Distributed state monitoring and fault diagnosis method for direct-current micro-grid

Abstract

The invention provides a distributed state monitoring and fault diagnosis method for a DC micro-grid. Belongs to the technical field of DC micro-grids, and solves the technical problem that the existing state monitoring method in practical application cannot accurately monitor the DC micro-grid state with load disturbance, process uncertainty and measurement noise at the same time. According to the technical scheme, the method comprises the following steps: a, establishing a distributed state space model of each power generation unit in the DC micro-grid; b, designing a time-invariant parameter matrix of the distributed observer by using a full-rank decomposition technology, and determining a time-varying parameter matrix of the distributed observer under the meaning of a minimum mean square error; and c, constructing a distributed residual generator based on local state monitoring. The method has the beneficial effects that the difficulty of high false alarm rate caused by fault propagation is overcome, the communication calculation amount is reduced, the risk of network attack is reduced, and the system has higher reliability and expandability.

Description

Technical Field

[0001] The present invention belongs to the technical field of direct current microgrids, and in particular relates to a distributed state monitoring and fault diagnosis method for a direct current microgrid. Background Art

[0002] DC microgrids do not have problems such as phase synchronization and harmonic control, and have the advantages of small size, simple structure, and good economy. They are widely used in data centers, electric vehicles, military bases and other fields. Condition monitoring is an important part of the DC microgrid energy management system, but due to the volatility of new energy and load disturbances, the condition monitoring of DC microgrids is extremely challenging. In addition, due to extreme weather, equipment aging, etc., component failures, actuator failures, sensor failures and malicious network attacks are inevitable. Therefore, the condition monitoring and fault diagnosis of DC microgrids have received widespread attention. DC microgrids are typical large-scale distributed interconnected systems. Each power generation unit is highly coupled. Once the load suddenly increases or a fault occurs, it will affect the status of other power generation units. Centralized or distributed condition monitoring methods and fault diagnosis require interactive information, which not only affects the accuracy of condition monitoring and fault diagnosis, but also requires large communication and computing resources and is at high risk of network attacks. Therefore, centralized and distributed condition monitoring methods and fault diagnosis have disadvantages.

[0003] In summary, the decentralized state monitoring and fault diagnosis method for DC microgrids has important practical significance and application value. A decentralized state monitoring and fault diagnosis method for DC microgrids is proposed. Summary of the invention

[0004] Purpose of the invention: The purpose of the present invention is to provide a distributed state monitoring and fault diagnosis method for a DC microgrid in response to the above-mentioned technical problems. The distributed observer under this method decouples the influence of local load disturbances and the state of neighboring power generation units, and also realizes the state monitoring of the DC microgrid in the sense of minimum mean square error under the influence of process uncertainty and measurement noise, and realizes DC microgrid fault diagnosis when actuator faults, voltage and current sensor faults occur simultaneously.

[0005] The inventive concept of the present invention is as follows: first, considering the influence of load disturbance, system uncertainty and fault, based on Kirchhoff's voltage and current laws and the quasi-steady-state approximation of power lines, a distributed state space model of each power generation unit in a DC microgrid is established; then, the full rank decomposition technology is used to design the time-invariant parameter matrix of the distributed observer and the time-varying parameter matrix of the distributed observer is determined in the sense of minimum mean square error, so as to realize the decoupling of the local load and the state of the neighboring power generation unit and obtain accurate DC microgrid state monitoring; then, a distributed residual generator is constructed based on local state monitoring, and concurrent fault diagnosis of the DC microgrid is realized through residual evaluation; finally, a DC microgrid composed of 6 distributed power generation units verifies the effectiveness of the proposed method.

[0006] In order to achieve the above-mentioned object of the invention, the present invention provides a distributed state monitoring and fault diagnosis method for a DC microgrid, comprising the following steps:

[0007] a. Considering the influence of load disturbance, system uncertainty and fault, Kirchhoff's voltage and current laws and the quasi-steady-state approximation of power lines, a distributed state space model of each power generation unit in the DC microgrid is established;

[0008] b. Design the time-invariant parameter matrix of the distributed observer using the full-rank decomposition technique and determine the time-varying parameter matrix of the distributed observer in the sense of minimum mean square error, so that the distributed observer can accurately monitor the status of the local power generation unit in real time;

[0009] c. A decentralized residual generator is constructed based on local state monitoring, and concurrent fault diagnosis decisions of the DC microgrid are realized through residual evaluation.

[0010] Furthermore, in step a, the influence of process uncertainty, measurement noise and load disturbance, as well as the influence of actuator failure, voltage and current sensor failure, Kirchhoff's voltage and current law and the quasi-steady-state approximation of the power line are considered to establish the DGU in the DC microgrid. i The distributed state space model of is as follows:

[0011]

[0012] where x [i] (k) = y [i] (k)=[V i , I ti ] T ,u [i] (k)=[V ti ],d [i] (k) = [I Li ],w [i] (k), v [i](k) are system state variables, output variables, control signals, unknown input disturbances, process uncertainties, and measurement noise; V i and I ti Respectively represent the voltage and filter current of the ith common coupling point; V ti Represents the control voltage of the Buck converter; I Li Represents the unknown load current; R ti , L ti , C ti are the resistance, inductance and capacitance of the RLC filter circuit respectively; V j Represents the common coupling point voltage of neighboring power generation units; N i Indicates DGU i The set of all neighbor power generation units j and the total number of neighbors is recorded as |N i |=s i ,s i ≥1; R ij , L ij are the equivalent resistance and inductance of the power line respectively; w [i] (k) and v [i] (k) are all unrelated Gaussian white noise, satisfying: is the control voltage deviation affected by the actuator fault, ΔV i is the measurement deviation of the voltage sensor, ΔI ti is the measurement deviation of the current sensor, and the corresponding known system parameter matrix is:

[0013]

[0014] In step b, design (DO i )The distributed observer is as follows:

[0015]

[0016] in Yes DO i status, It's DGU i The state estimate, F i ∈R 2×2 , T i ∈R 2×2 , H i ∈R 2 ×2 is the time-invariant parameter matrix to be designed, G i (k+1)∈R 2×2 is the time-varying parameter matrix to be designed.

[0017] The full rank decomposition technique is used to decouple the load disturbance of the local power generation unit and the state of the neighboring power generation unit, and the time-invariant parameter matrix is ​​designed to satisfy:

[0018]

[0019] T i =[IH i C i ];

[0020] F i =[IH i C i ]A zii ;

[0021] in The state monitoring of the DC microgrid is realized in the sense of minimum mean square error, and the time-invariant parameter matrix is ​​designed to satisfy:

[0022]

[0023] Where P [i] (k) is the state estimation error covariance of the local power generation unit, satisfying:

[0024]

[0025] The residual generator constructed in step c is as follows:

[0026]

[0027] The decision logic for realizing concurrent fault diagnosis of DC microgrid through residual evaluation is as follows:

[0028]

[0029] Among them J [i] (k+1) is the test statistic, satisfying J th，i is the corresponding threshold, satisfying

[0030] Compared with the prior art, the present invention has the following technical effects:

[0031] 1) The designed distributed observer-based state monitoring and concurrent fault diagnosis method decouples the influence of local load disturbance and neighboring power generation unit state, and realizes the state monitoring of each power generation unit in the sense of minimum mean square error under the influence of process uncertainty and measurement noise, and the monitoring results are more accurate;

[0032] 2) The designed distributed observer-based state monitoring and concurrent fault diagnosis method solves the problem of high false alarm rate caused by fault propagation in practical applications;

[0033] 3) The designed distributed observer-based state monitoring and concurrent fault diagnosis method solves the difficulty of large communication computing resources in centralized or distributed state monitoring and concurrent fault diagnosis methods, avoids the risk of network attacks during information exchange, and makes the system reliable and scalable. BRIEF DESCRIPTION OF THE DRAWINGS

[0034] Figure 1 A flowchart of the steps of the distributed state monitoring and fault diagnosis method of a DC microgrid provided by the present invention;

[0035] Figure 2 DGU in the present invention i Circuit model diagram of

[0036] Figure 3 This is an example diagram of a DC microgrid composed of 6 power generation units in the present invention;

[0037] Figure 4 It is a schematic diagram of the state monitoring result of a DC microgrid based on a distributed observer when the load suddenly increases in the present invention;

[0038] Figure 5 It is a schematic diagram of the state monitoring result of the DC microgrid based on the distributed observer when concurrent faults occur in the present invention;

[0039] Figure 6 It is a schematic diagram of the concurrent fault diagnosis results of the DC microgrid based on residual evaluation in the present invention. DETAILED DESCRIPTION

[0040] Example 1

[0041] Reference Figure 1 Flowchart, this embodiment provides a distributed state monitoring and fault diagnosis method for a DC microgrid, comprising the following steps:

[0042] Step a: Refer to Figure 2 DGU shown i The circuit model of DGU in DC microgrid is established by considering the influence of process uncertainty, measurement noise and load disturbance, as well as the influence of actuator failure, voltage and current sensor failure, Kirchhoff voltage and current law and quasi-steady-state approximation of power line. i The distributed state space model of is as follows:

[0043]

[0044] where x [i] (k) = y [i] (k)=[V i ,I ti ] T ,u [i](k)=[V ti ],d [i] (k) = [I Li ],w [i] (k),v [i] (k) are system state variables, output variables, control signals, unknown input disturbances, process uncertainties, and measurement noise; V i and I ti Respectively represent the voltage and filter current of the ith common coupling point; V ti Represents the control voltage of the Buck converter; I Li Represents the unknown load current; R ti , L ti , C ti are the resistance, inductance and capacitance of the RLC filter circuit respectively; V j N represents the common coupling point voltage of neighboring power generation unit j; i Indicates DGU i The set of all neighbor power generation units j and the total number of neighbors is recorded as |N i |=s i ,s i ≥1; R ij , L ij are the equivalent resistance and inductance of the power line respectively; w [i] (k) and v [i] (k) are all unrelated Gaussian white noise, satisfying: is the control voltage deviation affected by the actuator fault, ΔV i is the measurement deviation of the voltage sensor, ΔI ti is the measurement deviation of the current sensor, and the corresponding known system parameter matrix is:

[0045]

[0046] Step b: Design the decentralized observer (DO i )as follows:

[0047]

[0048] in, Yes DO i status, It's DGU i The state estimate, F i ∈R 2×2 ,T i ∈R 2×2 ,H i ∈R 2×2 is the time-invariant parameter matrix to be designed, G i (k+1)∈R 2×2is the time-varying parameter matrix to be designed.

[0049] In order to make the state monitoring not affected by the load disturbance of the local power generation unit and the state of the neighboring power generation unit, d [i] (k) and x [j] (k) are all considered unknown inputs, then DGU i The distributed state space model of is rewritten as:

[0050]

[0051] in And r i =rank(E i ) satisfies: 0<r i ≤min{2,1+2s i}.

[0052] Using the full rank decomposition technique, we can get:

[0053]

[0054] in is a column full rank matrix, is a row full rank matrix. The design time invariant parameter matrix satisfies:

[0055]

[0056] T i =[IH i C i ];

[0057] F i =[IH i C i ]A zii ;

[0058] The DC microgrid state monitoring based on the distributed observer can decouple the load disturbance of the local power generation unit and the state of the neighboring power generation unit. The time-invariant parameter matrix is ​​designed to satisfy:

[0059]

[0060] Where P [i] (k) is the state estimation error covariance of the local power generation unit, satisfying:

[0061]

[0062] The state monitoring of the DC microgrid is realized in the sense of minimum mean square error.

[0063] Step c: Construct the residual generator as follows:

[0064]

[0065] The decision logic for realizing concurrent fault diagnosis of DC microgrid through residual evaluation is as follows:

[0066]

[0067] Among them J [i] (k+1) is the test statistic, satisfying J th，i is the corresponding threshold, satisfying α is the given significance level

[0068] In this embodiment, in the MatlabR2020a environment, a DC microgrid composed of 6 distributed generation units is taken as an example. Figure 3 The method designed in this embodiment is verified, and the electrical parameters of a specific example are given in Table 1. The covariance of the process uncertainty and the measurement noise is set to be Q wi =diag(0.001,0.001),Q vi =diag(0.001,0.001), the simulation time is set to 30μs.

[0069] Example 1: Initially, the initial load current of the six power generation units is 5A; after 5μs, the load suddenly increases and the load current increases to 10.

[0070] Example 2

[0071] Based on Example 1, in Example 2, after 10 μs, DGU1 has an actuator failure, and DGU4 has a voltage sensor failure and a current sensor failure, and the fault amplitudes are respectively Given a significance level of α = 0.05, the threshold is J th,i =5.9915,i=1,...,6.

[0072] Table 1 Specific electrical parameters of DC microgrid

[0073] DG <![CDATA[R ti (Oh)]]> <![CDATA[C ti (mF)]]> <![CDATA[L ti (mH)]]> DGU(i,j) <![CDATA[R ij (Oh)]]> <![CDATA[L ij (mH) <!-- 5 -->]]> 1 0.2 2.2 1.8 (1,2) 0.05 2.1 2 0.3 1.9 2.0 (1,3) 0.07 1.8 3 0.1 1.7 2.2 (1,6) 0.1 2.5 4 0.5 2.5 3.0 (2,4) 0.04 2.3 5 0.4 2.0 1.2 (3,4) 0.06 1.0 6 0.6 3.0 2.5 (4,5) 0.08 1.8 (5,6) 0.08 3.0

[0074] To this end, the time-invariant parameter matrix F of the decentralized observer is calculated i ,T i ,H i and the time-varying parameter matrix G that converges iteratively over time i (k+1), i=1...6 are as follows:

[0075]

[0076]

[0077]

[0078]

[0079]

[0080]

[0081] Result description:

[0082] Figure 4 This is the state monitoring result of the DC microgrid when the load suddenly increases. It can be seen that under the influence of load disturbance, process uncertainty and measurement noise, the designed distributed observer can accurately monitor the voltage and current status of each power generation unit, while the traditional state monitoring method is greatly affected by noise. In particular, when the load suddenly increases after 5μs, the state monitoring method based on the distributed observer can still achieve accurate DC microgrid state monitoring.

[0083] Figure 5 This is the state monitoring result diagram of the DC microgrid during concurrent faults. It can be seen that after the 10μs concurrent fault occurs, only the state monitoring of the faulty DGU1 and DGU4 has a slight deviation, while the state monitoring of other power generation units is still accurate.

[0084] Figure 6 Based on the diagnosis results of concurrent faults in the DC microgrid based on residual evaluation, it was directly diagnosed that alarms occurred in DGU1 and DGU4 without any false alarms.

[0085] Based on the distributed state monitoring and fault diagnosis of the DC microgrid, this embodiment first considers the influence of load surge, system uncertainty and fault in practical applications, establishes the state space model of the distributed power generation unit, and then designs a distributed observer, determines the time-invariant parameter matrix through the full rank decomposition technology, decouples the influence of the load disturbance of the local power generation unit and the state of the neighboring power generation unit, determines the time-varying parameter matrix in the sense of minimum mean square error, reduces the influence of process uncertainty and measurement noise, and makes the state monitoring more accurate. The distributed design method does not require information interaction, reduces communication computing resources, greatly reduces the risk of network attacks, overcomes the difficulty of high false alarm rate caused by fault propagation, and successfully realizes concurrent fault diagnosis of DC microgrids.

[0086] Technical features not described in the present invention can be achieved through or by adopting existing technologies and will not be described in detail here. Of course, the above description is not a limitation of the present invention, and the present invention is not limited to the above examples. Changes, modifications, additions or substitutions made by ordinary technicians in this technical field within the essential scope of the present invention should also fall within the scope of protection of the present invention.

Claims

1. A distributed state monitoring and fault diagnosis method for a DC microgrid, characterized in that: The steps include: a. Considering the influence of load disturbance, system uncertainty and fault, Kirchhoff's voltage and current laws and the quasi-steady-state approximation of power lines, a distributed state space model of each power generation unit in the DC microgrid is established; b. Design the time-invariant parameter matrix of the distributed observer using the full-rank decomposition technique and determine the time-varying parameter matrix of the distributed observer in the sense of minimum mean square error, so that the distributed observer can accurately monitor the status of the local power generation unit in real time; c. A decentralized residual generator is constructed based on local state monitoring, and concurrent fault diagnosis decisions of the DC microgrid are realized through residual evaluation.

2. According to claim 1, a distributed state monitoring and fault diagnosis method design for a DC microgrid is characterized in that: In step a, the influence of process uncertainty, measurement noise and load disturbance, as well as the influence of actuator failure, voltage and current sensor failure, Kirchhoff's voltage and current law and the quasi-steady-state approximation of the power line are considered to establish the i-th generation unit DGU in the DC microgrid. i The distributed state space model of is as follows: where x [i] (k) = y [i] (k)=[V i ,I ti ] T ,u [i] (k)=[V ti ],d [i] (k) = [I Li ],w [i] (k),v [i] (k) are system state variables, output variables, control signals, unknown input disturbances, process uncertainties, and measurement noise; V i and I ti Respectively represent the voltage and filter current of the ith common coupling point; V ti Represents the control voltage of the Buck converter; I Li Represents the unknown load current; R ti , L ti , C ti are the resistance, inductance and capacitance of the RLC filter circuit respectively; V j N represents the common coupling point voltage of neighboring power generation unit j; i Indicates DGU i The set of all neighbor power generation units j and the total number of neighbors is recorded as |N i |=s i ,s i ≥1; R ij , L ij are the equivalent resistance and inductance of the power line respectively; w [i] (k) and v [i] (k) are all unrelated Gaussian white noise, satisfying: is the control voltage deviation affected by the actuator fault, ΔV i is the measurement deviation of the voltage sensor, ΔI ti is the measurement deviation of the current sensor, and the corresponding known system parameter matrix is:

3. The design of a distributed state monitoring and fault diagnosis method for a DC microgrid according to claim 1 is characterized in that: In step b, the distributed observer DO is designed i as follows: in, Yes DO i status, It's DGU i The state estimate, F i ∈R 2×2 ,T i ∈R 2×2 ,H i ∈R 2×2 is the time-invariant parameter matrix to be designed, G i (k+1)∈R 2×2 is the time-varying parameter matrix to be designed; The full rank decomposition technique is used to decouple the load disturbance of the local power generation unit and the state of the neighboring power generation unit, and the time-invariant parameter matrix is ​​designed to satisfy: T i =[I-H i C i ]; F i =[I-H i C i ]A zii ; in The state monitoring of the DC microgrid is realized in the sense of minimum mean square error, and the time-invariant parameter matrix is ​​designed to satisfy: Where P [i] (k) is the state estimation error covariance of the local power generation unit, satisfying:

4. The design of a distributed state monitoring and fault diagnosis method for a DC microgrid according to claim 1 is characterized in that: The residual generator constructed in step c is as follows: The decision logic for realizing concurrent fault diagnosis of DC microgrid through residual evaluation is as follows: Among them J [i] (k+1) is the test statistic, satisfying J th,i is the corresponding threshold, satisfying

Images