1 Introduction

Modern control techniques and high-sensitive applications promote the use of power electronics components in power systems. This has become more and more popular in recent years. Machinery electrical systems are becoming increasingly complex, as they are composed of microprocessor circuits, power delivery networks, safety control circuits, communication circuits, snubber circuits, sensors, drives, and electronic ignitions. These nonlinear components create harmonic currents that downgrade the power quality. Harmonic distortions cause problems, such as equipment overheating, motor failures, misoperation of protective equipment, inaccurate metering, and sometimes interference with communication circuits. To ensure the power quality now and in the future deregulated industry, harmonic source detection is important [1]. Besides harmful harmonics, the inter-harmonics sources could be more troublesome and harmful than harmonics. In addition to the typical problems caused by harmonics such as overheating and useful life reduction, inter-harmonics create new problems, such as sub-synchronous oscillations, voltage fluctuations, and light flicker, even for low-amplitude levels [2]. Inter-harmonics can be observed in an increasing number of components such as static frequency converters, cycloconverters, sub-synchronous converter cascades, adjustable speed drives for induction or synchronous machines, arc furnaces, and all loads not pulsating synchronously with the fundamental power system frequency. Besides all these components, partial failures such as inter-turn short-circuits cause the presence of inter-harmonics. Therefore, the condition monitoring and detection of the inter-harmonic polluters has become a concern for the utility industry. Identifying the inter-harmonic sources leads to the illustration of the power quality problem, which is utility companies’ concern. To accelerate the mitigation process, practical and reliable methods for identifying the harmonic and inter-harmonic polluters in the system are needed. Many research efforts have been implemented for the characterization and detection of the source of harmonics and inter-harmonics [2,3,4,5,6]. In [2], the principle of the harmonic and inter-harmonic was discussed and the proposed model was suggested.

The assessment of inter-harmonics is considered with particular attention to the problem of frequency resolution and the computational burden associated with the analysis of periodic steady-state waveforms. Finally, modeling of different types of inter-harmonic sources and the extension of the classical models developed for power system harmonic analysis to include inter-harmonics are discussed. The modeling for the assessment of the sources is discussed further in [3]. A new algorithm termed hybrid blind source separation was used for the localization. Although the application of this localization is for high-frequency components, identifying and locating the sources on the basis of predicted measures, which are used in this paper, can be inferred and used for low-frequency power components. Additional work such as [4] and [5] focused on the improving the measurement and extraction techniques to enhance the investigation of harmonics and inter-harmonics. The enhancements were related to higher degree of accuracy, structural/performance robustness, and frequency adaptivity. In [6], simulation and experimental study for the identification of inter-harmonic source location in power systems are implemented. The method was based on the inter-harmonic impedances measured at the metering points and comparing them with the harmonics of the utility system. The main idea behind this method is that the inter-harmonic impedance of the system is much smaller than that of an inter-harmonic-generating load.

In addition to inter-harmonics, other tools were also used for condition monitoring and source identification. In [7], an approximate technique was proposed for the reconstruction of magnetic-field distribution in the proximity of unknown sources based on two nested optimization algorithms. Moreover, in [1], the cascade correlation network was used for harmonic source detection. The current-injection-based harmonic power flow was used to calculate the bus voltages and total harmonic distortion. Srinivasan et al. [8] proposed a neural-network (NN)-based approach for nonintrusive harmonic source identification. The identification was particularly implemented based on the measurement of current at the incoming supply point and comparing the magnitude of different components from home appliances to computer apparatuses and power components. Moreover, other valuable researches have been conducted in this area using different types of filters such as Kalman filter [9], multilevel converter as active filter [10].

In this paper, the nonintrusive condition monitoring of power components such as induction motor, DC motor, synchronous generator and power converter is proposed. The accurate and fast condition monitoring of every component in electric grid increases the reliability and resiliency of power system, which is a major concern nowadays. The magnetic and electric signatures are observed in several distances under different conditions. The components are monitored in healthy and faulty conditions. Several cases of short-circuit conditions in the machine’s winding as well as the unbalanced currents to the system are applied to the machines to differentiate the healthy and faulty components. Moreover, several other conditions are studied. The procedure of the identification method is discussed in Sect. 2. The test procedure and different case studies are covered in Sect. 3. Finally, the neural network as the identification method was used and is explained in Sect. 4.

2 Power component monitoring procedure

The recognition is based on nonintrusive signature identification, without human intervention. The harmonic components of the radiated electric and magnetic fields are measured as the sources of valuable information for signature identification. Then, a processing network was trained to identify the devices from which the harmonics strayed. The overall typical scheme of the studied system with the procedure of the recognition is shown in Fig. 1. As shown, the proposed components consist of the power converter and typical rotating machines such as DC motor, induction motor and synchronous generator.

Fig. 1
figure 1

Scheme of the system and identification procedure

Full size image

The monitoring process is implemented in the following steps. The procedure of these processes is explained in details in the following sections.

2.1 Data acquisition and preliminary processing unit

In order to analyze the signatures which show the characteristics of the components, the radiated magnetic or electric fields are measured using the magnetic coil and the electric rod antennas. These specific antennas can capture low frequencies down to 30 Hz, so the power frequencies are covered. The electric rod antenna is used for the most component identifications and healthy and faulty components detection, because it can capture the electric radiated fields at far distances. On the other hand, the magnetic antenna does not capture the magnetic stray further than 30–40 cm which is because of the weakness of the radiated fields in far distances. Therefore, the magnetic coil antenna is used for more specific investigation such as inter-turn short-circuits.

After capturing the fields with the antennas, the data are sent to the EMI receiver. The technical details of the measurement devices are mentioned in the next section. The frequency responses of the radiated fields are monitored through the EMI receiver. Then, these data are sent to the data acquisition unit for further processing.

The acquired frequency response consists of the main harmonic orders of the radiated fields and the inter-harmonics in between the main harmonics. The main harmonics and critical sub-harmonics orders are picked up for the comparison between cases. The attained data are used in the trained processing unit.

2.2 Trained processing unit

In order to investigate the radiated harmonic and sub-harmonic orders, the physics of them in the machines and converters should be investigated. The comprehensive study about the origin and details of the harmonic sources including machines, transformer, converters and other components is explained in [10, 11]. The related discussion of the harmonics produced by the components is expressed in the following.

The speed of the synchronous rotating field of the stator of an induction motor is the fundamental frequency times the wavelength, i.e., \(f_{1}\lambda \). For a slip s, the rotor speed is thus \(f_{1}\lambda (1-s)\) and the frequency of the rotor currents \(sf_{1}\).

Time harmonics are produced by induction motors as a result of the harmonic content of the field and m.m.f. distribution and are speed dependent.

A harmonic of order n in the rotor m.m.f. (1) has a wavelength \(\lambda /n\); (2) travels at a speed \(\pm (sf)\lambda /n\) with respect to the rotor; and (3) travels at a speed \(f\lambda (1-s)\pm (sf)\lambda /n\) with respect to the stator.

The harmonic respect to the stator induces an e.m.f. in the stator at a frequency equal to the ratio speed/wavelength, i.e.,

$$\begin{aligned} f_{iii}^n =\frac{f\lambda (1-s)\pm sf(\lambda /n)}{\lambda /n}=f(n-s(n\pm 1)) \end{aligned}$$
(1)

The positive sign of the harmonic is due to the opposite direction of the rotor m.m.f to the fundamental. If the electrical asymmetry due to the fault or unbalanced current or any other abnormal reasons occurs, the positive and negative phase sequence currents will flow, giving field forward and reverse directions. The induced field due to these asymmetry travels at speed \(\pm sf\lambda \) with respect to the asymmetric part which can be stator or rotor. The induced frequencies by these fields could be f and \((1-2s)f\). These are the indices in the investigation of the frequency response to detect the faults or abnormal conditions of the induction machines.

The harmonic behavior of the power converter as the purpose of speed drive, pulse width modulation-adjustable speed drives (PWM-ASD), is different in the DC link, supply side, and output side [11]. Therefore, each of them should be considered for the diagnosis purpose. For the DC link, the harmonic is:

$$\begin{aligned} f_{hdc}^i (m_f ,j,r)=h_{dc} (m_f ,j,r)\cdot f_o =\left| {m_f j\pm r} \right| \cdot f_o \end{aligned}$$
(2)

where j is corresponds to the harmonic orders of the inverter, while j and r integers depend on the modulation ratio [12]. The \(f_{o}\) is the output frequency. The dependency from modulation ratio \((m_{f})\) is related to the switching strategy. For the supply side, the harmonic is:

$$\begin{aligned} f_{h s}^i (\nu )= & {} h_s (\nu )\cdot f_s =\left| {(\nu -1)q_s \pm 1} \right| \cdot f_s \nonumber \\ \nu= & {} 1,2,3,... \end{aligned}$$
(3)

where \(h_{s}\) is the order of the supply side harmonic and \(\nu \) corresponds to the harmonic order of the rectifier. The \(f_{o}\) is the output frequency. The \(q_{s}\) is the rectifier number of pulses. Finally, the harmonic for the output side is:

$$\begin{aligned} f_{h o}^i (m_f ,j,k)=h_o (m_f ,j,k)\cdot f_o =\left| {m_f j\pm k} \right| \cdot f_o \end{aligned}$$
(4)

where \(h_{s}\) is the order of the supply side harmonic and k corresponds to the harmonic order of the output side [13].

The harmonics in DC machines as well as synchronous generators are characteristics. The harmonics radiated from the DC motors are mainly from the winding of the commutation side which creates high-frequency harmonics due to the presence of the brushes. As the current to the rotor coils is frequently connected and disconnected to the DC source through the commutator segments, arcing at the brushes is produced as a result of the periodic interruption of the current in the rotor coils (inductors). This arcing has a spectral content [14]. Similarly, the excitation part of the synchronous generator produces the harmonic orders in addition to the harmonic orders of the power frequency. The harmonic contents of the synchronous generator are studied in [15].

All of the above hints help in the differentiation of the components as well as the recognition of their normal or faulty condition. This process is considered as the detection unit which is discussed in the next part.

2.3 Detecting and monitoring unit

After importing the radiated field and extracting the harmonic orders, the harmonics are investigated based on the physics of the components as discussed in Sect. 2.2 and the identification of the type of component is implemented. The flowchart of this unit is shown in Fig. 2.

Fig. 2
figure 2

Flowchart of the component identification

Full size image

The processing unit of each decision block in Fig. 2 was trained based on the characteristics of the mentioned component. The \(f_a\) frequencies, in which the harmonic orders of the DC motor \((h_i)\) appear, can be considered from the first to the higher orders (31th). Similarly, \(\hbox {f}_{\mathrm{b}}\) for the induction motor, \(\hbox {f}_{\mathrm{c}}\) for the converter and the other frequencies for synchronous generator and other components can be considered. It is mentioned in the main decision blocks, shown in colors, that if more than 80–85% of the harmonic orders were located in the right (expected) frequencies, then the component would be identified. The mentioned percentage (P) is because of the possible failures or abnormal condition of the proposed components. That is to say, the occurrence of short-circuits or unbalanced current may lead to the appearance of a new harmonic order or disappearance of a usual harmonic order. Therefore, some harmonic order may be missed and 100% of the harmonic orders may not match the expectance. The 80–85% number is obtained based on the experience. The percentage (P) varies between 80 and 85 due to the power level of the machine and the distance of the antenna to the machine.

Since the converter can consist of the DC-link, inverter, and rectifier, further identification of the specific part can be employed by considering the related Eqs. (2)–(4). Identifying the other components is also possible by locating their charts in the continuation of the main flowchart. So, the extension of main flowchart is based on the number of available components in the system.

Table 1 The characteristic of the components
Full size table

After detecting the type of the components, the condition monitoring in order to find out any failures or faults was also considered. The related flowchart is shown in Fig. 3. As shown in the figure, two symptoms are considered for diagnosing the faults. The first one is the percentage of “the harmonic orders located in the expected frequencies” (P), and the second one is the difference of the amplitude of harmonic orders in comparison with the healthy one \((A_{\mathrm{hi}})\). Both indices are defined based on the characteristics of the components. If any component meets both indices, the component is healthy; otherwise, it is faulty. The type and severity of the fault can be identified by further analysis by looking at the level of deviation from the healthy condition. This can be achieved by analyzing the data using signal processing techniques.

Fig. 3
figure 3

Flowchart of the component identification

Full size image

3 Test implementation and case studies

Following the explanation of the basics and procedure of the study, the setups were implemented through various case studies.

The characteristics of the equipment used in the test through the different case studies are mentioned in Table 1.

The case studies are designed in a way to monitor various examples of power system setups, and they are set from the simplest setup to the setup with the most components. The case studies are implemented in the following six steps:

  • Step 1: \(\hbox {IM}_{2}\) (10-HP induction motor) with and without mechanical load

  • Step 2: \(\hbox {IM}_{1}\) (7.5-HP induction motor) and \(\hbox {IM}_{2}\) (10-HP induction motor) at different locations of antenna

  • Step 3: \(\hbox {IM}_{1}\) (faulty: inter-coil short-circuit), \(\hbox {IM}_{2}\) (healthy) at different locations of antenna

  • Step 4: \(\hbox {IM}_{1}\) (unbalanced current, healthy), \(\hbox {IM}_{2}\) (unbalanced current, healthy) at different locations of antenna

  • Step 5: Converter connected to the \(\hbox {IM}_{1}\) at different switching frequencies at different locations of antenna with and without fault

  • Step 6: Synchronous generator connected to the DC motor and \(\hbox {IM}_{3}\) (3-HP induction motor) at different locations of antenna

Fig. 4
figure 4

Electric radiated field of the step 1 in \(\hbox {dB}\upmu \hbox {V/m}\) (main harmonic orders)

Full size image

3.1 Step 1: \(\hbox {IM}_{2}\) with and without mechanical load

As the first case, the 10-HP induction motor is tested with and without load to figure out the effect of mechanical load on the signatures. The purpose of this step is to investigate the effect of the mechanical load on the harmonic orders.

The test is employed at 180 V with three different loads (0-NM, 3-NM, 10-NM).

As mentioned in the procedure, the harmonic and inter-harmonic orders of these three cases are extracted and shown in Figs. 4 and 5, respectively.

As shown in the Fig. 4, the harmonic orders of the electric radiated field (E-field) slightly increase, while the load is connected. However, the harmonic orders do not change orderly by increasing the amount of load. Therefore, for identifying the motor with load from the motor without load, the study of inter-harmonics is essential.

The inter-harmonics of the setup of step 1 are shown in Fig. 5. The main frequency of the setup is 60 Hz, and the machine (IM2) is 4pole. Based on the fact that mechanical frequency is (2/number of Pole) times the electrical frequency, the mechanical frequency is 30 Hz. Hence, the convolution of the mechanical frequency and electrical frequency yields the inter-harmonics at 90, 210 Hz and so forth which are the 1.5th, 3.5th harmonic orders of the main frequency, respectively. As shown in Fig. 5, the inter-harmonic orders show more changes by applying the mechanical load. Therefore, the algorithm for monitoring the machine with load can be designed in a way that the inter-harmonics compare with and without load. The amplitude of inter-harmonic orders of the machine with load is more than the one without load. Identifying the machine with load helps in the monitoring of the system with several machines to diagnose the faults.

Fig. 5
figure 5

Electric radiated field of the step 1 in \(\hbox {dB}\upmu \hbox {V/m}\) (inter-harmonic orders)

Full size image

Note that the amplitude of harmonic or inter-harmonic orders do not decrease based on the theory of harmonic orders of current/voltage which is because of the effect of the previous harmonics; see Figs. 4, 5. In other words, the previous harmonic orders of a specific harmonic order produce sub-harmonic orders themselves. Therefore, these sub-harmonic orders are being convolved with the harmonic orders of the main frequency and build a higher magnitude than it was expected.

3.2 Step 2: \(\hbox {IM}_{1}\) and \(\hbox {IM}_{2}\) on and off

After recognizing the effect of the mechanical load on the electric signature, the load is removed from the setup and another motor is added. Adding another induction motor with the same number of poles but different power level was the identification of the components with similar operational mechanism but different power level. This is the required step before studying the detection of the faulty machine which is explained the next step.

This test is implemented while the two IM motors were fed with 180 V. The location of the antenna is changed from A to F.

Fig. 6
figure 6

The locations of the antennas and motor in the tests of step 2, 3, 4

Full size image

The machines are tested in all six positions to recognize the effect of each of them at different locations. In addition, the optimum location of antenna for better investigation is the other reason for testing at different locations and height of the antenna. Besides, the machines are switched on and off and located at different locations as shown on the bottom side of Fig. 6. Consequently, the setup was tested in 38 circumstances. The results are applied to the algorithm, and the identification is fulfilled. As an example, the comparison of some of the results in which the antenna was located at point A is shown in Fig. 7.

As illustrated in Fig. 7, the amplitudes of more than 90% of the harmonic orders of the case in which \(\hbox {IM}_{2}\) is switched on and \(\hbox {IM}_{1}\) is switched off are more than those of the case in which \(\hbox {IM}_{1}\) is switched on and \(\hbox {IM}_{2}\) is switched off. The reason is the horsepower of the \(\hbox {IM}_{2}\) is more than \(\hbox {IM}_{1}\). Therefore, higher current passes through the winding of \(\hbox {IM}_{2}\) and consequently the induced voltage is higher. This fact leads to the generation of more electric radiated field. Accordingly, the amplitudes of the harmonic orders of the case in which both motors are switched on are higher than those of the two other cases.

This case along with the other 36 tested cases built an algorithm to identify the larger machine in different locations.

Fig. 7
figure 7

Electric radiated field of step 2 in \(\hbox {dB}\upmu \hbox {V/m}\) (antenna located at point A)

Full size image

3.3 Step 3: \(\hbox {IM}_{1}\) faulty: short-circuited) and \(\hbox {IM}_{2}\) (healthy)

Following the identification of the machine in terms of the power level, the recognition of the faulty machine is investigated while a healthy machine is located very close to the faulty one; see Fig. 6.

Fig. 8
figure 8

Location of the selected turn for inter-coil and turn-terminal short-circuits (\(\hbox {T}_{\mathrm{B}}\): terminal of phase B, \(\hbox {S}_{1}\) and \(\hbox {S}_{2}\): selected turns of phase B)

Full size image

The inter-coil as well as turn-terminal short-circuits are implemented into \(\hbox {IM}_{1}\). The locations of the short-circuited turns are shown in Fig. 8. The \(\hbox {T}_{\mathrm{B}}\) is connected to \(\hbox {S}_{1}\) for the turn-terminal and \(\hbox {S}_{1}\) to \(\hbox {S}_{2}\) for the inter-coil short-circuit. The leads of these points are connected through a rheostat to avoid damages. This test is also done, while the two IM motors were fed with 180 V. The location of the antenna is changed from A to F.

The test cases in this step are implemented similar to the cases in step 2 and the algorithm which is shown in Fig. 3 is used for fault detection. For instance, the antenna is located at point B and the two mentioned types of short-circuits are implemented. The harmonic orders of the radiated E-field are shown in Fig. 9.

As demonstrated in Fig. 9, the harmonic orders of the faulty cases have higher amplitude compared to the healthy one. As the result, \(A_{\mathrm{hi}}\) of Fig. 3 is not equal to \(\hbox {A}_{\mathrm{h-base}}\). Accordingly, the faulty system can be identified. Besides, the identification of faulty component, the type of fault can be detected. Comparing the harmonic orders of the two faulty cases, the harmonic orders of the turn-terminal short-circuit have higher amplitude.

Fig. 9
figure 9

Electric radiated field of the step 3 in \(\hbox {dB}\upmu \hbox {V/m}\) (the antenna is located at point B and both machines are switched on)

Full size image

To identify which of the placed machines are faulty, the level of harmonic-order changes (\(\Delta A_h=A_{hi}-A_{{h-\mathrm{base}}})\) can be compared. Perceiving from the steps 2 and 3, the \(\Delta A_{h}\) while the fault is placed in the larger machine is more. This is discussed more in the next step.

3.4 Step 4: \(\hbox {IM}_{1}\) (unbalanced current, healthy) and \(\hbox {IM}_{2}\) (unbalanced current, healthy)

The other abnormal condition of the electrical machine which is tested here is passing unbalanced currents into the phases of the windings. A resistor box is located in the way of the cables to the machines, and the each three phases of both motors are unbalanced individually. The results are measured similar to the two previous cases as shown in Fig. 6.

Since this step is alike faulty step, the procedure of the detecting the faulty component is similar to the previous one. The \(P_{\mathrm{component}}\) and \(A_{\mathrm{hi}}\) of each case are compared with the healthy case, and the faulty component would be identified. To identify which machine of the system is faulty, the unbalanced current at the phase A of the \(\hbox {IM}_{1}\) and \(\hbox {IM}_{2}\) are compared with each other and the healthy case and displayed in Fig. 10. The result shows that the harmonic orders of both machines cases increase which is due to the increase of current in the machines. Note that the unbalanced current in \(\hbox {IM}_{2}\) has more effect in comparison with \(\hbox {IM}_{1}\). As mentioned before, the reason is the level of the power that \(\hbox {IM}_{2}\) is larger than \(\hbox {IM}_{1}\). Accordingly by comparing the level of changes \(\Delta \hbox {A}_{\mathrm{h}}\), the machine which carries the unbalanced current can be detected.

The figure shows that some harmonic orders may decrease or do not change. For these kinds of cases and especially for the cases with insignificant faults, the other locations of antenna such as points B and E should be tested.

Fig. 10
figure 10

Electric radiated field of the step 4 in \(\hbox {dB}\upmu \hbox {V/m}\) (the antenna is located at point B, both machines are switched on while there is an unbalance current in phase A)

Full size image

3.5 Step 5: Converter connected to the \(\hbox {IM}_{1}\) at different switching frequencies with and without fault

Following the fault and abnormal conditions monitoring, the healthy and faulty conditions of the \(\hbox {IM}_{1}\) motor connected to a multifunctional converter is studied in this case. The converter is controlled through a dSpace connected to the MATLAB/SIMULINK control blocks. Through this, the switching frequency of the converter as well as the frequency of the machine is adjusted. The locations and distances of the components in this step are shown in Fig. 11.

Fig. 11
figure 11

The locations of the antennas and components in the tests of step 5

Full size image

The converter is tested in three switching frequencies (2, 2.5, and 3 kHz) to see the effect of changing the switches in the radiated E-field. Moreover, the frequency of motor is altered from 60 to 59 and 58 Hz.

Figure 12 displays the radiated E-field from the setup while the three switching frequencies (2, 2.5, and 3 kHz) are applied. Since the frequency band is expanded due to the presence of the converter, it’s impossible to demonstrate the comparison of all harmonic orders as shown in previous sections. Therefore, the continuous frequency response is used. As datatips of the figure show, the main frequencies are the power frequency of the machine (60 Hz) and the switching frequency and its harmonics. Therefore, the type of components can be identified. Through this, not only the type of components can be identified, but also the failures or fault can be monitored. For example, if there is any problem in the switches, the peaks at the switching frequency may shift along the frequency band. Moreover, if the slip of the induction motor changes, the harmonic orders of the machine may not be located at the expected frequencies.

Fig. 12
figure 12

Electric radiated field of the step 5 in \(\hbox {dB}\upmu \hbox {V/m}\) (the antenna is located at point B)

Full size image

The short-circuit which is employed in step 3 is applied again to the \(\hbox {IM}_{1}\) and the effect of the faults on the radiated E-fields at the switching frequencies are shown in Fig. 13. It is shown that the fault in one component, IM1 in this case, may not affect the other components significantly; however, the peaks at the switching frequency and its harmonics are changed marginally. Consequently, the faults of the inverter can be investigated without significant interference of the other components.

Fig. 13
figure 13

Electric radiated field of the step 5 in \(\hbox {dB}\upmu \hbox {V/m}\) while the IM1 is faulty (the antenna is located at point B)

Full size image

3.6 Step 6: Synchronous generator connected to the DC motor and IM3 (3-HP induction motor)

As the last case study, the synchronous generator is connected to a DC and induction motors to identify the different type of machines. The purpose of this step is to identify the different type of machines. The situation of the machines and antenna is shown in Fig. 14.

Fig. 14
figure 14

The locations of the antennas and components in the tests of step 6

Full size image

As demonstrated in Fig. 14, the generator and motors are not at the same height. Similar to the previous cases, the radiated fields are measured with several combinations of these components. Three cases are selected among numerous tests to be shown in Fig. 15. As displayed, the harmonic orders of the case in which the generator is connected to the DC machine (GEN + DC) are considerably different from the other two cases in terms of amplitude. This is due to the difference between the structures of the DC motor from the induction motor. The presence of four types of winding in DC motor such as field, armature, compensation and commutation winding creates different amplitude of the radiated E-field. It is also shown that the presence of the all machines in one of the test cases would lead to the highest amplitudes of the harmonic orders. However, the difference between the case GEN+IM3 and GEN+IM3+DC is less which is due to the existence of two similar machines, generator and induction motor, in terms of working frequency.

Fig. 15
figure 15

Electric radiated field of step 6 in \(\hbox {dB}\upmu \hbox {V/m}\), with antenna at point A

Full size image
Fig. 16
figure 16

Electric radiated field of step 6 in \(\hbox {dB}\upmu \hbox {V/m}\), with antenna at point B

Full size image

In addition to the difference of amplitudes between the above cases, the location of the harmonic orders along the frequency band should be compared as mentioned in the main flowchart. Figure 16 shows the frequency response of the above cases measured at point B. The amplitudes of the 12th harmonic order in the three cases are shown in the datatips. As revealed, the harmonics of the first two cases appeared at 727.5 Hz instead of 720 Hz which is because of the presence of the induction motor in these two cases that have slip frequency. The slip frequency shifts the rotors frequency which yields to the shift of the harmonic orders along the frequencies. This is one of the main hints in identifying the type of machines.

4 Conclusion

The identification of the types of power system components is implemented in this paper. The electromagnetic radiated field of the component was used for the study. An algorithm is designed for the identification and monitoring. More than 150 circumstances of the combinations of the typical power components are tested experimentally, and the identification is explained. The application of the identification, the fault and failure detections are also experimented and studied.

The results show that it is possible to identify the type of components as well as the faulty components by comparing the amplitudes of their harmonic orders as well as of the location of harmonic orders along the frequency band. This comparison can be processed through the explained flowcharts. The identification using the radiated fields is nonintrusive and can be used for the setups that cannot go offline and dismantled.