Improvements in Micro-Grid Fault Detection

Micro-grids include low voltage distribution systems with distributed energy resources and controllable loads that can operate in medium voltage grid connected mode or in islanded mode. It provides environmental and economic benefits for end-user, customers, utilities and society. However, their implementation creates great technical challenges, such as the protection. Micro-grid system has bidirectional power flows, which make traditional fault insufficient. An algorithm combining of cumulative sum and power flow method is found to be simple and faster for detecting micro-grid faults in both modes of operation... - Vikrant J. Majarikar, K.S. Swarup

A protection system defends the power system from the harmful effects of a sustained fault. Power system protection has main aim to provide maximum sensibility to faults and unnatural conditions and to restrict false alerts during normal state of operations. The protective relays are more of a preventive device which comes in to picture only after a fault has occurred which tries to help in reducing the duration of fault and limiting the damage, outage time and related problems. In general, the first step in the power system relaying algorithms is to detect the faults and the next step is to isolate defected part from the healthy system.

Amicro-gridsystem is a flexible bi-directional powerflow distribution network that is able to suit combination of loads, Distributed Generation Units (DGUs), storage systems like batteries and power conditioning units. This structural characteristic of micro-grid system allows it to function as a single controllable system within its service domain, which generates and distributes electric power to its loads. The bi-directional power flow characteristics of micro-grid systems can be beneficial in providing benefits to utility grid operators and investors, Distributed Generation Units (DGU) owners and customers. Such advantages may include exchanging active and reactive powers, reducing transmission system overloading and improving power transmission and distribution. However, along with these benefits, micro-grids have also aroused an important challenge of protection the provision of rightly coordinated and authentic protection system.

Micro-grid and need of protection

Micro-grids have granted a viable option to achieve increase in power demands by certain load centres through connecting more installed distributed generating units to distribution feeders, rather than expanding present distribution networks. The most positive characteristics of micro-grids are the comparatively short length between the generation and loads and low generation and distribution voltage level.

A micro-grid is compiled of interconnected distributed energy resources (renewable and non-renewable), which are able to provide sufficient and continuous energy to the most part of the micro-grid inner load requirements. A micro-grid must be able to operate both in grid connection mode and in autonomous mode from the utility grid seamlessly with little or no disturbances to the loads within the micro-grid during a disruption. It is a network of small-scale power supply, which is contrived to provide power to a small residential area or small community. The key concept that discerns this approach from a traditional power service is that micro-grids have small power generators and are spread and settled in close proximity to the energy users. This very crucial form of de-centralized electricity supply ensures large environment profits. A typical micro-grid configuration is shown in Fig.1.

Protection issues in micro-grids

Connecting distributed generators to main grid like micro-grid, which can work in stand-alone mode also, changed system properties significantly. Voltage and Current profiles and dynamic demeanor of the whole power system are changed. As a result, the classical protection techniques which are sufficient to detect faults in radial system become inadequate and insufficient. The problems concerns are as follows:

Loss of Selectivity

System protection is more selective when the protective device nearest to the transmission fault is activated as fast as possible to disconnect or isolate the fault from the system. In Micro-grid, the power flow is bidirectional, so traditional over-current protective system becomes inadequate. A possible scenario is the disconnection of a healthy feeder by its own protective relay – because it contributes to the short circuit current flowing through a fault in a neighbouring feeder.

Blinding of Protection

The presence of generator reduces the fault current or over-current detected at the beginning of a feeder or the fault current supplied by other local circuit current, the voltage drop over the feeder section between the generators. Because of the generator’s contribution to the short – and the fault increases, this results in a lower fault current from the grid. If this reduction is sufficiently large, current detected at various points are too low to trigger fast disconnection. This can result in prolonged over-current or earth faults.

Protective Disconnection of Generators

Distributed Generators in micro-grids have to be saved against all types of short-circuits, over- and under- current and voltages, unnatural frequencies, harmonic distortions, etc. Depending on the position of the fault in a micro-grid, the protection system of generators should employ different time delay, which assures the selectivity of the system.

Islanding Mode of Micro-grid Operation

Micro-grid islanding operation through intentional islanding has to be studied as an alternative option, which drastically increases the reliability of system because generators in micro-grid are capable to supply loads in it even if grid is disconnected.

Single-Phase Connection in Micro-grid

Some of the distributed generation units outputs single-phase power in micro-grid system e.g. batteries, small photovoltaic systems or stirling engines. This affects the three-phase current, leading more current in the neutral conductor and stray currents in the earth.

The power output of distributed generators in micro-grids is frequently irregular. Due to this, the behaviour of the system during faults alters constantly. In addition, governable islands of different shape, size and capacity can be formed as a result of faults inside a micro grid. Therefore to deal with bi-directional power flows and low short-circuit current levels in micro grids dominated by micro-generators, a newfault detection method comprising of Cumulative Sum and power flow method is proposed.

Cumulative sum algorithm

The traditional fault detection algorithms mainly use the differential principle for detecting faults – and because of this they are sensitive to noises or spikes or variation of parameters in the signal. To overcome these drawbacks, Cumulative Sum Algorithm (CUSUM), an integral approach is used where adaptive sorts of filters are utilized which offer valuable advantages over traditional detectors. CUSUM method works in time domain and uses sample elements of voltage and current waveform to detect faults. As power signals alternate, the two-sided CUSUM algorithm is appropriately designed for the purpose of fault detection in power system by using the current samples (Sk) of any phase and prepares two complementary signals such as

Using the above mentioned two signals, the two-sided CUSUM test is expressed as:

Where gk ( ) and v are represented by test statistics and drift parameter respectively. A fault is registered if

Where h is an arbitrary constant and which should be ideally zero during normal state of operation.

The max-operation in above mentioned relations provides a positive or zero value for the gk ( ). In the above relation, v provides the low-pass filtering effect which determines the performance of the algorithm. In general, the value of v is little more than amplitude of current signals. With an increased level in the current signal due to transient disturbances, if any Sk > v, the corresponding gk starts growing and if it provides an index higher than the threshold h, a fault is registered. As observed from Fig.2 (a), the gk value increases by a factor of the difference between Sk and v. The two-sided approach fits here as a power system signal alternates and is advantageous from the detection-speed point of view. To understand the principle of above approach, a current signal of 50 Hz frequency and 1kHz sampling rate is used and the behaviour of cumulative sum algorithms is checked under case studies as given in table 1.

The working of the CUSUM algorithm corresponding to the 3 cases is given in Fig 2. From the simulation it is clear that the output is zero during normal sinusoidal wave of the input signal. As the algorithm works on the sampled value at an instant it is immune to frequency variation and by selecting higher threshold value algorithm output is unaffected by noise.

Algorithm implementation

The previous section has shown that the CUSUM algorithm can facilitate detecting transient disturbances into faults and non-faults. It is found that the power output of distributed generators like synchronous generators, induction generators and inverter interfaced protection units is unpredictable due to which whenever there is a fault, power output of these DG sources changes. The ability of the DG units to detect external faults depends strongly on the technology of the generator and the type of fault. In case of Synchronous generator, the initial fault contribution can reach more than six times the generator full-load current and can decay over several seconds below generator full-load current as the generator field collapses. In case of asynchronous generator, the fault is of the order of synchronous generator but for only two to three cycles time period. And in case of inverters based-DG, most of them cannot supply current under external fault conditions; usually no more than 1.2 to 1.5 times of their rated load current. Fault detection schemes using over-current principles, which are universally applied in radial system, are not usually effective in all these power sources.

As traditional over-current fault detection is complex and inefficient for low fault current levels, proposed algorithm is fashioned to detect low level transient current disturbances immediately and isolate the defected region using circuit breaker. CUSUM method is capable to detect lower level faults provided h is set to lower value. A trip signal is initiated if a transient disturbance level is above the acceptable range. Also, it is faster in terms of speed these properties can be used to find faults in presence of asynchronous generators. The concept of Load flow is incorporated along with CUSUM method for accurately locate fault in looped system. The flow chart of the proposed Fault Detection Algorithms is shown in Fig.3.

The steps for detecting faults using proposed algorithm are:

Step 1: Initialize ϑ,gk-1 for Cumulative sum algorithm
Step 2: Data Acquisition of every bus in Micro- grid system
Step 3: Applying CUSUM algorithm for detecting faults in all 3 input current phases.
Step 4: Evaluate the condition gk>h if it is True, check whether it is a load bus else continue step 2.
Step 5: Check whether the faulted bus is load bus or not. (This is done by introducing delay in all generator bus fault                detection blocks to improve selectivity).
Step 6: Tripped that specific load bus which has fault detection signal near it.
Step 7: Monitor the power flows and direction between all generator buses, Pij.
Step 8: Tripped the generator bus which has high amount of power flows in to it.
If Pij> 0, Trip gen i (High fault current at gen i)
Else Trip gen j

Due to bidirectional flow property of micro-grid, the proposed algorithm first detects the fault in micro-grid system later power flow method is utilised to exactly locate the fault. Delay is introduced in the algorithm at generator buses to improve selectivity of the system.

To check the functionality of algorithm it is implemented on a laboratory micro-grid model, which composed of the following components

  • A 208 V, 60 Hz supply (the utility grid);
  • A 2.5 kVA, 4-pole, 208 V synchronous generator (HU);
  • A 1.8 kVA, 4-pole, 208 V synchronous generator (HU);
  • Two 1.8 kW Y-connected resistive loads;

The fault detection blocks were used at all 4 buses to detect transient disturbances in the system. The current through each bus were collected using current sensors.The operating voltage of the laboratory micro-grid was set for one level, which is 208, in order to match the grid voltage. The single line diagram of above 4-bus micro-grid system is shown in Fig 4.

The step by step procedure of the algorithm was realized using Simulink SimPowerSystems tool box. Each fault detection block includes current sensor, which reads samples at a sampling rate of 1 KHz from the current flowing through that bus.

The output signals of current sensor were fed to algorithm which detects the presence of transient disturbances in the system. If due to presence of transients, the current amplitude goes above predefined threshold value, fault is registered. The φ value used in the above test micro-grid system is 10 (110% of highest current value of the system under normal state). So, if current amplitude goes higher than 11 (ideally h = 0, but to avoid noise from the system the value of h is considered as 1) fault is detected. The working of the algorithm is limited to system under study. In order to demonstrate the performance of the proposed algorithm, simulated transient disturbances were investigated in the following section.

Simulation results

Case 1: Single phase fault at bus 4 in islanded mode

A single phase-to-ground fault was created through activating a 1Ø controlled switch to connect phase A to ground point near bus 4 in islanded mode of operation of a test micro-grid system. Due to fault, higher level transient current triggers at all generator buses and a 4th bus which is a load bus as per Fig.5. Now by introducing delay in each protective device near all generator buses, block near load bus will detect the transient current quickly so that the defected region was tripped immediately to avoid further damage to the system. Delay introduction helps to improve the selectivity of the protective system. The simulation test result shows encouraging performance in terms of accuracy, sensitivity, reliability and simple implementation of real time relaying.

Case 2: 3-phase fault at bus 2 in grid connection mode

The 3φ-to-ground fault at terminals of the load at Bus 2 was performed for purposes of investigating the responses of the proposed algorithm to a high level fault current in the grid connected mode. The loads at buses 2 and 4 were being supplied by the generators at bus1 and 3 when a 3Ø–to-ground fault was connected to the ground at Bus 2. The detected 3Ø current signals for each protective device near all buses are shown in Fig. 6.

The simulation indicates that the digital relay at bus 2 only changed the status of its trip signal (from 1 to 0 in less than 3ms after the starting of the fault) in response to the 3Ø-to-Gnd fault at bus 2. Due to incorporation of delay in all protective device near generator buses selectivity of the system improves.

Case 3: 2-Phase fault at bus 1 in islanded mode

The line-to-line fault on the terminals of the generator Bus1 was performed in order to investigate the responses of proposed algorithm to an asymmetrical fault within the domain of the micro-grid in islanded mode of operation. As the generator buses 1 and 3 units were supplying the load buses, phase A was suddenly connected to phase B on the terminals of the generator at bus 1.

The simulation results from Fig. 7 shows that the line-to-line fault at the terminal of synchronous generator triggered high level transient disturbances in 3Ø current at all generator buses. By using the power flow method the exact location of fault is detected as due to fault the power flow was high near defected region. Due to incorporation of delay for improving selectivity of the system, the output trip signal given to circuit breaker near generator bus 1 shows some time lag.

It is found that a fault at any bus cause transient disturbances in overall system. The faults at load bus can be easily detected by using CUSUM algorithm by incorporating some delay at generator bus to improve selectivity and accuracy of the system.

Whereas in case for fault at generator bus the concept of Power Flows is used along with CUSUM method (which detect higher level transient at all gen. buses) to detect exact location of fault. By monitoring theta (θ) values i.e. power flows through all generator buses, we can find the actual location of faults in loop system.

The test result of the fault relaying for micro-grid demonstrated significant abilities to accurately detect, classify, coordinate and responds to different types of transient disturbances.

Furthermore performance result showed high degree of insensitivity to parameters of protected components, harmonics contents and/or micro-grid mode of operation.

In all tested faults, the algorithm was able to detect, locate and clear faults under milli-seconds through disconnecting only the faulty region while maintaining an operating status for the micro-grid.


A simple and fast proposed algorithm consisting of power flow assisted cumulative sum method pose an improvement for detecting faults in MG. A CUSUM based fault detection method alone is insufficient for detecting faults in looped system like micro-grid – so Load Flow theory is incorporated along with it to find the exact location of the fault.

The algorithm is both robust and efficient in fault detection. The simplicity, reliability, accuracy and fast response to transient disturbances of the proposed relaying method comprising of CUSUM and Load flow validate its application for protection of micro-grid and their host utility grid.