Multi-agent universal detection and accurate positioning method for short-circuit fault of alternating-current micro-grid

Abstract

The invention provides a universal detection and accurate positioning method for a short-circuit fault of a multi-agent alternating-current micro-grid, and the method comprises the steps: dividing a cable of the alternating-current micro-grid into a plurality of regions through a multi-agent system, determining a fault region through the communication between agents, and carrying out the accurate positioning of the fault through the agents responsible for the region. The influence of different working modes and topological change of the micro-grid is self-adapted; the method can adapt to influences of different working modes and topological changes of the micro-grid, and has universality.

Description

Technical Field

[0001] The present invention relates to the technical field of power grid operation and maintenance, and in particular to a general detection and precise positioning method for short-circuit faults in a multi-agent AC microgrid. Background Art

[0002] As the proportion of renewable energy generation in my country's energy structure gradually increases, the stable operation of microgrids has received more and more attention. However, microgrid protection faces many problems, such as the bidirectionality of short-circuit current direction, the input and removal of different capacities may change the flow direction, and the protection device may flow reverse current; the power electronic interface device limits the short-circuit current amplitude at the outlet to within twice the rated current, making the fault characteristics unclear when a short circuit occurs. The existence of such problems makes the traditional short-circuit fault protection scheme no longer applicable to microgrids, so it is necessary to study new short-circuit fault protection methods for microgrids.

[0003] Although some progress has been made in the existing research on short-circuit faults in microgrids, there are still some shortcomings. First, some studies can achieve regional positioning of short-circuit faults, but do not fully consider the differences between the two operating modes of grid-connected and islanded. Second, the accuracy of fault detection and positioning decreases significantly over long distances, and does not consider changes in capacity and topology, thus lacking versatility.

[0004] In order to solve the above problems, the present invention adopts a multi-agent system to divide the cables into multiple areas. The fault area is first determined through communication between agents, and then the agent responsible for the area accurately locates the fault. At the same time, it can adapt to the different working modes and topology changes of the microgrid and has universality. Summary of the invention

[0005] The present invention proposes a general detection and precise positioning method for short-circuit faults in AC microgrids with multiple intelligent agents, which can adapt to the influence of different working modes and topological changes of microgrids and has universality.

[0006] The present invention adopts the following technical solutions.

[0007] A general detection and precise positioning method for short-circuit faults in AC microgrids using a multi-agent system is proposed. The method uses a multi-agent system to divide the cables of the AC microgrid into multiple areas. The fault area is first determined through communication between the agents, and then the agent responsible for the area accurately locates the fault to adapt to the different working modes and topological changes of the adaptive microgrid.

[0008] The intelligent agent in the multi-agent system is composed of a detection module, a signal processing module, a fault judgment module, a communication module and a positioning module. It judges the occurrence of a fault and transmits the information to adjacent intelligent agents by real-time sampling of current signals and performing wavelet transform. The positioning module judges the fault area and accurately locates the fault.

[0009] The detection module is responsible for real-time sampling of current signals.

[0010] The signal processing module performs wavelet transform to extract features of the current signal.

[0011] The communication module uses RS485 communication protocol for communication. After the intelligent agent at the cable confirms the fault, the fault information is sent to the intelligent agent at the adjacent rear cable through the communication module.

[0012] The agents of the multi-agent system are arranged sequentially at the cable. During the operation of the AC microgrid, each agent collects the current waveform through a current sensor and extracts the wavelet energy spectrum value E4 using a binary wavelet energy spectrum transform algorithm. If an agent detects that the E4 value exceeds the set fault threshold Xd, it means that a short circuit fault has occurred in the microgrid. At this time, the agent transmits the fault information to the rear agent through the RS485 communication protocol. If the rear agent also detects a fault, it means that the fault point is not in the area between the two agents. If the front agent detects a fault signal but the rear agent does not, the rear agent can determine that the fault is located in the area between the two agents and feed back the information to the front agent through the RS485 communication module.

[0013] The fault threshold Xd is 5 times greater than the E4 value during normal operation.

[0014] For a three-phase three-wire system, the agent determines the fault type by detecting the zero-sequence current. The zero-sequence current formula is:

[0015]

[0016] For a three-phase four-wire system, the agent distinguishes between ground faults and non-ground faults by detecting residual current. The residual current formula is:

[0017]

[0018] In the AC microgrid, for different initial fault phase angles θ c , the first peak value E4 of the fault current wavelet energy spectrum waveform max1 As the distance d from the short-circuit point to the detection point increases, the E4 that can be measured by the agent decreases. max1 And the fault phase angle θ c , Distributed power capacity SDG Accurately locate the short-circuit fault; the fault accurate location method is based on the S DG ,θ c 、E4 max1 is a variable, and the distance d from the short-circuit point to the agent in the area is the objective function; when a fault occurs, the fault type and fault area are first determined by the cooperation between multiple agents, and then the S detected by the agent in the area is DG ,θ c 、E4 max1 Substitute into the precise positioning formula to calculate the exact value of the fault location;

[0019] The method for accurately locating faults is based on a BP neural network algorithm to obtain an accurate positioning formula; the accurate positioning formula obtained through BP neural network training is:

[0020] d=tansig(W1*A+B1)*W2+B2Formula 3;

[0021] d is the predicted distance from the fault point to the intelligent agent in the area, W1 is the weight from the input layer to the hidden layer, B1 is the bias of the hidden layer, W2 is the weight from the hidden layer to the output layer, B2 is the bias of the output layer, A is the input amount, and tansig is the neural network activation function;

[0022] The expression of input quantity A is

[0023]

[0024] The expression of the neural network activation function tansig is

[0025]

[0026] For the agent responsible for a certain area, when training the BP neural network algorithm, take S DG ∈{80,85,90,95,100,105,110,115,120}, θ c ∈{0,30,60,90,120,150}, d∈{75,150,225,300,375,450}, combined with all single-phase ground short-circuit faults in the area that the agent is responsible for, a total of 9×6×6=324 single-phase ground short-circuit faults in the area are obtained. max1 , that is, the total number of sample points is 324, 300 sample points are randomly selected as the training set of the BP neural network algorithm, and the remaining 24 sample points are the test set.

[0027] The present invention takes into account the two operating modes of the microgrid, grid-connected and islanded, as well as changes in the capacity of distributed power sources. It can adapt to the influence of different operating modes and topological changes of the microgrid, has universality, and verifies the effectiveness of the positioning method after changing the cable type.

[0028] The method of the present invention has taken into account capacity changes, including the switching of distributed power sources, so changes in topology, such as the removal of distributed power sources, will not affect the detection results of each intelligent agent. In addition, the method can also achieve accurate positioning for different types of cables, such as replacing copper core cables with aluminum core cables, and has good versatility.

[0029] The advantages of the present invention also lie in:

[0030] 1. Accurately locate long-distance cable short-circuit faults through the cooperation between wavelet energy spectrum values, BP neural network, and multi-agents.

[0031] 2. This method can adapt to topology and capacity changes and achieve accurate positioning of short-circuit faults in AC microgrids in all modes. BRIEF DESCRIPTION OF THE DRAWINGS

[0032] The present invention is further described in detail below with reference to the accompanying drawings and specific embodiments:

[0033] Attached Figure 1 This is a schematic diagram of a microgrid simulation model built based on MATLAB / Simulink;

[0034] Attached Figure 2 It is a schematic diagram of the process flow of accurately locating a short-circuit fault according to the present invention. DETAILED DESCRIPTION

[0035] As shown in the figure, a general detection and precise positioning method for short-circuit faults in AC microgrids of multiple agents is provided. The method uses a multi-agent system to divide the cables of the AC microgrid into multiple areas. The fault area is first determined through communication between agents, and then the agent responsible for the area accurately locates the fault to adapt to the different working modes and topological changes of the adaptive microgrid.

[0036] The intelligent agent in the multi-agent system is composed of a detection module, a signal processing module, a fault judgment module, a communication module and a positioning module. It judges the occurrence of a fault and transmits the information to adjacent intelligent agents by real-time sampling of current signals and performing wavelet transform. The positioning module judges the fault area and accurately locates the fault.

[0037] The detection module is responsible for real-time sampling of current signals.

[0038] The signal processing module performs wavelet transform to extract features of the current signal.

[0039] The communication module uses RS485 communication protocol for communication. After the intelligent agent at the cable confirms the fault, the fault information is sent to the intelligent agent at the adjacent rear cable through the communication module.

[0040] The agents of the multi-agent system are arranged sequentially at the cable. During the operation of the AC microgrid, each agent collects the current waveform through a current sensor and extracts the wavelet energy spectrum value E4 using a binary wavelet energy spectrum transform algorithm. If an agent detects that the E4 value exceeds the set fault threshold Xd, it means that a short circuit fault has occurred in the microgrid. At this time, the agent transmits the fault information to the rear agent through the RS485 communication protocol. If the rear agent also detects a fault, it means that the fault point is not in the area between the two agents. If the front agent detects a fault signal but the rear agent does not, the rear agent can determine that the fault is located in the area between the two agents and feed back the information to the front agent through the RS485 communication module.

[0041] The fault threshold Xd is 5 times greater than the E4 value during normal operation.

[0042] For a three-phase three-wire system, the agent determines the fault type by detecting the zero-sequence current. The zero-sequence current formula is:

[0043]

[0044] For a three-phase four-wire system, the agent distinguishes between ground faults and non-ground faults by detecting residual current. The residual current formula is:

[0045]

[0046] In the AC microgrid, for different initial fault phase angles θ c , the first peak value E4 of the fault current wavelet energy spectrum waveform max1 As the distance d from the short-circuit point to the detection point increases, the E4 that can be measured by the agent decreases. max1 And the fault phase angle θ c , Distributed power capacity S DG Accurately locate the short-circuit fault; the fault accurate location method is based on the S DG ,θ c 、E4 max1 is a variable, and the distance d from the short-circuit point to the agent in the area is the objective function; when a fault occurs, the fault type and fault area are first determined by the cooperation between multiple agents, and then the S detected by the agent in the area is DG ,θ c 、E4 max1Substitute into the precise positioning formula to calculate the exact value of the fault location;

[0047] The method for accurately locating faults is based on a BP neural network algorithm to obtain an accurate positioning formula; the accurate positioning formula obtained through BP neural network training is:

[0048] d=tansig(W1*A+B1)*W2+B2Formula 3;

[0049] d is the predicted distance from the fault point to the intelligent agent in the area, W1 is the weight from the input layer to the hidden layer, B1 is the bias of the hidden layer, W2 is the weight from the hidden layer to the output layer, B2 is the bias of the output layer, A is the input amount, and tansig is the neural network activation function;

[0050] The expression of input quantity A is

[0051]

[0052] The expression of the neural network activation function tansig is

[0053]

[0054] For the agent responsible for a certain area, when training the BP neural network algorithm, take S DG ∈{80,85,90,95,100,105,110,115,120}, θ c ∈{0,30,60,90,120,150}, d∈{75,150,225,300,375,450}, combined with all single-phase ground short-circuit faults in the area that the agent is responsible for, a total of 9×6×6=324 single-phase ground short-circuit faults in the area are obtained. max1 , that is, the total number of sample points is 324, 300 sample points are randomly selected as the training set of the BP neural network algorithm, and the remaining 24 sample points are the test set.

[0055] Example:

[0056] In this example, a multi-agent system is used to divide the cables into multiple areas. When a fault occurs, the fault area is first determined through communication between the agents, and then the agent responsible for the area accurately locates the fault.

[0057] Figure 1The figure shows the microgrid simulation model. Taking the three agents A, B, and C as an example, the cable system is divided into three areas, and each agent is responsible for the precise fault location task in its corresponding area: agent A is responsible for the AB section (i.e. area 1), agent B is responsible for the BC section (i.e. area 2), and agent C is responsible for the last section. If the system has more than three sections, the tasks are divided and assigned in sequence according to this mode.

[0058] When a fault occurs, the intelligent agent detects the fault, determines the fault area through mutual communication, and determines the fault type. Then the intelligent agent responsible for the precise positioning task of the area substitutes the detected feature value into the trained BP neural network positioning formula to calculate the precise location of the fault. Figure 2 The figure shows the precise positioning flow chart when a short circuit fault occurs.

[0059] The fault area location method in this example is based on S DG ,θ c 、E4 max1 is a variable, and the distance d from the short-circuit point to the agent in the area is the objective function.

[0060] by Figure 1 Taking region 1 in the figure as an example, to ensure the accuracy of the model, take S DG ∈{80,85,90,95,100,105,110,115,120}, θ c ∈{0,30,60,90,120,150}, d∈{75,150,225,300,375,450}, it is measured that when the single-phase ground short circuit fault occurs in area 1, the E4 of the detection point A max1 , we get a total of 9×6×6=324 E4 when single-phase ground short circuit occurs in area 1 max1 value, that is, the total number of sample points is 324. 300 sample points are randomly selected as the training set of the BP neural network algorithm, and the remaining 24 sample points are the test set. The precise positioning formula shown in formula (3) is obtained through BP neural network training. In formula (3), d is the predicted distance from the fault point to the intelligent agent in the area, W1 is the weight from the input layer to the hidden layer, B1 is the bias of the hidden layer, W2 is the weight from the hidden layer to the output layer, B2 is the bias of the output layer, A is the input quantity, and tansig is the neural network activation function. Formula (4) is the expression of the input quantity A, and formula (5) is the expression of the neural network activation function tansig. When a fault occurs, the fault type and fault area are first determined by the cooperation between multiple intelligent agents, and then the S detected by the intelligent agent in the area is used. DG ,θ c 、E4 max1 Substitute it into the precise positioning formula to calculate the exact value of the fault location.

Claims

1. A general detection and precise positioning method for short-circuit faults in AC microgrids based on multiple agents, characterized by: The method adopts a multi-agent system to divide the cables of the AC microgrid into multiple areas. The fault area is first determined through communication between the agents, and then the agent responsible for the area accurately locates the fault to adapt to the different working modes and topological changes of the adaptive microgrid.

2. The method for universal detection and precise positioning of short-circuit faults in AC microgrids of multiple agents according to claim 1 is characterized in that: The intelligent agent in the multi-agent system is composed of a detection module, a signal processing module, a fault judgment module, a communication module and a positioning module. It judges the occurrence of a fault and transmits the information to adjacent intelligent agents by real-time sampling of current signals and performing wavelet transform. The positioning module judges the fault area and accurately locates the fault.

3. The method for universal detection and precise positioning of short-circuit faults in a multi-agent AC microgrid according to claim 2 is characterized in that: The detection module is responsible for real-time sampling of current signals.

4. The method for universal detection and precise positioning of short-circuit faults in AC microgrids of multiple agents according to claim 2 is characterized in that: The signal processing module performs wavelet transform to extract features of the current signal.

5. The method for universal detection and precise positioning of short-circuit faults in a multi-agent AC microgrid according to claim 2 is characterized in that: The communication module uses RS485 communication protocol for communication. After the intelligent agent at the cable confirms the fault, the fault information is sent to the intelligent agent at the adjacent rear cable through the communication module.

6. The method for universal detection and precise positioning of short-circuit faults in a multi-agent AC microgrid according to claim 2 is characterized in that: The agents of the multi-agent system are arranged in sequence at the cable. During the operation of the AC microgrid, each agent collects the current waveform through the current sensor and extracts the wavelet energy spectrum value E4 using the binary wavelet energy spectrum transformation algorithm. If an agent detects that the E4 value exceeds the set fault threshold Xd, it means that a short circuit fault has occurred in the microgrid. At this time, the agent transmits the fault information to the rear agent through the RS485 communication protocol. If the rear agent also detects a fault, it means that the fault point is not in the area between the two agents. If the front agent detects a fault signal but the rear agent does not, the rear agent can determine that the fault is located in the area between the two agents and feed this information back to the front agent through the RS485 communication module.

7. The method for universal detection and precise positioning of short-circuit faults in a multi-agent AC microgrid according to claim 6 is characterized in that: The fault threshold Xd is 5 times greater than the E4 value during normal operation.

8. The method for universal detection and precise positioning of short-circuit faults in a multi-agent AC microgrid according to claim 6 is characterized in that: For a three-phase three-wire system, the agent determines the fault type by detecting the zero-sequence current. The zero-sequence current formula is: For a three-phase four-wire system, the agent distinguishes between ground faults and non-ground faults by detecting residual current. The residual current formula is:

9. The method for universal detection and precise positioning of short-circuit faults in a multi-agent AC microgrid according to claim 8, characterized in that: In the AC microgrid, for different initial fault phase angles θ c , the first peak value E4 of the fault current wavelet energy spectrum waveform max1 As the distance d from the short-circuit point to the detection point increases, the E4 that can be measured by the agent decreases. max1 And the fault phase angle θ c , Distributed power capacity S DG Accurately locate the short-circuit fault; the fault accurate location method is based on the S DG ,θ c 、E4 max1 is a variable, and the distance d from the short-circuit point to the agent in the area is the objective function; when a fault occurs, the fault type and fault area are first determined by the cooperation between multiple agents, and then the S detected by the agent in the area is DG ,θ c 、E4 max1 Substitute into the precise positioning formula to calculate the exact value of the fault location; The method for accurately locating faults is based on a BP neural network algorithm to obtain an accurate positioning formula; the accurate positioning formula obtained through BP neural network training is: d=tansig(W1*A+B1)*W2+B2 Formula 3; d is the predicted distance from the fault point to the intelligent agent in the area, W1 is the weight from the input layer to the hidden layer, B1 is the bias of the hidden layer, W2 is the weight from the hidden layer to the output layer, B2 is the bias of the output layer, A is the input amount, and tansig is the neural network activation function; The expression of input quantity A is The expression of the neural network activation function tansig is 10. The method for universal detection and precise positioning of short-circuit faults in a multi-agent AC microgrid according to claim 9, characterized in that: For the agent responsible for a certain area, when training the BP neural network algorithm, take S DG ∈{80,85,90,95,100,105,110,115,120}, θ c ∈{0,30,60,90,120,150}, d∈{75,150,225,300,375,450}, combined with all single-phase ground short-circuit faults in the area that the agent is responsible for, a total of 9×6×6=324 single-phase ground short-circuit faults in the area are obtained. max1 , that is, the total number of sample points is 324, 300 sample points are randomly selected as the training set of the BP neural network algorithm, and the remaining 24 sample points are the test set.

Images