Communication-less ensemble classifier-based protection scheme for DC microgrid with adaptiveness to network reconfiguration and weather intermittency
Introduction
In recent years, the soaring dissemination of Distributed Energy Resource (DER) systems such as photovoltaic (PV), fuel cells and energy storage devices coupled with the wider use of DC loads such as Electric Vehicles (EVs), light-emitting diode (LED) lights and electronic gadgets, has led to the adoption of DC microgrids [1], [2], [3]. As compared to AC microgrid, DC microgrid involves a reduced number of power conversion stages, thereby minimizing the losses and complexity associated with the design and control of power converters [4], [5], [6], [7]. In spite of the advantages, the wider acceptance of DC microgrid has been hindered by the complexity of designing a reliable protection scheme.
The major challenges in DC microgrid protection concern the substantial (8–10 times) and sudden increase in fault current, grounding issues, absence of zero-crossing and high possibility of sustained arcs [8]. As compared to AC microgrids, the protection of DC microgrids demands higher level of reliability and rapidity of response. Achieving the advantages of DC microgrid quite often demands reconfiguration of the microgrid topology through opening/closing of switches leading to outage/connection of the DERs. The reconfiguration leads to a dynamic microgrid, with alterations in the connection of DERs as per the operating condition and load-demand response. The dynamic alterations in the network topology add to the complexity of the protection task. Thus, for reconfigurable DC microgrids, the relay must be insensitive towards operating scenarios involving unnecessary connection/disconnection of DERs and load changes. Also, the stochastic behaviour of weather dependent DERs like solar and wind sources significantly impacts their power output, thereby causing fluctuations in voltage and current profile in the power distribution network [9], [10], [11]. The conventional threshold-based relays may malfunction during the unanticipated stochastic variations in voltage and current magnitude arising because of weather intermittency.
In the last few years, researchers have shown increased interest in the design of effective protection schemes to overcome the challenges associated with the detection and classification of faults in DC microgrid. In this context, some of the significant reported works include wavelet transform based efficient protection scheme for low-voltage DC microgrid [5], overcurrent protection [12], protection of smart DC microgrid with ring configuration using parameter estimation approach [4], hierarchical protection method [13], centralized unit base protection scheme [14], superimposed current based unit protection scheme [15], non-unit protection [16], travelling wave-based methodology [17], non-iterative fault location scheme [18], cumulative sum based fault location [19] and adaptive thresholding [20]. Unit protection schemes involving communication networks have been reported in [4], [21], [22]. A pictorial representation of the notable reported techniques and challenges/issues in DC microgrid protection is given in Fig. 1. These schemes rely on the rate of change of line current and estimation of cable parameters using the data transmitted through the communication links. Communication-based techniques require either data synchronization or a high bandwidth network. In spite of their effectiveness in providing fast and reliable protection, the performance of communication assisted protection scheme are compromised during non-synchronizing of sensor data, communication failure and data loss. In this regard, local measurement-based backup protection schemes have been proposed [23], [24], [25] to address the issues related to data loss, latency and communication link vulnerability.
The review of literature reveals that majority of the reported techniques fails to simultaneously address the following issues:
(1) Adaptability to the varying configuration of microgrid.
(2) Concurrently imparting a high degree of robustness against operating mode, loading condition and possible DER outage(s).
(3) Immunity to weather based stochastic variation in DER dynamics.
(4) Operate satisfactorily under communication network failure and sensor data loss.
(5) Performing the multiple tasks of fault detection, fault classification and faulty section identification.
Motivated by the need to develop a protection scheme, capable of performing reliably under varying configuration (including DER outage(s)) of DC microgrid with reduced sensitivity to variation in DER output (due to weather intermittency), communication link failure, operating mode (grid connected/islanded) and loading scenarios, the present work proposes a scheme based on the combined framework of online topology identification, weather dependent DER output modelling and ensemble of classifier.
Reduced sensitivity of the protection scheme to changes in the operating scenario can be achieved by providing adaptiveness in the threshold settings of current and/or voltage [26]. However, for microgrids, which are characterized by a high uncertainty in configuration/topology and DER operation, adaptive schemes may fail to provide reliable protection because of the non-inclusion of all possible dynamic variations in the relay setting [27]. In this context, in the proposed work, the uncertainty in network topology has been addressed by online identification of the topology using local measurements. The use of local information avoids retrieving data through high bandwidth communication channels, thereby avoiding the issues associated with data loss and latency. Further, the influence of weather intermittency in the relay setting has been addressed by modelling the randomness in solar irradiance and wind speed using a probability distribution function. The model allows for compensating the impact of weather-dependent variation in DER output on the performance of the protection scheme. In [8], [9], the stochastic modelling of weather intermittency has been carried out for AC microgrid with fixed topology.
The proposed scheme initiates the identification of the network in real-time from the local current and voltage information. After the identification of the network topology, a set of classifiers is designed to perform the protection tasks of fault detection, classification and section identification. For each of the configuration resulting from the outage of a particular DER, a separate classifier is designed using the simulated fault data generated while considering the weather intermittency (as reflected in the stochastic model). Prior to generating the dataset, the probability distribution functions replicating the randomness in solar irradiance and wind velocity has been derived from the time-series dataset. In literature, works have been reported on the use of the various classification techniques for AC microgrid protection [8], [9], [28], [29]. Majority of these works are based on the extraction of discriminatory features from the raw time domain data. However, the unidirectional and stationary nature of voltage/current in DC microgrid allows for directly feeding the raw data to the classifier. To avoid possible the biasness of an individual classifier towards a particular class, an ensemble of classifier approach has been adopted in the present work. In the ensemble scheme, the same input data is processed simultaneously by multiple classifiers and the final output is derived based on a voting strategy. The involvement of multiple decisions followed by polling, results in an output which is either equal to the most accurate classifier or (in majority of the cases) better than the output achieved with each of the standalone classifier.
The proposed protection scheme has been validated for diverse faults, evolving topologies, weather intermittency and DER operating scenarios under both modes (grid connected and islanded) of microgrid operation. Further, the proposed scheme has been compared with other classifier based protection schemes in terms of different statistical indices.
The major highlights/contributions of the proposed work can be summarized as:
- (1)
Development of a local measurement-based protectionscheme for reconfigurable DC microgrid under varying operating scenarios.
- (2)
Formulation of the fault detection/classification and faulty section identification task as a multi-class classification problem and further solving it using an ensemble of classifier approach.
- (3)
Development of a scheme for online identification of network topology and DER outage.
- (4)
Implementation of a protection scheme with immunity to network reconfiguration and weather intermittency using a joint probability model.
- (5)
Performance assessment of the proposed scheme under varying stressed scenarios of the microgrid and network reconfiguration along with comparison with the existing schemes.
Section snippets
DC test microgrid system under study
For the present study, a 350 V, 500 kW DC microgrid system with ring topology has been considered (Fig. 2) [4]. The entire microgrid extends over a length of 6 km with six sections of each 1 km. The overall system consists of three DERs (PMSG wind generator of 500 kW, photovoltaic array of 125 kW, and synchronous generator of 186 kVA), an energy storage element and AC and DC loads. The DC microgrid model simulated using MATLAB/Simulink includes all the major DERs (i.e. synchronous generator,
Probabilistic modelling of PV and wind intermittency
As outlined in the introduction section, in addition to an outage, the intermittent behaviour of PV and wind source also has a significant impact on the operation of protective relays in the microgrid. The false operation of traditional predefined threshold setting based overcurrent and distance relays because of fault like voltage and/or current waveform during weather intermittency demands adaptiveness with the prevailing weather scenario. The impact of weather intermittency on relay
Local information and ensemble classifier based DC microgrid protection
As discussed earlier, the distinct level of fault currents contributed under complete DER penetration and varying levels of DER outage demands the development of a robust protection scheme, which should be able to detect/classify and identify the faulty section irrespective of the outage in any DER. Further, fault levels are also dependent on the variation in wind speed and solar irradiance during weather intermittency. In this context, the present work aims at the development of local
Performance analysis
In this section, the effectiveness of the proposed protection scheme has been evaluated for different faults scenarios using the local information at bus B. The validation has been carried out to evaluate the ability of the ensemble classifier-based approach to adapt itself as per the changes in the network configuration and weather condition. In this regard, different fault scenarios involving wide variation in the fault parameters have been simulated under DER outages while varying the wind
Conclusion
The variability in the operational dynamics of a DC microgrid arising due to the variation in the network configuration and weather condition increases the complexity of the protection task. The scenario is further challenged during data loss and/or delay in the communication channel, used for retrieving the sensor information from different locations of the microgrid. In this regard, in the present work, a communication-less DC microgrid protection scheme with robustness against network
CRediT authorship contribution statement
Shankarshan Prasad Tiwari: Data curation, Formal analysis, Investigation, Methodology, Software, Validation, Visualization, Writing - original draft. Ebha Koley: Conceptualization, Investigation, Methodology, Supervision, Validation, Visualization, Writing - original draft, Writing - review & editing. Subhojit Ghosh: Conceptualization, Supervision, Writing - original draft, Writing - review & editing.
Declaration of Competing Interest
The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.
References (33)
- et al.
Comparison of integer and fractional order robust controllers for DC/DC converter feeding constant power load in a DC microgrid
Sustain. Energy, Grids Netw.
(2017) - et al.
Procedures for fault location and isolation to solve protection selectivity problems in MV distribution networks with dispersed generation
Electr. Power Syst. Res.
(2009) - et al.
Microgrid protection under wind speed intermittency using extreme learning machine
Comput. Electr. Eng.
(2018) - et al.
Control techniques in AC-DC and hybrid AC-DC microgrid a review
IEEE J. Emer. Sel. Top. Power Electr.
(2018) - et al.
Sustainable and reliable hybrid AC/DC microgrid planning considering technology choice of equipment
Sustain. Energy, Grids Netw.
(2020) - et al.
Protection of smart DC microgrid with ring configuration using parameter estimation approach
IEEE Trans. Smart Grid
(2018) - et al.
Efficient protection scheme for low voltage DC micro grid
IET Gener. Transm. Distribution
(2018) - et al.
Efficient and seamless power management of hybrid generation system based-on DFIG wind sources and microturbine in DC microgrid
Sustain. Energy, Grids Netw.
(2020) - et al.
DC microgrid small-signal stability and control, sufficient stability criterion and stabilizer design
Sustain. Energy, Grids Netw.
(2021) - et al.
DC microgrid protection a comprehensive review
IEEE J. Emer. Sel. Top. Power Electr.
(2019)