State Inference and Evaluation Technology of Relay Protection Based on Digital Twinning

The problems of relay protection equipment in such aspects as the insufficient depth of data mining, not timely grasp of equipment status, and poor timeliness of fault warning restrict the development of the operation and maintenance mode of relay protection, which aims at “autonomous and controllable, safe and reliable, advanced and applicable, intensive and efficient”. This article studies the comprehensive state evaluation and action simulation deduction technology of intelligent substation relay protection equipment based on digital twins. It uses multi-source state monitoring data of relay protection equipment for state quantity correlation analysis. It constructs a key state feature index library and uses digital twin technology to model the relay protection function and software and hardware states. Through the collaborative linkage between physical devices and digital twin models, the twin can dynamically track the operational status of devices and automatically update the twin status. Finally, based on the established high-precision digital twin model, relay protection action deduction and status evaluation are carried out. Then the effectiveness of the scheme is demonstrated through example testing of the relay protection twin system.


Introduction
With the continuous construction of a new generation of autonomous and controllable substation secondary systems, relay protection equipment is gradually optimizing its operation and maintenance mode with the goal of "autonomous and controllable, safe and reliable, advanced and applicable, and intensive and efficient".To comprehensively improve the centralized monitoring ability of relay protection intelligent remote inspection system on the whole life cycle state of relay equipment and to create a relay "digital team" suitable for the modern power grid, it has become the basic link of building a new mode of intelligent operation and maintenance of relay protection.However, there are still some problems in the deep mining of operation and inspection data, centralized monitoring, and early warning of the status of relay protection equipment, which does not meet the requirements of lean equipment management.In addition, the traditional operation and maintenance methods of power plants and substations are difficult to accurately evaluate the status of equipment, and the timeliness of misoperation warnings and the automation level of fault tracing mechanisms are low [1][2].
Many experts and scholars have carried out relevant research on how to strengthen the data processing and intelligent analysis ability in the substation and realize the fault warning.Liu et al. [3] propose a new method of online monitoring of relay protection based on the characteristics of a smart substation system; they establish a fault information model of the smart substation, automatically identify the key state of relay protection, and realize fault location of relay protection.Xu et al. [4] use graph theory to establish a connection matrix to describe the actual connection relationship of physical devices, analyze the causal relationship of various devices in actual cases, and obtain improved matrix criteria to quickly search and locate faults.Dai et al. [5] establish a probability model for the misoperation and rejection of bus differential protection under-sampling synchronization error to evaluate the impact of sampling synchronization error on bus differential protection.Researchers in [6][7][8] respectively study the influence of various failure factors, such as human factors, external environmental factors, and system factors, on the reliability of the protection system.The existing research mainly relies on relatively static data.The timeliness is weak and is not conducive to the emergency treatment of the relay fault and other problems.
Based on digital twin technology, a set of physical information systems can be constructed which can keep synchronization with physical objects and carry out a parallel deduction.It is of great significance to improve the visibility of equipment running state, the ability of equipment operation and maintenance management, and the ability of equipment failure mechanism analysis.It has been widely used in aerospace, intelligent manufacturing, construction, and other industries [9][10].It has been widely used in aerospace, intelligent manufacturing, building construction, and other industries.In this paper, the state inference and evaluation technology of relay protection based on digital twinning is studied.Multi-source state monitoring data of relay equipment are used for state quantity correlation analysis, and a key state characteristic index database is constructed.Twin modeling of relay protection function and software and hardware state is carried out based on digital twinning technology.Through the cooperative linkage between the physical device and the digital twin model, the dynamic tracking of the operating state of the device and the automatic updating of the state of the twin are realized.Finally, based on the established high-precision digital twin model, the secondary physical equipment is monitored, and the closed-loop simulation analysis is carried out in the virtual space.Through the established health status evaluation system, the operation status of the existing secondary equipment is monitored and evaluated.

Construction of relay digital twin system
Based on the multi-source state monitoring data of relay protection and the objective electrical principle, the key state characteristic index database of relay protection equipment is constructed.Taking the fully independently developed domestic relay equipment as the physical mapping object, the twin model of relay protection function logic and hardware and software running state of 220 kV new-generation smart substation and the twin model of the external operating system of smart substation are established, as shown in Figure 1.The digital twin system of relay protection is built in a Simulink time domain simulation environment.It is designed according to the relevant specification of protection communication in the IEC61850 standard and the action logic of physical equipment.The system designs virtual and real data interaction interfaces based on XML format to receive physical device protection fixed value, control word, platen fixed value, and device parameters in real-time so as to improve the operation logic.
The relay protection twin system includes a protection logic function simulation subsystem and software and hardware running state twin simulation subsystem.This article takes the bus differential protection subsystem as an example to introduce the logic function simulation process.The bus differential protection subsystem includes voltage power frequency change starting element, current power frequency change starting element, differential current starting element, conventional ratio differential element, power frequency change proportion differential element, fault bus selection element, CT saturation element, voltage locking element, and charging locking element.
The voltage power frequency variation starting element acts when the voltage power frequency variation of any phase voltage at both ends of the bus is greater than the threshold, and its criterion is shown in Equation (1). 0.05  is the instantaneous value of phase voltage power frequency variation, 0.05 N U is the fixed threshold, is the floating threshold, which adjusts itself as the output of the variable quantity changes.
The current power frequency variation starting element acts when the braking current power frequency variation is greater than the threshold, and its criterion is shown in Equation (2).0.05  is the instantaneous value of braking current power frequency variation, 0.5 N I is the fixed threshold, T SI  is the floating threshold, which adjusts itself as the output of the variable quantity changes.
The differential current element acts when the differential current of any phase is greater than the differential current starting value, and its criterion is shown in Equation (3).
In Equation (4), K is the ratio braking factor, j I is the current of the j-th connector, cdzd I is the differential current starting value.In order to prevent the sensitivity of the large differential ratio differential element from being insufficient in the event of a fault on the bus on the weak power side when the busbar switch is disconnected, the ratio braking coefficient of the proportional differential element is set to two fixed values: high and low.The high value of the large difference is fixed at 0.5, the high value of the small difference is fixed at 0.6, the low value of the large difference is fixed at 0.3, and the low value of the small difference is fixed at 0.5.When the large difference high value and small difference low-value act simultaneously, or when the large difference low value and the small difference high-value act simultaneously, the proportional differential element acts.
In order to improve the ability of protection equipment to resist transition resistance and reduce the influence of the power angle relationship before failure, the differential current is used to form the conventional proportional differential element, and the power frequency variation current is used to form the power frequency variation proportional differential element.The power frequency variation proportional differential element and the conventional proportional differential element with a fixed braking coefficient of 0.2 are combined to form a fast differential protection, and its criterion is shown in Equation ( 5).
In Equation ( 5), j I  is the power frequency variation current of the j-th connecting element, is the floating threshold for differential flow initiation, and cdzd DI is the fixed threshold for differential flow initiation, determined by cdzd I .' K is the proportional braking coefficient of power frequency variation.To solve the sensitivity problem of braking coefficient under different main wiring methods, the ratio braking coefficient is set to two fixed values: high and low.The high value of the large difference is fixed at 0.65, the high value of the small difference is fixed at 0.65, the low value of the large difference is fixed at 0.3, and the low value of the small difference is fixed at 0.5.When the large difference high value and small difference low-value act simultaneously, or when the large difference low value and small difference high-value act simultaneously, the power frequency change proportional differential element acts.
The system calculates the large difference current based on the current sampling values of all connected components on the bus, forming a large difference ratio differential element as the fault discrimination element in the differential protection zone.The system calculates the small difference current of each busbar based on the switch position of each connecting component, forming a small difference ratio differential component as the fault busbar selection component.
When the large differential anti-saturation bus differential acts and any small differential ratio differential element acts, the bus differential protection acts and trips the contact switch of the relevant bus.When the small difference ratio differential element and the small difference harmonic braking element are simultaneously opened, the bus differential acts and trips the corresponding bus.
When the large differential anti-saturation bus differential acts and there is no bus trip, two times limit backup protections are set to prevent the protection from rejecting.The first-time limit backup protection cuts off the branches with current and no switch position input, as well as the busbar (segmented) switch with voltage lockout open.The second time limit backup protection cuts off all branches with a current greater than 2 N I to minimize the impact on the system and minimize unnecessary cuts.In addition, the system will exit all anti-saturation measures after the ratio differential continuous action for 500 ms; only the ratio differential is retained.If its action still does not return, the corresponding bus will trip.
The CT saturation detection element is designed to prevent misoperation of the bus protection under the condition that the CT is heavily saturated when an out-of-area fault occurs near the bus.According to the characteristics of the CT saturation waveform, two CT saturation detection elements are set in the twinning system to judge whether the differential current is caused by CT saturation due to out-of-area fault.If it is, the differential protection outlet is closed; otherwise, the protection outlet is opened.The first CT saturation element uses the adaptive impedance weighted anti-saturation method, and the second uses the harmonic braking principle.This configuration makes it highly resistant to CT saturation, and it can act quickly in the case of in-zone fault, general conversion fault, or the same phase conversion fault after CT saturation of out-of-zone fault.
The voltage locking element criterion is shown in Equation (6).
In Equation ( 6), U  is the phase voltage, 0 U is self-generated zero-sequence voltage, 2 U is negative sequence voltage, bs U is phase voltage locking value, fixed as 0.7 N U , 0bs U and 2bs U are zero sequence voltage locking value and negative sequence voltage locking value, fixed as 6V and 4V, respectively.When any of the above three criteria are met, the voltage-blocking component opens.
In order to prevent the bus from tripping due to dead zone faults during charging, when the busbar switch is detected to switch from 0 to 1 in the charging standby state, the differential tripping bus will be locked within 300 ms from the start of the large differential current.The differential tripping busbar (segmented) will not be delayed.After the TWJ of the busbar returns for more than 500 ms or the busbar switch is turned upside down for 1 s, the bus differential function returns to normal.In addition, if there is a busbar flow or the busbar split operation pressure plate is put in during the charging, it indicates that the fault is not charging to the dead zone, and the delay of bus tripping is immediately lifted.
The design logic diagram of the bus differential protection simulation subsystem is shown in Figure 2. The modeling principle of other subsystems is similar to that of the bus differential protection subsystem, which will not be described here.
The input of the busbar brake changes from 0 to 1 The busbar TWJ changes from 0 to 1  In the model, the hardware and software state twinning is based on the objective principle of the actual operation of the equipment to establish the logical model of analogy input acquisition loop, switching input-output loop, optical fiber link, CPU, DSP, etc. Through the mapping with the hardware and software loop state of physical equipment, the hardware and software state of physical equipment are simulated.If the software and hardware running status are abnormal, the specific logic loop or module will be interfered with or locked, and an alarm signal will be sent.It is made sure to present the real situation of a physical entity in the digital world.
The operating environment simulation subsystem can simulate and output various kinds of power grid faults, including the fault inside and outside the area, permanent and developmental faults, transfer faults inside and outside the area, short circuits through the transition resistance, system oscillation, system frequency deviation, and so on.The modeling method is consistent with the above bus protection modeling method.

Relay protection twin model status tracking and updating
Digital Twin builds a digital simulation system in the information platform by integrating physical feedback data, supplemented by artificial intelligence, machine learning, and software analysis.This simulation system will automatically make corresponding changes with the changes of the physical entity according to the feedback.Ideally, the digital twin system can self-learning according to multiple feedback source data and present the real situation of a physical entity almost in real-time.In this paper, the MMS communication stack protocol is used to map the data between the real data of relay protection and the digital twin to realize state tracking and automatic updates of the twin system.
To realize MMS communication between physical devices and digital twins, it is necessary to complete the development of MMS communication, which focuses on three parts: object mapping between IEC61850 and MMS, ASN.1 codec, and connection-oriented underlying network delivery mechanism.Considering the complexity of IEC61850 and MMS mapping, this article utilizes the "MMS-ESAE LITE" software package from SISC to achieve communication mapping and underlying communication transmission.The mapping structure is shown in Figure 3. MMS server uses MVL to obtain parse device SCL files.In the process of parsing SCL files, all kinds of objects in MMS are dynamically created, such as logical devices, logical nodes, data objects, etc., and memory allocation size is determined.The mapping between data type objects, VMD objects, and MMS user data is completed by analyzing and mapping short addresses of data attributes in data objects.
The state tracking and updating data interaction process of the relay protection twin model is shown in Figure 4.The secondary online monitoring system at the station communicates with the protection equipment in actual operation through DLT860/IEC61850 protocol.It obtains all kinds of parameters of the protection equipment, such as fixed value, control word, pressure plate, remote signal, and telemetry, which are actively pushed to the digital twin system of relay protection through MMS protocol in XML format.The digital twin system of relay protection obtains the data and updates the parameters of the relay equipment in real-time.For example, the system receives a protection device setting file, as shown in Figure 5, retrieves the modules to be assigned in the system based on the<desc>description in the file, and assigns the corresponding values in <val> in the file to the corresponding modules.Therefore, it achieves automatic updates of parameters such as protection settings, control words, and pressure plate settings.

Relay protection state inference and evaluation based on digital twins
Based on the established high-precision digital twin model, the secondary physical equipment can be monitored and analyzed in virtual space.The operating status of the existing secondary equipment can be evaluated through the established health status evaluation system.
The state diagnosis method based on Belief Rule Base (BRB) is shown in Figure 6.Based on the key state characteristic indicators of protection, combined with the equipment state evaluation and maintenance guidelines and expert evaluation experience, the dimension of state quantity of state evaluation guidelines is expanded to make the scoring factors more comprehensive.The real situation of the equipment is evaluated by the maintenance situation and the maintenance report.The confidence rule technique is used to realize the fusion processing of the evaluation guidelines and the qualitative and quantitative heterogeneous data included in the critical state feature database and train the state reasoning model.Under the boundary constraints of the national grid standard evaluation guidelines, we gradually make the evaluation guidelines flexible, able to cover more variables and have finer interval evaluations and health status diagnoses.

Multi-source data fusion analysis of relay protection equipment
Obtain device state data association knowledge from dynamic feature vector  In the BRB model, introducing premise attribute weights and rule weights during the regularization process can enhance the model's ability to handle random and fuzzy uncertainty information.As described below, the k-th rule of the BRB model is shown in Equation (7).
With a rule weight and attribute weight where k R is the k-th rule of the BRB evaluation model, k i A (i=1, 2…, M) refers to the reference value of the i-th input prerequisite attribute in rule k, N D represents the result set, , represents the confidence degree of j D relative to the j-th evaluation result of the output part in rule k, and N represents the total number of confidence rules.k  (k =1, 2, …, L) is the rule weight of rule k, , i k  (i=1, 2…, M) represents the weight of the i-th premise attribute in rule k, and its initial value is assumed to be 1.
In order to carry out overall state comprehensive assessment and fault diagnosis based on numerous dynamic feature vectors of relay protection equipment, ER algorithm is introduced to integrate multisource state data, and each type of input data is a class of attribute values.The following steps are performed to obtain the diagnosis result of the status interval and fault type.
The confidence of the input prerequisite attribute is calculated.The confidence ij  of the i-th attribute i x relative to the reference value is shown in Equation (8). , The confidence rule activation weight k  is calculated.The activation weight of each input attribute on k rules is different, which is calculated by Equation (9).
The confidence ' j  of the model output results is calculated.By combining L rules of the model, including the expert system mechanism model designed based on the guidelines, fuzzy evaluation algorithm, and deep learning neural network, confidence degree relative to the evaluation result j D can be obtained.The calculation is carried out by the ER parsing algorithm, as shown in Equation (10).
The final output of the BRB system can be obtained by combining all the rules in the BRB system, as shown in Equation ( 11), and the confidence values of different states and fault types can be obtained.Finally, the diagnosis result of the state with the highest confidence is selected as the output result of the confidence rule reasoning model. , Digital twinning can be used for analysis, prediction, diagnosis, training, etc., and the simulation results can be fed back to the physical object to optimize the equipment condition evaluation and maintenance decision.

Case verification
To verify the good status monitoring and action deduction effect of the digital twin system for relay protection, this paper constructs an ideal 220 kV double bus single-segmented main wiring model, as shown in Figure 7.The bus, power supply, and outgoing lines are ideal components, and the load adopts a three-phase load with an active power of 103 kW and a reactive power of 103 kVAR.During the operation, two 220 kV ideal power sources are input, which are respectively defined as the main transformer 1 branch and the main transformer 3 branch in the system.Two outgoing terminals are input, which are respectively defined as branch 6 and branch 16.The transformer is replaced by a Simulink current and voltage measuring element in the diagram.Figure 7. 220 kV busbar operating environment structure diagram.This article takes the three-phase short circuit fault on the I bus as an example to verify the performance of the system.The system simulation time is set to 1000 ms, with a simulation step size of 0.833 ms.The three-phase short circuit fault is set as a permanent fault from 400 ms to 1000 ms, and the fault occurs on the I bus.The test grounding resistance value is 0.01 Ω, and the protection operates normally without any rejection.The operation of bus differential protection is shown in Figure 8.When the fault occurs on the I bus at 400 ms, the direction of the branch current changes, and the bus large difference is not zero.At this time, the large difference current starting module in the bus differential protection starts at 401.2 ms, and the large difference power frequency change ratio differential module and the large difference ratio differential module also act simultaneously at 401.2 ms.On the other hand, due to the fault located on the I bus, the small difference of the I bus becomes a non-zero value when the fault occurs at 400 ms.The I bus power frequency change ratio differential module and I bus ratio differential module also act at 401.2 ms.The Ⅱbus differencial protection outputs a trip busbar signal according to the I bus steady state differential element.
In summary, at 401.2 ms, the differential protection tripped all branches on the I bus and the busbar 1 switch.After the switch jumps off, it undergoes a dynamic process of about 10 ms.The current of each branch transformer on the I bus and the current of the busbar 1 transformer become 0. Subsequently, both the large differential current startup module and the conventional ratio differential module return at 427.5 ms, and the power frequency variation differential module returns at 447.8 ms.Finally, the I bus differential protection module returns at 427.5 ms, and theⅡ bus differential protection differential tripping busbar output returns at 427.5 ms.During the entire process, the Ⅲ bus differential protection module did not act.This paper also designed the busbar Ⅰ failure fault, busbar 1 joint dead zone fault, busbar 1-bit dead zone fault, and circuit breaker failure, which effectively verified the excellent performance of the relay protection twin system.

Conclusions
This article studies the digital twin technology of relay protection and designs a digital twin system of relay protection based on the digital twin of protection function and software and hardware status.It realizes the deduction and evaluation of relay protection status based on the digital twin, which can effectively improve the business collaboration efficiency of protection equipment and intelligent remote inspection system.It fully responds to the centralized monitoring business requirements of the inspection system for unattended substations, as well as the construction goal of a "digital team".

Figure 1 .
Figure 1.Relay protection digital twin system d I is large differential phase current, cdzd I is the differential current starting value.The action criterion of the conventional ratio differential element is shown in Equation (4).

Figure 2 .
Figure 2. Bus differential protection simulation subsystem design logic diagram

Figure 4 .
Figure 4. Relay protection twin model status tracking and updating.

Figure 6 .
Figure 6.State diagnosis using confidence rule reasoning