The flow chart of the compound criterion is shown in Figure 1. In the power system, no matter the system is working normally or in fault status, KCL is always satisfied. Besides the occurrence of abnormal data, there are two other conditions where KCL may not be satisfied:

1) When there is a fault on the busbar, KCL is not satisfied. The sum of the current phasors equals to the short-circuit current.

2) When the current transformer breaks down, KCL may not be satisfied.

In order to use KCL to detect the existence of abnormal data, it is necessary to overcome the influence of the two conditions above.

In the compound criterion, the status of current transformers and busbar protection status are firstly inspected. If the current transformers work well

and the busbar protection doesn’t start, then the satisfaction of KCL is verified. If the sum of current phasors satisfies KCL, the system unlocks all the protection; if KCL isn’t satisfied, the abnormal data detection algorithm is processed in each line. The system will lock the protection of the line detected with abnormal and unlock the protection of the line without abnormal data.

By this criterion, only the protection on the line with abnormal data is locked. The protection safety and reliability is improved. The abnormal data occurs in a relatively low frequency, so in most cases the judgment system only spends time on verifying KCL. The calculation time spent on verifying KCL is very short, so this criterion works much faster than the traditional criterions which only use abnormal data detecting algorithm, as the criterion in [1] [2] [3]. This is a very big advantage of this compound criterion.

According to the power system theory, no matter the system is in normal operation or fault, the waveform of the current is always continuous. When an abnormal data occurs, there is a mutation on that point and the waveform is not continuous. Using this character, the abnormal data can be distinguished.

Supposing f(x) is the sampled signal, the abnormal data occurs in the n^th point.

1) There is a mutation in the abnormal data, so we have:

2) The waveform is not continuous at the abnormal point, so we have:

and are two threshold values and could be calculated as following:

For a sinuous signal,

So and.We can choose,. is the sample time interval, K₁ and K₂ are two safety factors.

In order to increase the speed of the detection algorithm, for the n^th point, we can use the average amplitude calculated by point n − 1, n − 2 and n − 3 directly instead of calculating the amplitude at the n^th point.

It is very important to know whether the CT (current transformer) is functioning normally in this algorithm. According to the real engineering experience, the CT status can be classified to short-circuit, open-circuit and normal operation and the status is related to the CT current loop impedance. Reference [4] proposes a CT status monitoring method, as shown in Figure 2.

The current loop consists of CT₀ and CT₁. CT₀ is a double-output current transformer. The status of CT₀ can be obtained by measuring the impedance value of CT₁. When CT₀ is in short-circuiting, normal operation and open-circuit status, the impedance of CT₁ changes much, which leads to the output voltage in the secondary side varies differently. The experimental results by [4] is shown in Table 1.

From the experimental result, we can find that there is a big gap of the output voltage among the three statuses. By observing the output voltage, we can monitor the current transformer’s real time status.

In order to sum up the current phasor, it is necessary to measure the current phase. In this paper, the soft phase lock loop (Figure 3) is chosen to obtain the real-time phase as shown in [5].

The three-phase current vector is,

I is the phase-to-ground amplitude of the current,. The phase angle of phase A is, the estimated phase angle of phase A is, is the error of the estimated phase.

After exhausting the synchronous reference frame transformation with chosen as the rotating phase, we can get

and are the components on axis d and axis q.

We can find from Equation (9): if there is no error in the estimated phase, ,; if, the value of is related to the value of. After the lead/leg adjustment component, the error of estimated angular frequency is obtained. By the adjustment of angular frequency, the accurate phase and angular frequency of the current is obtained.