A Nonintrusive IGBT Open-Circuit Fault and Current Sensor Fault Diagnosis Method for Grid-Tied Three-phase Three-Wire Inverter with Two Current Sensors

In some cases, only two current sensors are available in three-phase three-wire (3P3W) inverters. This occurs when one of three sensors is faulty, or the inverter is equipped with only two sensors to save cost. There has been no method addressing the diagnosis of both IGBT open-circuit faults and current sensor faults in the 3P3W inverter with only two current sensors. In order to solve this problem, a nonintrusive method based on average output voltage deviations is proposed. The deviations between the measured and estimated output line and phase voltages are utilized to detect and locate the fault. With average model, this method can diagnose the fault fast with only signals available in the controller. Therefore, no extra sampling or diagnosing circuits are needed. Besides, the error-adaptive thresholds are adopted to ensure the robustness. Finally, experimental results verified the effectiveness.


INTRODUCTION
Inverters are widely used in renewable energy systems, electrical tractions systems and so on. In inverters, IGBTs are one of the most vulnerable devices [1]. IGBT faults mainly include short-circuit (SC) faults and open-circuit (OC) faults. Both faults may cause the system malfunction. Compared with SC faults, OC faults are less catastrophic and meanwhile are more difficult to be detected. A lot of papers have been published focusing on IGBT OC fault diagnosis. These methods are mainly voltage signal based [2]- [4], current signal based [5]- [7] and model based [8]- [10].
Inverters may also suffer from sensor faults [11]. Sensor faults may cause currents or voltages to rise fast due to closeloop control, which may cause further damages on other devices and loads. Therefore, it is necessary to diagnose sensor fault timely as well. There are some reports on diagnosis of sensor faults. Most of them are based on observers [12]- [14] and current analysis [15]- [16].
However, all these methods mentioned above only consider IGBT faults or sensor fault. The method considering  In recent years, some methods have been developed to address both IGBT fault and sensor fault [17]- [18]. In [17], the current deviations generated by a Luenberger observer are used to diagnose both IGBT OC faults and current sensor faults. In order to improve diagnosis speed, [18] proposes a method based on average bridge arm pole-to-pole voltage deviations. In both methods, the sum of three phase currents are used to distinguish IGBT faults from current sensor faults. Hence, these methods are only suitable for three-phase threewire (3P3W) inverter with three current sensors.
The literature review shows the problem of diagnosing both IGBT OC faults and current sensors fault in the 3P3W inverters with only two current sensors has not been investigated. For the circumstance where only two current sensors are available in the 3P3W inverters, a novel method based on the average output voltage deviations are proposed to diagnose both kinds of faults. The method is nonintrusive, which means the method needs only existing signals in the controller, therefore no extra sampling or diagnosis circuits are added. The method can be embedded in the system easily. Fig.1 shows the grid-tied 3P3W inverter with only two current sensors. In this paper, a inverter with current sensors in phase A and phase B is taken as an example. In the inverter, two phase currents, three phase grid voltages and DC voltage are sampled every switching periods for control. According to the loop shown in Fig.2 and Kirchoff law, the output line voltages vxy (x,y =a,b,c) is

II. FAULTY CHARACTERISTICS OF OUTPUT VOLTAGE DEVIATIONS
Where X,Y = A, B, C. Based on (1), the output line voltages can be estimated as Where ix^ is the sampled phase current, and VXY * is the estimated bridge arm pole-to-pole voltage.

III. PROPOSED FAULT DIAGNOSIS METHOD
The diagnosis principle is shown in Fig.4. The average output line and phase voltage deviations are used for diagnosis.

A. Calculate Average Output Voltage Deviations
Define average model as , , More detailed derivation of the calculation model can be found in [8]. n  are needed to avoid false alarm caused by these calculation error. The error-adaptive method proposed in [8] is applied to determine the thresholds.
The determination of sampling errors and inductance error is explained in [19].
Then, according to the faulty characteristics shown in TABLE I, the criteria for diagnosing IGBT OC fault and current sensor fault are given in TABLE II. In order to further improve robustness against disturbances, the minimum time judging rule is implemented in the method. The fault diagnosis result have to remain for the minimum time Tmin to be considered reliable. Higher Tmin leads to better robustness and meanwhile longer detection time. In this paper, Tmin is set as 2TS.

IV. EXPERIMENTAL RESULTS
In order to verify the effectiveness of the method, experiments have been carried out on OPAL-RT 4510 platform. The experiment specification is shown in TABLE III. The experimental results of T1 OC fault and CSa fault are shown in Fig.5 and Fig.6 respectively.
In Fig.5, before the fault comes, all the diagnosis variables are within thresholds. After the fault is triggered at t1, the voltage deviation polarities (Δab, Δbc, Δca, ΔaN, ΔbN, ΔcN) change to (P, Z, N, P, N, N) soon. According to criteria in TABLE II, the T1 fault is diagnosed at t2.
In Fig.6, after the current sensor fault is triggered at t1, the voltage deviation polarities change to (N, N, P, N, Z, P) soon. According to criteria in TABLE II, the CSa fault is diagnosed at t2.
These experiments verify that the proposed method can diagnose the IGBT faults and current sensor faults effectively. A novel diagnosis method based on average output voltage deviations is proposed in the paper. This method can diagnose both IGBT OC faults and current sensor faults quickly in gridtied 3P3W inverters with two current sensors by utilizing signals available in the controller. The fault analysis, method principle and model calculations are introduced. Finally, the effectiveness of the proposed method is verified by experiments.