Table 7 Summary of protection schemes for AC-MGs
Protection Scheme | Operation description | Merits | Demerits | Â | Â |
|---|---|---|---|---|---|
Traditional approaches | Adaptive protection | Predefined group settings are included in digital relays to handle different configurations. Intelligent controllers can be employed in centralized or decentralized frameworks | Adapt automatically to any changes and conditions Relay settings are updated via external signals Setting groups can be calculated online or offline | Requires communication infrastructure Needs a huge database of simulated topologies High implementation costs (digital relays, controllers, and communication links) | Â |
| Â | Differential protection | Compares electrical quantities at input/output terminals of protected facility | High selectivity and sensitivity Detects high impedance faults Simple and high computational efficiency | Requires communication infrastructure Current transformers saturation or mismatch False activation on heavy external faults | Â |
| Â | Distance protection | Apparent impedance to fault point is defined based on voltage/current measurement at relay location This impedance identify fault and its location | Based on local measurement Simple implementation | Errors to fault resistance, lines loadability, and infeed currents Depends on line parameters Time consuming Fundamental components extraction | Â |
| Â | Overcurrent protection | Traditional | Trip decision based on current magnitude comparison with the pickup value | Simple Low cost | Bidirectional flow of fault current Low fault current contribution of inverter-based DERs Influenced by operating mode High impedance faults |
| Â | Â | Directional | Trip decision based on current magnitude comparison with the pickup value and fault direction (forward/reverse) | Simple Addresses bidirectional flow of currents | Low fault current contribution of inverter-based DERs Influenced by operating mode High impedance faults |
| Â | Voltage protection | Depends on voltage level comparisons to decide faults | Simple Low cost | High impedance faults Influenced by operating mode and system configuration Difficult discrimination of voltage sag in fault and normal events | Â |
Signal processing-based approaches | Wavelet transform | System signals are transformed into time–frequency domain to extract fault related features for further fault identification | Dependable and secure Adjustable data window for signal processing | Needs classification learning models High cost implementation Impacted by signals noise Requires high capability software |  |
| Â | Travelling waves | Based on analysis of induced electromagnetic waves at faults | High accuracy Independent of network data | Un required reflections due to laterals Complex implementation More expensive High sampling rate of fault recorders | Â |
|  | S-transform | System signals are transformed into time–frequency domain to extract fault related features for further fault identification | Dependable and secure Adjustable data window for signal processing | Needs classification learning models High cost implementation Impacted by signals noise Requires high capability software |  |
|  | Hilbert–Huang | A time–frequency signal processing approach that computes instantaneous frequency signal of input data to be compared with a threshold value | Dependable and secure Adjustable data window for signal processing | Needs classification learning models High cost implementation Impacted by signals noise Requires high capability software |  |
| Â | Harmonic contents | Depends on harmonic content of output currents/voltages due to inverter-based DERs | Simple Mimics traditional networks in the higher-frequency domain | Dependent to network configuration and penetration level of DERs High impedance faults Inaccurate in a harmonic rich system | Â |
Knowledge-based approaches | Artificial Neural Network | System signals, relays history, breakers status for offline trainings to define faults | Addresses data uncertainties Simple and feasible implementation | Time consuming during training Requires wide-range of data for training Requires high capability software | Â |
| Â | Fuzzy logic | System voltages/currents are processed using fuzzy model to decide faults | Fast and simple Absence of training Addresses data uncertainties | Low accuracy Dependent to network configuration | Â |
| Â | Decision trees | Classification and regression model requires data for training to decide abnormalities | Easy to grasp Clear Visualization Handles outliers and missed values | Consumes more memory High computational burden Influenced by noise Overfitting and limited for small database | Â |
| Â | Support vector machine | Classification and regression model requires data for training to decide abnormalities | Regularization capabilities, thus no overfitting Stable against data variation | Consumes more memory High computational burden Difficult to understand | Â |
Multiagent-based approaches | Comprises three layers of different responsibilities: equipment/substation/system | Layers-based scheme Simple, reliable, and flexible Easy to understand | Requires reliable communications High impedance faults Limited to small-scale systems | Â | Â |
External helping devices | Fault current limiters | Inserts a series impedance to limit fault current | Simple Fast response | High installation and maintenance costs Challenges regarding their size, location, parameters tuning Type selection and associated limitations | Â |
| Â | Energy storage systems | Additional storage is provided to support the short circuit capacity to be sensible by relays | Simple Easy implementation | High installation and maintenance costs More suitable for islanded modes with inverter-based DERs Needs islanding detection schemes | Â |
| Â | Intelligent electronic devices | Voltage and current data are monitored at different locations to decide faults | Simple implementation | Requires reliable communications Time consuming when combined to learning classifiers High cost Low accuracy and sensitivity | Â |