation indexes that the characteristic points affect the most. Therefore, the method to determine characteristic points based on EMI evaluation indexes under dynamic conditions is as follows: f U_max is a frequency concerned by the EMI test for new energy vehicles under dynamic conditions. (2) When U down (U down = U T ) and w are constant, the larger ρ is, the more EMI that is generated by new energy vehicles. Therefore, the frequencies in the area whose ρ is larger than the threshold (especially the area whose ρ is maximum) should be included in the EMI test for new energy vehicles under dynamic conditions.

Implementation Flow
Based on the method to identify dynamic accelerating, sliding, and braking conditions, and to determine the characteristic points, EMI tests for new energy vehicles under dynamic conditions can be implemented. A flow chart of an EMI test based on FFT and DFS for new energy vehicles under dynamic conditions is shown in Figure 7, which mainly includes rough sweep based on FFT and precise sweep based on FFT and DFS. Firstly, rough sweep obtains a distribution diagram of the characteristic points under different dynamic conditions. Secondly, p max (U max , f U_max ) and the area whose ρ is maximized are calculated. Finally, precise sweep takes place. In the second step, it is very important to select U T and w reasonably.

Construction of the Experimental System
According to Figure 1, the experimental system of the EMI test based on FFT and DFS for new energy vehicles under dynamic conditions was constructed, as shown in Figure 8. The whole system was arranged in a control room and a shielded room. The control room contained a signal analyzer, an EMI receiver, a computer from an EMI test subsystem, an IPC, and an ADC from a roller subsystem. The signal analyzer was N9030A PXA from Agilent. The EMI receiver was N9038A MXE from Agilent. The computer contained a B85-HD3-A board, an i7-4790K CPU, 16 GB RAM, and Windows 10 operating system. The velocity was transformed into voltage in the range from −10 to 10 V, which was outputted with a BNC on an IPC. The ADC was NI 9220 from National Instrument. The shielded room contained an antenna from the EMI test subsystem and a pneumatic manipulator, a test vehicle, and a chassis dynamometer from roller subsystem. The pneumatic manipulator was customized by Festo AG & Co. KG. GA5 PHEV produced by Guangzhou Automobile Group Co., Ltd., and the electric test vehicles of Zotye Auto Co., Ltd. and BYD Co., Ltd. were chosen. The chassis dynamometer was made by AVL List GmbH. Under the condition that the wheel and the roller do not slip relative to each other, the vehicle speed was equal to the linear speed of the roller. The roller could handle a range of velocity (0~200 km/h) in both front and reverse directions. Its velocity sensor had a range of accuracy of 0.05-0.1 km/h.

Construction of the Experimental System
According to Figure 1, the experimental system of the EMI test based on FFT and DFS for new energy vehicles under dynamic conditions was constructed, as shown in Figure 8. The whole system was arranged in a control room and a shielded room. The control room contained a signal analyzer, an EMI receiver, a computer from an EMI test subsystem, an IPC, and an ADC from a roller subsystem. The signal analyzer was N9030A PXA from Agilent. The EMI receiver was N9038A MXE from Agilent. The computer contained a B85-HD3-A board, an i7-4790K CPU, 16 GB RAM, and Windows 10 operating system. The velocity was transformed into voltage in the range from −10 to 10 V, which was outputted with a BNC on an IPC. The ADC was NI 9220 from National Instrument. The shielded room contained an antenna from the EMI test subsystem and a pneumatic manipulator, a test vehicle, and a chassis dynamometer from roller subsystem. The pneumatic manipulator was customized by Festo AG & Co. KG. GA5 PHEV produced by Guangzhou Automobile Group Co., Ltd., and the electric test vehicles of Zotye Auto Co., Ltd. and BYD Co., Ltd. were chosen. The chassis dynamometer was made by AVL List GmbH. Under the condition that the wheel and the roller do not slip relative to each other, the vehicle speed was equal to the linear speed of the roller. The roller could handle a range of velocity (0~200 km/h) in both front and reverse directions. Its velocity sensor had a range of accuracy of 0.05-0.1 km/h. Symmetry 2019, 11, x FOR PEER REVIEW 10 of 21

Experimental Process and Results
According to Section 5.1, three experiments were carried out: (1) An experiment to identify dynamic conditions verified the method.  (1) The experiment to identify dynamic conditions. The experiment to identify dynamic conditions used an FFT spectrum analysis to calculate EMI data at multiple frequencies. Then, a distribution diagram of characteristic points was obtained. According to the identification results of different dynamic conditions from speed and acceleration, different condition distribution diagrams of characteristic points were distinguished.
The experiment processes were as follows: The test vehicle was driven on the chassis dynamometer where the wheel and the roller coincide. The computer controlled the pneumatic manipulator to handle the vehicle and controlled the signal analyzer to measure EMI.
A group of EMI test results, based on FFT for new energy vehicles under dynamic conditions, is shown in Figure 9, where Figure 9a is the speed and acceleration curves, and Figure 9b is the EMI intensity graph. According to

Experimental Process and Results
According to Section 5.1, three experiments were carried out: (1) An experiment to identify dynamic conditions verified the method. (1) The experiment to identify dynamic conditions. The experiment to identify dynamic conditions used an FFT spectrum analysis to calculate EMI data at multiple frequencies. Then, a distribution diagram of characteristic points was obtained. According to the identification results of different dynamic conditions from speed and acceleration, different condition distribution diagrams of characteristic points were distinguished.
The experiment processes were as follows: The test vehicle was driven on the chassis dynamometer where the wheel and the roller coincide. The computer controlled the pneumatic manipulator to handle the vehicle and controlled the signal analyzer to measure EMI.
A group of EMI test results, based on FFT for new energy vehicles under dynamic conditions, is shown in Figure 9, where Figure 9a is the speed and acceleration curves, and Figure 9b is the EMI intensity graph. According to Table 1, when the condition a i ≥ a min > a T_acc > 0 is detected, it indicates that the test vehicle is accelerating. When the conditions v idling < v min ≤ v i and a max < −a T_slide are detected, it indicates that the test vehicle is sliding. When the condition a i ≤ a max < a slide − a T_brake < 0 is detected, it indicates that the test vehicle is braking. Based on experience, a T_acc = a T_slide = 0.1m/s 2 , a T_brake = 0.5m/s 2 , and v idling = 10km/h. Then, the vehicle was identified to accelerate during 0~21.3 s, slide during 21.3~58.5 s, and brake during 90.0~96.2 s. It can be seen that the values of a T_acc , v idling , a T_slide , a T_brake would affect the identification results.
EMI data of different conditions were obtained by combining the identification results and the EMI test results, which is shown in Figure 9b. According to Section 4.1, distribution diagrams of the characteristic points in different dynamic conditions are shown in Figure 10. All abscissas are frequency (MHz), and all ordinates are amplitude (dBµV). Figure 10a-c depicts distribution diagrams of characteristic points during acceleration, sliding, and braking, respectively.
Using the identification results to fit the EMI data, any dynamic condition distribution diagram of characteristic points can be obtained. The method to identify acceleration, sliding, and braking was correct and effective. EMI data of different conditions were obtained by combining the identification results and the EMI test results, which is shown in Figure 9b. According to Section 4.1, distribution diagrams of the characteristic points in different dynamic conditions are shown in Figure 10. All abscissas are frequency (MHz), and all ordinates are amplitude (dBμV). Figure 10a-c depicts distribution diagrams of characteristic points during acceleration, sliding, and braking, respectively.   EMI data of different conditions were obtained by combining the identification results and the EMI test results, which is shown in Figure 9b. According to Section 4.1, distribution diagrams of the characteristic points in different dynamic conditions are shown in Figure 10. All abscissas are frequency (MHz), and all ordinates are amplitude (dBμV). Figure 10a-c depicts distribution diagrams of characteristic points during acceleration, sliding, and braking, respectively. Using the identification results to fit the EMI data, any dynamic condition distribution diagram of characteristic points can be obtained. The method to identify acceleration, sliding, and braking was correct and effective. The area with ρ max relates to the values of U T and w. The analyses are as follows: (1) Constant w and changing U T . Based on Figure 10, let w = 0.9 MHz. The relationship between U T and ρ max is shown in Figure 11. Figure 11a-c is the relationship between U T and ρ max under accelerating, sliding, and braking conditions, respectively, when −42.6dBµV ≤ U T ≤ −39.9dBµV. It can be seen that different U T values correspond to different ρ max values. Among the curves in Figure 11, the maximum of ρ max , ρ max|max , was 22.3 pt/MHz/dBµV when U T = −39.9dBµV under sliding conditions, and the minimum of ρ max , ρ max|min , was 4.2 pt/MHz/dBµV when U T = −41.6dBµV under braking conditions. ρ max|max and ρ max|min were largely different. If ρ max was too small, which means that there were few points per MHz and per dBµV, this area had no practical guidance. Therefore, U T should be selected carefully when w is constant.
(2) Constant U T and changing w. Based on Figure 10, let U T = −41.6dBµV. The relationship between w and ρ max is shown in Figure 12. Using the identification results to fit the EMI data, any dynamic condition distribution diagram of characteristic points can be obtained. The method to identify acceleration, sliding, and braking was correct and effective.   It can be seen that different w values correspond to different ρ max values. ρ is an index to evaluate local regions, reflecting the level of EMI generated by new energy vehicles. When w → 0 , ρ max would be much larger but has little practical guidance.
Therefore, it is reasonable that U T = −41.6dBµV and w = 0.9 MHz. The ρ max and the frequency range when U T = −41.6dBµV and w = 0.9 MHz under different conditions are shown in Table 2. The experimental EMI test based on FFT and DFS identifies accelerating, sliding, and braking conditions in real time from the speed and acceleration. Meanwhile, it uses DFS to measure the EMI of f U_max and uses FFT spectrum analyses to measure the EMI of ρ max area. The specific experimental ideas are analyzed as follows.
(1) In the experiment of the EMI test based on FFT and DFS, the FFT frequency span f span = w = 0.9 MHz, the frequency resolution f resolution = 2.5 kHz, and the frequency accuracy f accuracy = ±3.7 kHz. In the dynamic conditions identification experiment, the FFT frequency span f span = 29.85 MHz, the frequency resolution f resolution = 37.3 kHz, and the frequency accuracy f accuracy = ±57.5 kHz. Therefore, by reducing f span , f resolution and f accuracy were greatly improved with the same instruments.
(2) When the vehicle was accelerating, the FFT spectrum analysis was applied from 20.0 to 20.9 MHz with ±1.31 dB amplitude accuracy, and DFS was applied to f U_max = 21.38 MHz with ±0.36 dB amplitude accuracy. When the vehicle was sliding, the FFT spectrum analysis was applied from 21.0 to~21.9 MHz with ±1.31 dB amplitude accuracy, and DFS was applied to f U_max = 22.05 MHz with ±0.36 dB amplitude accuracy. When the vehicle was braking, the FFT spectrum analysis was applied to 21.4 to~22.3 MHz with ±1.31 dB amplitude accuracy, and DFS was applied to f U_max = 21.49 MHz with ±0.36 dB amplitude accuracy.
The experiment of the EMI test based on FFT and DFS was similar to the dynamic conditions identification experiment, with the difference that an EMI receiver was added to measure EMI.
A group of condition identification results of the EMI test based on FFT and DFS is shown in Figure 13. According to Table 1 and experience, a T_acc = a T_slide = 0.1m/s 2 , a T_brake = 0.5m/s 2 , and v idling = 10km/h. Then, the vehicle can be identified that it was accelerating from 0 to 30.5 s, sliding from 30.5 to 84.0 s, and braking from 125.0 to 136.0 s. of U _ max f and uses FFT spectrum analyses to measure the EMI of max ρ area. The specific experimental ideas are analyzed as follows.
(1) In the experiment of the EMI test based on FFT and DFS, the FFT frequency span span 0.9 MHz = = f w , the frequency resolution resolution 2.5 kHz = f , and the frequency accuracy (2) When the vehicle was accelerating, the FFT spectrum analysis was applied from 20.0 to 20.9 MHz with ±1.31 dB amplitude accuracy, and DFS was applied to U _ max 21.3 Hz 8 M = f with ±0.36 dB amplitude accuracy. When the vehicle was sliding, the FFT spectrum analysis was applied from 21.0 to ~21.9 MHz with ±1.31 dB amplitude accuracy, and DFS was applied to U _ max 22.05 MHz = f with ±0.36 dB amplitude accuracy. When the vehicle was braking, the FFT spectrum analysis was applied to 21.4 to ~22.3 MHz with ±1.31 dB amplitude accuracy, and DFS was applied to U _ max 21.49 MHz = f with ±0.36 dB amplitude accuracy.
The experiment of the EMI test based on FFT and DFS was similar to the dynamic conditions identification experiment, with the difference that an EMI receiver was added to measure EMI.
A group of condition identification results of the EMI test based on FFT and DFS is shown in Figure 13. According to Table 1 and experience,   A group of EMI intensity graphs in ρ max area and f U_max amplitude are shown in Figure 14, where Figure 14a-c depicts the EMI intensity graph in ρ max area and f U_max amplitude under accelerating, sliding, and braking conditions.
A spectrogram of EMI test results based on FFT from 21.0 to 21.9 MHz with f span = 30 MHz is shown in Figure 15. A spectrogram of EMI test results based on FFT and DFS from 21.0 to 21.9 MHz with f span = 0.9 MHz is shown in Figure 16. It can be seen that there are 24 points in Figure 15 with 37.3 kHz frequency resolution and 361 points with 2.5 kHz frequency resolution in Figure 16. Therefore, f resolution can be significantly improved by reducing f span .  Figure 15. A spectrogram of EMI test results based on FFT and DFS from 21.0 to 21.9 MHz with span 0.9 MHz = f is shown in Figure 16. It can be seen that there are 24 points in Figure 15 with 37.3 kHz frequency resolution and 361 points with 2.5 kHz frequency resolution in Figure 16.
Therefore, resolution f can be significantly improved by reducing span f .    Figure 16.
Therefore, resolution f can be significantly improved by reducing span f .

Test Results of Other Test Vehicles
In order to prove the applicability and effectiveness of the method on different vehicles, the EMI test method, based on FFT and DFS for new energy vehicle under dynamic conditions, was applied to the electric test vehicles of Zotye Auto Co., Ltd. (hereinafter referred to as "test vehicle #1") and BYD Co., Ltd. (hereinafter referred to as "test vehicle #2"). Figure 17 shows the test sites of the vehicles.

Test Results of Other Test Vehicles
In order to prove the applicability and effectiveness of the method on different vehicles, the EMI test method, based on FFT and DFS for new energy vehicle under dynamic conditions, was applied to the electric test vehicles of Zotye Auto Co., Ltd. (hereinafter referred to as "test vehicle #1") and BYD Co., Ltd. (hereinafter referred to as "test vehicle #2"). Figure 17 shows the test sites of the vehicles.   A group of EMI test results based on FFT for test vehicles under dynamic conditions is shown in Figure 18, where Figure 18a is the result of test vehicle #1, and Figure 18b is the result of test vehicle #2. In order to prove the applicability and effectiveness of the method on different vehicles, the EMI test method, based on FFT and DFS for new energy vehicle under dynamic conditions, was applied to the electric test vehicles of Zotye Auto Co., Ltd. (hereinafter referred to as "test vehicle #1") and BYD Co., Ltd. (hereinafter referred to as "test vehicle #2"). Figure 17 shows the test sites of the vehicles.     The following can be seen from The following can be seen from (2) The point ratio η of test vehicle #1 under braking conditions was 0.90. The point ratio η of test vehicle #2 under accelerating conditions was 0.94. This indicates that there were many characteristic points exceeding the respective U T . As for the comprehensive index to evaluate the overall region, η, it was necessary to pay special attention to the characteristic points under braking conditions of test vehicle #1 and accelerating conditions of test vehicle #2.
(3) The maximum ρ t (29.0~30.0 MHz) of test vehicle #1 was 132.7 pt/MHz/dBµV/min under braking conditions. The maximum ρ t (16.5~17.5 MHz) of test vehicle #2 under braking conditions was 48.9 pt/MHz/dBµV/min. As for the index to evaluate the local region ρ t , it was necessary to pay attention to the concentrated characteristic points between 29.0 and 30.0 MHz of test vehicle #1 and the characteristic points between 16.5 and 17.5 MHz of test vehicle #2 under braking conditions.
Through the EMI test based on FFT and DFS for test vehicles under dynamic conditions, the dynamic conditions can be identified. And the dynamic EMI can be measured with higher accuracy. The harshest cases can be judged from U max . The amplitude range of characteristic points can be judged from η. The regions with concentrated characteristic points can be judged from ρ t . These help us to evaluate and improve the vehicle's EMI performance.

Comparison Among Different EMI Test Methods
The results of different EMI test methods are shown in Figure 21. Figure 21a-d depicts the test results from previous studies [12][13][14] and this paper. We marked the elements with numbers as follows: (1) driving conditions (e.g., speed/torque); (2) frequency span; (3) amplitude; (4) time; and (5) characteristic points. Figure 21a includes driving conditions, amplitude and time, which constitute the result of DFS. Figure 21b includes frequency span and amplitude, which constitute the result of frequency sweeping. Figure 21c includes frequency span, amplitude, and time, which constitute the result of FFT. The test results of this paper shown in Figure 21d include elements (1)-(4). In addition, characteristic points were calculated out for more precise measurement.

Comparison Among Different EMI Test Methods
The results of different EMI test methods are shown in Figure 21. Figure 21a-d depicts the test results from previous studies [12][13][14] and this paper. We marked the elements with numbers as follows: (1) driving conditions (e.g., speed/torque); (2) frequency span; (3) amplitude; (4) time; and (5) characteristic points. Figure 21a includes driving conditions, amplitude and time, which constitute the result of DFS. Figure 21b includes frequency span and amplitude, which constitute the result of frequency sweeping. Figure 21c includes frequency span, amplitude, and time, which constitute the result of FFT. The test results of this paper shown in Figure 21d include elements (1)-(4). In addition, characteristic points were calculated out for more precise measurement.

Conclusions and Prospects
This paper proposes an EMI test method based on FFT and DFS for new energy vehicles under dynamic conditions, which makes full use of FFT's rapidity and DFS's accuracy to effectively expand their applications, making EMI testing under dynamic conditions possible.
Based on the features of accelerating, sliding, and braking conditions, a method to identify accelerating, sliding, and braking conditions was proposed. Different dynamic conditions can be identified from any group of speed and acceleration curves, realizing the automation of identifying dynamic conditions during vehicle the measurement process.
A comprehensive EMI key evaluation index system for new energy vehicles was built, including characteristic points of maximum amplitude p max U max , f suspect_max , area Π, ratio η, and density coefficient ρ of high-amplitude characteristic points. U max is an index to evaluate extreme values. η is a comprehensive index to evaluate the overall region. And ρ is an index to evaluate the local region. The calculation formula of each index is deduced, while the physical significance of each index is expounded.
The EMI test method, based on FFT and DFS for new energy vehicle under dynamic conditions, was applied to GA5 PHEV, produced by Guangzhou Automobile Group Co., Ltd., and the electric test vehicles of Zotye Auto Co., Ltd. and BYD Co., Ltd. The results indicate that the EMI of new energy vehicles can be tested under dynamic conditions with high accuracy, and the evaluation indexes are operable. With the same instruments, by reducing f span from 29.85 to 0.9 MHz, f resolution improves from 37.3 to 2.5 kHz, and f accuracy improves from ±57.5 kHz to ±3.7 kHz. DFS is used in the EMI test with a ±0.36 dB amplitude accuracy.
In the future, more experiments will be carried out to develop EMI test standards and specifications for new energy vehicles under dynamic conditions. Author Contributions: S.Z. and G.L. conceived the study and analyzed the results. S.Z. performed the experiments and wrote the paper.