Fault diagnosis model of main equipment in substation based on multi-mode time-frequency combination

The conventional fault diagnosis method of the main equipment in substations has strong limitations and the fault diagnosis accuracy is low, so this paper introduces the multi-mode time-frequency combination method and designs the fault diagnosis model of the main equipment in substations. First, the transmission mode of fault information of main transformer equipment is analyzed and the fault source is diagnosed. Secondly, the UHF partial discharge method is used to detect the partial discharge signal of the main substation equipment. On this basis, the main equipment fault diagnosis model of the substation is established based on the multi-mode time-frequency combination, and the equipment fault degree is obtained to complete the fault diagnosis. According to the experimental results, the proposed fault diagnosis model is highly feasible, and the accuracy of an insulation fault, overheating fault and discharge fault diagnosis is more than 98%.


Introduction
The substation is an important component of power transmission and distribution in power systems.Fault diagnosis and accident handling in substations are of great significance for improving the safety and reliability of power systems and preventing malignant power outages.Most of the existing fault diagnosis models for substation power transmission and transformation equipment only use local information, mainly based on equipment status monitoring and fault characteristic information from various physical and chemical tests.It is difficult to achieve an accurate diagnosis of complex faults [2].Especially when multiple modes of data are required for fault diagnosis, the information fusion effect is poor [3].Therefore, the fault diagnosis technology of main equipment in substations needs to be continuously improved [4].Multimode time-frequency combination is a common idea, which maps them to the same vector space and then fuses them to improve the reliability of data [5].Based on this, this paper proposes research on the fault diagnosis model for main equipment in substations based on the multimode time-frequency joint method.

Diagnosis of fault source
The fault diagnosis data source of the substation main equipment is the diagnostic basis [6] for fault diagnosis, which is collected by the data acquisition and monitoring system (SCADA) and the intelligent high-voltage equipment status monitoring and management system (ENA-CBMS) through the change of switching value and electrical value of the corresponding equipment caused by the interference of natural factors or the fault of the equipment itself.Therefore, maintenance personnel needs to clearly understand the source of fault information in order to make accurate judgments about the fault.
First of all, this paper draws the schematic diagram of fault information transmission of the main equipment of the substation as shown in Figure 1.

Figure 1. Schematic Diagram of Fault Information Transmission of Main Substation Equipment
According to Figure 1, when a fault occurs, first, the current, voltage, power and other power will suddenly change.The relay protection device can issue a command to enable the circuit breaker to trip [7] according to the set action rules.Based on this, the fault types and causes of the main equipment of the substation are analyzed using MATLAB simulation analysis software.Its structure is shown in Table 1.

Detection of partial discharge signal of main substation equipment
After the diagnosis fault source of the main substation equipment is obtained based on the above, the UHF partial discharge method is used to detect the partial discharge signal of the main substation equipment.The UHF partial discharge detection field wiring diagram of the main substation equipment designed in this paper is shown in Figure 2. As shown in Figure 2, the UHF probe, oscilloscope, and partial discharge instrument are used to detect the partial discharge signal of the main transformer equipment, and the UHF sensor is placed in the box of the main transformer equipment.The detection accuracy is improved through shielding the main transformer equipment shell and strong anti-interference ability.The high-voltage side threephase of the main substation equipment is connected to the high-voltage side three-phase marked on the instrument panel one by one, and the low-voltage side of the main substation equipment is connected to the low-voltage side of the instrument to ensure good contact [8].The switch waveform measured in this test project can be used to determine whether the switch is correct, and whether there are any faults such as mechanism jamming, insulation fault, contact burning, and transmission shaft fracture, providing strong data support for partial discharge detection of substation main equipment [9].The detection of UHF signals generated by small defects in the main equipment of a substation can be used to determine partial discharges through signal distribution characteristics within the cycle, providing a reference for subsequent fault diagnosis.

Building fault diagnosis model for main equipment of substation based on multi-mode timefrequency combination
The minimum functional unit that requires communication is obtained after the decomposition of equipment functions.We select the appropriate class to model the function according to the list of logical node classes specified in IEC61850-7-4.We need to determine whether the existing LN class of the standard meets the functional requirements.If so, we use the appropriate LN class; if not, we determine whether the core requirements of the modeling function are met.If  As shown in Table 2, it is the logic node of the fault diagnosis model of the main substation equipment designed in this paper.The logic node describes the IED configuration and parameters, communication system configuration, substation system structure and their relationships related to the main substation equipment in the intelligent substation based on the IEC61850 standard, which can be used to exchange fault information data.According to SCL, the format of alarm information ID is set as "IED logical device/logical node $constraint type $data $data attribute".The logical node LN is the basic unit to complete various functions in the intelligent substation.When obtaining information, it is mainly based on the logical node to identify the required information.On this basis, the initial information of the fault diagnosis model was allocated.In an uncertain power system, the initial probability is ( ) ( ) Then, the mathematical expectation of the information entropy of the fault diagnosis model X is expressed as: where ( ) H X represents a collection of fault diagnosis models X Information entropy.After the fault diagnosis model receives the circuit breaker and main protection action alarm information, the set of all possible faults corresponding to these alarm information can be expressed as , ,..., l K b b b = , where l b can be pilot protection, zero sequence overcurrent protection and overvoltage protection, according to the protection configuration of the substation [10].When a fault occurs in the main substation equipment, in order to better integrate the fault state data, the multi-mode time-frequency joint method is used.Assuming that the probability of occurrence of each component element in the fault state set and the fault feature set is equal, the fault state based on the multi-mode time-frequency joint method is i a .The confidence information entropy is: where i n represents the action information data of fault levels.Assuming that the probability of occurrence of each fault state and fault feature is equal, the formula for calculating the probability of occurrence of the fault of main equipment in the substation is: where k indicates the fault characteristics of the main substation equipment, and l b is the number of elements.The modified probability of each section is calculated based on the information entropy theory, and the corresponding depot is assigned.Combining multi-mode time-frequency joint and fuzzy reasoning, the minimum fault set is obtained to represent the fault degree of the main substation equipment, and realize the goal of fault diagnosis for main equipment of the substation [11].

Experiment preparation
Assuming that the long-term operation data of the oil-immersed power transformer in the main transformer of a 110 kV substation are simulated, the diagnostic sample information of the oilimmersed power transformer obtained is divided into two types: routine test data for power outages and live detection data.We select 500 sets of real-time data for oil-immersed power transformers with known faults, including 200 sets of insulation faults, 150 sets of overheat faults, and 150 sets of discharge faults.Using a diagnostic model, we diagnose the real-time data of another 1500 sets of oilimmersed power transformers, input live working data and power failure working data into the model, and synchronously import factory data, historical data, and real-time data.After fault diagnosis, we calculate the true classification rate, false positive classification rate, and true negative classification rate of the diagnosis results, using the formulas: TN TNR FP TN = + (7) where TDR , FPR and TNR represent the true class rate, false positive class rate and true negative class rate of diagnosis results respectively; TP represents the positive sample diagnosed as positive by the model; FN represents the positive sample diagnosed as negative by the model; FP indicates the negative sample diagnosed as positive by the model; TN indicates a negative sample diagnosed as negative by the model.Through calculation, it is concluded that the true rate of the main equipment fault diagnosis model is as high as 98.8%, the true negative rate is as high as 99.55%, and the negative-positive rate is only 0.49%.It shows that the correct rate of the diagnostic model proposed in this paper is high, which meets the requirements of actual production in substation operation and maintenance sites.

Result analysis
The diagnostic model proposed in this paper was used as the experimental group, and the fault diagnosis methods proposed in [1] and [2] were set to control A and B for comparative analysis.Abnormal data is recorded as 0 and normal data as 1.At the same time, we recorded three oilimmersed power transformer faults: insulation, overheating and discharge faults, numbered as 01, 02 and 03 respectively in Figure 3.
The diagnosis and field investigation data of oil-immersed power transformers were analyzed by MATLAB simulation analysis software.The number of experiments was 20, and the accuracy of 20 fault diagnosis results was averaged, and the diagnostic accuracy was compared, as shown in Figure 3.

Figure 3. Comparison Results of Fault Diagnosis Accuracy of Main Substation Equipment
It can be seen from the comparison results in Figure 3 that there is a large gap in the accuracy of fault diagnosis of the three main transformer equipment.Among them, the fault diagnosis model based on multi-mode time-frequency combination proposed in this paper shows significant advantages.The accuracy of insulation fault, overheat fault and discharge fault diagnosis is more than 98%, and the accuracy of diagnosis can meet the actual diagnosis needs.It can obtain more accurate fault information on main equipment of the substation.

Conclusion
This paper proposes a diagnosis model of the main equipment of the substation based on multi-mode time-frequency combination, and uses various fault characteristics and time-frequency data fusion to conduct a comprehensive fault diagnosis analysis of the substation.Through the research of this paper, the purpose of unified modeling and data sharing is achieved, which provides a complete, shared, object-oriented data resource for the fault diagnosis of the main substation equipment in intelligent substations, and realizes the goal of hierarchical and structured comprehensive fault diagnosis of the main substation equipment in intelligent substations.

Figure 2 .
Figure 2. Field Wiring Diagram of UHF Partial Discharge Detection of Main Substation Equipment self information of the fault diagnosis model X is:

Table 1 .
Fault Types and Causes of Main Substation Equipment satisfied, new data can be added to the LN class; if not, you can create a new LN class.If the appropriate LN class meets the requirements of the modeling function, the logical node instance should have all the required attributes of the LN class.The initial letter of the LN class name should meet the prefix requirements associated with the logical node group.The name of the newly created LN class cannot conflict with the existing name, and meets the requirements of the IEC61850 namespace.The design of the main logical nodes in the fault diagnosis model of the main substation equipment is shown in Table 2. Table 2. Main Logic Node Design of Fault Diagnosis Model for Main Substation Equipment