Micro-grid integrated protection method based on multi-criterion adaptive identification
By using a multi-criteria adaptive identification method, a perception layer-decision layer-execution layer architecture is constructed to achieve accurate identification and stable control of microgrid faults. This solves the problem of long fault handling time during microgrid mode switching, improves power supply reliability and adaptability, and reduces hardware costs.
Patent Information
- Application Number
- CN202511969230.3
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2025-12-24
- Publication Date
- 2026-05-05
AI Technical Summary
Existing technologies are difficult to adapt to the differences in fault characteristics between microgrid grid-connected mode and islanded mode, resulting in maloperation or failure of protection devices, affecting power supply stability. Furthermore, the lack of fault prediction mechanism during mode switching leads to longer fault handling time and reduced power supply reliability.
A multi-criteria adaptive identification method is adopted to construct a control architecture of perception layer-decision layer-execution layer. Through multi-dimensional fault feature fusion detection, accurate fault identification and stable control are achieved during the grid-connected/islanded mode switching process, including real-time acquisition of electrical quantity data, pattern recognition, fault judgment, execution layer operation, and fault isolation and recovery.
It achieves a dual breakthrough in fault detection accuracy and mode switching stability, reduces false positive and false negative rates, ensures smooth switching, meets GB/T 33593-2017 standard, improves power supply reliability and adaptability, and reduces hardware costs.
Smart Images

Figure CN121984150A_ABST
Abstract
Description
Technical Field
[0001] This invention belongs to the field of microgrid protection technology, specifically relating to an integrated microgrid protection method based on multi-criteria adaptive identification. Background Technology
[0002] Microgrids, as a key carrier for distributed energy consumption and flexible operation of distribution networks, need to frequently switch between grid-connected and islanded modes (e.g., switching to islanded operation during grid failures and switching back to grid-connected operation after grid recovery). However, the fault characteristics of the two operating modes are fundamentally different: Grid-connected mode: The fault current is supported by the large power grid, with large amplitude and slow attenuation. The fault types are mainly three-phase short circuit and two-phase short circuit, and there is harmonic interference on the grid side. Island mode: The fault current is provided only by distributed power sources (such as photovoltaic, energy storage, and micro-turbines). The magnitude is small and the decay is fast. The fault types are mostly single-phase grounding and transient faults caused by sudden load changes, and the voltage drop is larger.
[0003] Currently, existing technologies are ill-suited to the aforementioned differences in fault characteristics and have significant drawbacks: First, the use of a single fault detection algorithm (such as overcurrent protection or voltage surge detection) cannot take into account the fault characteristics of both modes, which can easily lead to maloperation or failure of protection devices, affecting power supply stability; second, the mode switching process lacks a fault prediction mechanism, and if a sudden fault occurs after the switching command is issued, it can easily cause voltage dips and frequency fluctuations, which may damage distributed power supply equipment in severe cases; third, the fault isolation and recovery strategies are not designed in conjunction with mode switching, resulting in long fault handling times, which in turn reduces the power supply reliability of the microgrid and fails to meet actual operational needs. Summary of the Invention
[0004] In view of the shortcomings of the prior art, the purpose of this invention is to provide a microgrid integrated protection method based on multi-criteria adaptive identification. Through multi-dimensional fault feature fusion detection and hierarchical collaborative control strategy, it can achieve accurate fault identification and stable control during grid-connected / islanded mode switching.
[0005] To achieve the above objectives, this invention provides a microgrid integrated protection method based on multi-criteria adaptive identification, comprising the following steps: S1. In the microgrid, a three-layer control architecture of perception layer, decision layer and execution layer is constructed. The perception layer collects electrical quantity data of the microgrid in real time, including voltage, current, frequency and harmonic content. The decision layer performs pattern recognition and fault judgment based on electrical quantity data. The execution layer performs mode switching and fault handling operations. S2. Real-time collection of electrical quantity data of the microgrid using the sensing layer, and real-time identification of the current operating mode of the microgrid based on the power flow direction and frequency deviation at the grid connection point at the decision layer. The operating modes include grid connection mode and islanding mode. S3. At the decision-making level, extract fault feature parameters, match weight coefficients for each fault feature parameter, and use a multi-dimensional feature fusion algorithm to fuse fault feature parameters in grid-connected mode or islanded mode to determine the fault status. S4. The decision-making layer identifies the current operating mode of the microgrid and generates instructions to be sent to the execution layer. When the execution layer performs grid-connected to islanded switching, it first predicts the power demand of islanded operation, and after cutting off non-critical loads, it performs circuit breaker tripping and distributed power controller mode switching. When performing islanded to grid-connected switching, it first adjusts the microgrid voltage and frequency to synchronize with the grid, and then uses a soft grid-connected strategy to control the closing angle to complete the switching. After switching, it monitors faults and triggers protection mechanisms. S5. When a fault is detected, locate and isolate the fault location. For momentary faults, attempt to restore the circuit by reclosing. For permanent faults, adjust the operating strategy to ensure power supply to non-faulty areas.
[0006] As a preferred embodiment of the present invention, in S1, the sensing layer specifically involves deploying one distributed detection unit at the outlet of each distributed power source and at the grid connection point where the microgrid connects to the power grid, to collect electrical quantity data in real time. Specifically, the decision-making layer consists of a central controller with a built-in fault feature database and control strategy library. Based on the data from the perception layer, it identifies the current operating mode of the microgrid in real time, makes fault judgments, and generates instructions to be sent to the execution layer. The execution layer includes grid-connected circuit breakers, distributed power controllers, and load switching switches, which receive instructions from the decision layer and execute operations.
[0007] As a preferred embodiment of the present invention, in S2, the microgrid grid-connected / islanded mode is identified in real time based on the power flow direction and frequency deviation at the grid connection point: if the total active power P at the grid connection point is ≥0 and the frequency deviation Δf is ≤±0.1Hz, the grid supplies power to the microgrid, and it is determined to be in grid-connected mode; if the total active power P at the grid connection point is <0 and the frequency deviation Δf is >±0.1Hz, the microgrid supplies power to the load, and it is determined to be in islanded mode; the rest are abnormal operating states.
[0008] As a preferred embodiment of the present invention, in step S3, fault characteristic parameters are extracted for grid-connected mode and islanded mode respectively. Specifically, for grid-connected mode, the fault current amplitude I is extracted using wavelet transform modulus maxima detection. f The voltage drop depth ΔU is extracted by calculating the effective value, and the effective value of the nth harmonic is extracted by the fast Fourier transform to obtain the total harmonic distortion (THD). For islanded mode, the frequency change rate df / dt is extracted using the first-order difference method, where f is the frequency and t is the time. The voltage sag duration t is obtained based on the duration during which the voltage is below 85% of the rated value. drop The rate of change of energy storage SOC, dSOC / dt, is calculated by integrating the energy storage charging and discharging power. Will I f ΔU and THD are used as characteristic parameters of grid-connected faults, and df / dt and t are used as parameters of grid-connected faults. drop dSOC / dt are used as island fault characteristic parameters. Each fault characteristic parameter is standardized and a corresponding weight coefficient is matched for each fault characteristic parameter. The fault characteristic parameters are fused through a multi-dimensional feature fusion algorithm to obtain fault judgment indicators F1=Σ(grid-connected characteristic parameter × weight coefficient) and F2=Σ(island characteristic parameter × weight coefficient). Set the grid connection mode threshold T grid And the threshold T of the island mode island In grid-connected mode, when F1≥T grid When F2 ≥ T, it is determined to be a fault state; in islanded mode, when F2 ≥ T island When this occurs, it is determined to be a fault state.
[0009] As a preferred embodiment of the present invention, T grid The method of obtaining it is: ; In the formula, This refers to the real-time operating capacity of the microgrid. This serves as a baseline value for microgrid capacity. To adjust the step size; T island The method of obtaining it is: .
[0010] As a preferred embodiment of the present invention, in S4, when a fault condition is determined to occur on the grid side in grid-connected mode, if the sensing layer collects I in real time... f If the current is ≥1.5 times the rated current, the decision-making level will immediately trigger the prediction mechanism to calculate the minimum power P required for islanded operation. min =Total load power - Maximum output of distributed power source; If P min If P ≤ 0, it indicates that the distributed power source can meet the load demand, and a switchover can be performed directly; if P min If the value is greater than 0, a load shedding command will be issued first, prioritizing the shedding of non-critical loads, including streetlights and commercial loads, until P... min ≤0; Execution layer execution switching: Disconnect the grid connection point circuit breaker, with a tripping time of ≤20ms, and simultaneously switch the distributed power controller from PQ mode to V / f mode, and switch the grid connection mode to islanding mode; The decision-making level monitors energy storage overload faults in islanded mode in real time. If an energy storage overload fault with dSOC / dt < -5% / min is detected, the backup energy storage unit is immediately activated. Within 300ms after the switch is completed, the execution layer adjusts the output of the distributed power supply to restore the frequency to 50±0.02Hz and the voltage to the rated value ±1%.
[0011] As a preferred embodiment of the present invention, in S4, in islanded mode, the sensing layer collects voltage and frequency data from the grid side, the decision layer determines whether the grid meets the synchronization conditions, if not, the microgrid frequency is adjusted by energy storage charging and discharging, and the voltage deviation is compensated by the static var generator, if the conditions are met, a switching command is triggered. The execution layer adopts a soft grid connection strategy, controls the closing angle of the circuit breaker at the grid connection point, gradually increases from 0° to 180°, the closing time is 50ms, the inrush current is ≤1.2 times the rated current, the distributed power controller switches from V / f mode to PQ mode, operates according to the preset power command, and switches the island mode to grid connection mode. The protection mechanism is as follows: if ΔU ≥ 30% is detected on the grid-connected side within 1 second after switching, the circuit breaker at the grid connection point will be immediately disconnected, and the system will return to islanded mode.
[0012] As a preferred embodiment of the present invention, in S5, fault isolation is performed as follows: the decision-making layer locates the fault location based on fault characteristic parameters, and the execution layer isolates the fault area by opening the circuit breaker. The isolation time is ≤150ms. If it is an instantaneous fault, the circuit breaker is reclosed within 3s after the fault isolation. After successful reclosing, normal operation is restored. If it is a permanent fault, the decision-making layer adjusts the operation strategy, including reducing the output of distributed power sources and disconnecting the load in the fault area to ensure normal power supply to non-fault areas.
[0013] An integrated microgrid protection device based on multi-criteria adaptive identification includes: One or more processors; Memory, used to store one or more computer programs; Specifically, when one or more programs are executed by one or more processors, the one or more processors execute the methods described above.
[0014] A computer-readable storage medium having executable instructions stored thereon, which, when executed by a processor, cause the processor to perform the methods described above.
[0015] The beneficial effects of this invention are: This invention achieves a dual breakthrough in fault detection accuracy and mode switching stability through a three-layer architecture (perception layer, decision layer, and execution layer) and a multi-dimensional feature fusion algorithm. On one hand, addressing the differences in fault characteristics between grid-connected and islanded modes, the grid-connected mode integrates fault current amplitude, voltage sag depth, and harmonic content, while the islanded mode integrates frequency change rate, voltage sag duration, and energy storage SOC change rate. Furthermore, the fault threshold is adaptively adjusted according to the microgrid capacity, reducing false positive and false negative rates, far superior to traditional single algorithms. On the other hand, the layered collaborative switching strategy ensures smooth switching. Power demand is predicted and load optimized before grid-connected to islanded switching, enabling circuit breaker tripping and controller mode conversion within 20ms and voltage and frequency stabilization within 300ms. The islanded to grid-connected switching adopts a soft grid-connection strategy with a 50ms gradual closing angle change, and the inrush current is ≤1.2 times the rated current, meeting the GB / T 33593-2017 standard, thus solving the problems of voltage sag, frequency fluctuation, and equipment damage risks in traditional switching.
[0016] This invention also boasts strong adaptability, high reliability, and cost advantages, fully adapting to the actual operational needs of microgrids. In terms of adaptability, it can cover various types of microgrids including photovoltaics, energy storage, and micro-gas turbines, with a capacity range of 100kW-10MW, eliminating the need for redesign for different scenarios. Regarding reliability, it can complete fault location and isolation within 150ms, reclosing to restore power in case of instantaneous faults within 3 seconds, and power outage time ≤500ms, significantly improving the power supply guarantee capability for critical loads. In terms of cost, it reuses existing monitoring units in the microgrid, such as smart meters and PMU synchronous phasor measurement devices, requiring only software algorithm upgrades. Compared to the traditional solution of additionally deploying transient waveform recorders and harmonic analyzers (each costing over 100,000 RMB), hardware costs are reduced by 60%, while also eliminating complex manual debugging, significantly reducing the difficulty of project implementation. Attached Figure Description
[0017] Figure 1 This is a flowchart illustrating the principle of this invention; Figure 2 This is a schematic diagram of the microgrid sensing layer-decision layer-execution layer control architecture in this invention; Figure 3 This is a flowchart of fault feature extraction in grid-connected and islanded modes in this invention; Figure 4 This is the timing diagram for the grid-connected → islanded mode switching control in this invention; Figure 5 This is the soft grid connection closing angle control curve for islanded → grid-connected mode in this invention; Figure 6 These are the voltage and frequency change curves during grid-connected to islanded switching in this invention. Detailed Implementation
[0018] The embodiments of the present invention will be further described below with reference to the accompanying drawings: Example 1: As Figure 1 As shown, a microgrid integrated protection method based on multi-criteria adaptive identification includes the following steps: S1. In the microgrid, a three-layer control architecture of perception layer, decision layer and execution layer is constructed. The perception layer collects electrical quantity data of the microgrid in real time, including voltage, current, frequency and harmonic content. The decision layer performs pattern recognition and fault judgment based on electrical quantity data. The execution layer performs mode switching and fault handling operations. S2. Real-time collection of electrical quantity data of the microgrid using the sensing layer, and real-time identification of the current operating mode of the microgrid based on the power flow direction and frequency deviation at the grid connection point at the decision layer. The operating modes include grid connection mode and islanding mode. S3. At the decision-making level, extract fault feature parameters, match weight coefficients for each fault feature parameter, and use a multi-dimensional feature fusion algorithm to fuse fault feature parameters in grid-connected mode or islanded mode to determine the fault status. S4. The decision-making layer identifies the current operating mode of the microgrid and generates instructions to be sent to the execution layer. When the execution layer performs grid-connected to islanded switching, it first predicts the power demand of islanded operation, and after cutting off non-critical loads, it performs circuit breaker tripping and distributed power controller mode switching. When performing islanded to grid-connected switching, it first adjusts the microgrid voltage and frequency to synchronize with the grid, and then uses a soft grid-connected strategy to control the closing angle to complete the switching. After switching, it monitors faults and triggers protection mechanisms. S5. When a fault is detected, locate and isolate the fault location. For momentary faults, attempt to restore the circuit by reclosing. For permanent faults, adjust the operating strategy to ensure power supply to non-faulty areas.
[0019] In S1, the sensing layer specifically involves deploying one distributed detection unit at the outlet of each distributed power source and at the grid connection point where the microgrid connects to the power grid. This unit collects electrical quantity data in real time, with a sampling frequency of no less than 2kHz and a data transmission delay of no more than 50ms. Specifically, the decision-making layer consists of a central controller with a built-in fault feature database and control strategy library. Based on the data from the perception layer, it identifies the current operating mode of the microgrid in real time, makes fault judgments, and generates instructions to be sent to the execution layer. The execution layer includes grid-connected circuit breakers (PQ control mode), distributed power controllers (V / f control mode), and load switching switches, which receive instructions from the decision layer and execute operations.
[0020] In S2, based on the power flow direction and frequency deviation at the grid connection point, the microgrid's grid-connected / islanded mode is identified in real time: if the total active power at the grid connection point P≥0 and the frequency deviation Δf≤±0.1Hz, the grid supplies power to the microgrid, and it is determined to be in grid-connected mode; if the total active power at the grid connection point P<0 and the frequency deviation Δf>±0.1Hz, the microgrid supplies power to the load, and it is determined to be in islanded mode; other situations are abnormal operating states. The mode recognition result is updated every 10ms to ensure real-time performance.
[0021] In S3, fault characteristic parameters are extracted for grid-connected mode and islanded mode respectively. For grid-connected mode, the fault current amplitude I is extracted by wavelet transform modulus maxima detection. f The voltage drop depth ΔU (sliding window length of 1 cycle) is extracted by calculating the effective value, and the effective value of the nth harmonic is extracted by fast Fourier transform to obtain the total harmonic distortion (THD). For islanded mode, the frequency change rate df / dt is extracted using the first-order difference method, where f is the frequency and t is the time. The voltage sag duration t is obtained based on the duration during which the voltage is below 85% of the rated value. drop The rate of change of energy storage SOC, dSOC / dt, is calculated by integrating the energy storage charging and discharging power. Will I f ΔU and THD are used as characteristic parameters of grid-connected faults, and df / dt and t are used as parameters of grid-connected faults. drop dSOC / dt are used as island fault characteristic parameters. Each fault characteristic parameter is standardized and a corresponding weight coefficient is matched for each fault characteristic parameter. The fault characteristic parameters are fused through a multi-dimensional feature fusion algorithm to obtain fault judgment indicators F1=Σ(grid-connected characteristic parameter × weight coefficient) and F2=Σ(island characteristic parameter × weight coefficient). In this embodiment, I f The weights of ΔU and THD are set to 0.4, 0.3, and 0.3, respectively, and the weights of df / dt and t are also set to 0.4, 0.3, and 0.3, respectively. drop The weights of dSOC / dt are set to 0.5, 0.3, and 0.2, respectively.
[0022] Set the grid connection mode threshold T grid And the threshold T of the island mode island In grid-connected mode, when F1≥T grid When F2 ≥ T, it is determined to be a fault state; in islanded mode, when F2 ≥ T island When this occurs, it is determined to be a fault state.
[0023] T grid T island You can set it directly (e.g., T) grid The value is 0.6, T island (The value can be 0.5), or it can be adaptively adjusted according to the microgrid capacity, for example, T.grid The method of obtaining it is: ; In the formula, This represents the real-time operating capacity of the microgrid, in kW. This is a baseline value for microgrid capacity, in kW; To adjust the step size; T grid The adjustment range is 0.55~0.65. If Less than 100kW, T grid The threshold is set to 0.55; if Greater than 10000kW, T grid The threshold is set to 0.65.
[0024] The initial value is 0.05, which can be dynamically adapted to changes in capacity. The larger the grid capacity, the higher the fault current amplitude supported by the large grid during grid connection, so the threshold needs to be appropriately increased to avoid misjudgment. The fault current of a 10000kW microgrid is prone to exceed twice the rated value, and a threshold of 0.65 can filter out small disturbances.
[0025] T island The method of obtaining it is: ; T island The adjustment range is 0.45~0.55. If Less than 100kW, T island The threshold is set to 0.45; if Greater than 10000kW, T island The threshold is set to 0.55. In islanded mode, the larger the capacity, the stronger the frequency and voltage regulation capabilities of distributed power sources such as energy storage and micro-turbines, and the higher the tolerance to voltage sags and frequency fluctuations. The threshold needs to be increased accordingly to avoid over-protection.
[0026] In S4, when a fault condition is detected on the grid side in grid-connected mode, if the sensing layer collects I in real time... f If the current is ≥1.5 times the rated current, the decision-making level will immediately trigger the prediction mechanism to calculate the minimum power P required for islanded operation. min =Total load power - Maximum output of distributed power source; If P min If P ≤ 0, it indicates that the distributed power source can meet the load demand, and a switchover can be performed directly; if P min If the value is greater than 0, a load shedding command will be issued first, prioritizing the shedding of non-critical loads, including streetlights and commercial loads, until P... min ≤0; During mode switching, the control voltage sag is ≤10% and the frequency fluctuation is ≤±0.2Hz; Execution layer switching: Disconnect the grid connection point circuit breaker, with a tripping time of ≤20ms, and simultaneously switch the distributed power controller from PQ mode to V / f mode, with voltage control accuracy ±2% and frequency control accuracy ±0.05Hz, and switch the grid connection mode to islanding mode; The decision-making level monitors energy storage overload faults in islanded mode in real time. If an energy storage overload fault of dSOC / dt < -5% / min is detected, the backup energy storage unit is immediately activated with a response time of ≤100ms. Within 300ms after the switch is completed, the execution layer adjusts the output of the distributed power supply to restore the frequency to 50±0.02Hz and the voltage to the rated value ±1%.
[0027] In island mode, the sensing layer collects voltage and frequency data from the grid side, and the decision layer determines whether the grid meets the synchronization conditions, namely, voltage deviation ≤ ±5%, frequency deviation Δf ≤ ±0.1Hz, and phase difference ≤ ±5°. If not, the microgrid frequency is adjusted by energy storage charging and discharging, and the voltage deviation is compensated by static var generator. If the conditions are met, a switching command is triggered. The execution layer adopts a soft grid connection strategy, controlling the closing angle of the circuit breaker at the grid connection point, gradually increasing it from 0° to 180°, with a closing time of 50ms, to avoid inrush current. The inrush current is ≤1.2 times the rated current. The distributed power controller switches from V / f mode to PQ mode, operates according to preset active and reactive power commands, and has a power adjustment rate ≤0.5kW / ms, switching the islanded mode to grid connection mode. The protection mechanism is as follows: if ΔU ≥ 30% is detected on the grid-connected side within 1 second after switching, the circuit breaker at the grid connection point will be immediately disconnected and the system will return to islanded mode to prevent the fault from spreading.
[0028] In S5, fault isolation involves the decision-making layer locating the fault location based on fault characteristic parameters, and the execution layer isolating the fault area by tripping the circuit breaker. The isolation time is ≤150ms. If it is a transient fault, including lightning strikes and transient short circuits, reclosing is attempted within 3 seconds after fault isolation. After successful reclosing, normal operation is restored. If it is a permanent fault, including line breaks and equipment damage, the decision-making layer adjusts the operation strategy, including reducing the output of distributed power sources and disconnecting the load in the fault area to ensure normal power supply to non-faulty areas.
[0029] The method of this embodiment will be verified using a specific example. The verification process is as follows: Taking a microgrid in an industrial park (capacity 500kW, including 200kW photovoltaic, 200kW energy storage, 100kW micro gas turbine, 300kW for important loads and 200kW for non-important loads) as an example, the implementation process of the method in this embodiment is explained. Step 1: System Deployment Sensing layer: Deploy one detection unit at each of the photovoltaic outlet, energy storage outlet, micro gas turbine outlet, and grid connection point to collect data on voltage 380V, current 0-1000A, and frequency 50Hz. Decision-making level: PLC is used as the central controller, with a built-in fault characteristic database. The fault threshold is 0.6 in grid-connected mode and 0.5 in islanded mode. Execution layer: The grid connection point uses a vacuum circuit breaker with a 15ms opening time, and the energy storage controller uses V / f-PQ dual-mode control.
[0030] Figure 2 The control architecture diagram of the microgrid, consisting of the sensing layer, decision-making layer, and execution layer, is shown. Figure 3 The flowchart for fault feature extraction in grid-connected and islanded modes is shown (in this verification example, the 3rd and 5th harmonic contents THD3 and THD5 are extracted by fast Fourier transform, and THD3 and THD5 are used to replace the total harmonic distortion rate THD in the calculation).
[0031] Step 2: Grid Connection → Islanding Handover Implementation: The sensing layer detected the fault current I in grid-connected mode. f =1.8 times the rated current (800A), voltage drop ΔU=40%, harmonic content THD3=5%, fusion index F1=0.4×1.8+0.3×0.4+0.3×5=2.34≥0.6, judged as grid-connected side fault; The decision-making level calculates the minimum power P for islanded operation. min =500kW-(200+200+100)kW=0, no load shedding is required, issue a switching command; The execution layer disconnects the grid connection point circuit breaker (trips within 15ms), the energy storage and micro-turbine switch to V / f mode, and the frequency recovers to 50±0.03Hz and the voltage recovers to 380V±2% within 200ms; After the switchover was completed, the system ran stably in islanded mode without any load interruption.
[0032] Figure 4 The timing diagram for switching between grid-connected and islanded modes is shown. Figure 5 The control curve for soft grid connection closing angle in islanded → grid-connected mode is shown.
[0033] Step 3: Islanding to Grid Connection Switching Implementation (Grid Restoration) The sensing layer detected a grid voltage of 380V±1%, a frequency of 50±0.05Hz, and a phase difference of 3°, which met the synchronization conditions. The decision-making level issued a soft grid connection instruction, and the execution level controlled the circuit breaker closing angle at the grid connection point to increase from 0° to 180° (50ms), with an inrush current of 700A (1.1 times the rated current). Distributed power sources switch to PQ mode, with photovoltaic power output at 200kW, energy storage at 0kW (standby), and micro-turbine power output at 100kW, achieving power stabilization within 1 second; There were no faults after the switchover, and the system resumed grid-connected operation.
[0034] Figure 6 The voltage and frequency variation curves during grid-connected to islanded switching in the embodiment are shown. From Figure 6 The data shows that the voltage drops below 228V (ΔU=40%) in the first 0-15ms, recovers to 372.4-387.6V (±2%) in the 15-200ms range, and stabilizes at 380V±1% in the 200-300ms range. Furthermore, the frequency response curve shows fluctuations to 49.8Hz (Δf=-0.2Hz) in the first 0-15ms range, recovers to 49.97-50.03Hz in the 15-200ms range, and stabilizes at 50Hz±0.02Hz in the 200-300ms range.
[0035] Through the above verification process, it can be concluded that the method of this embodiment has the following advantages: High fault detection accuracy: Through multi-dimensional feature fusion and pattern adaptive recognition, the fault false positive rate is ≤0.5% and the false negative rate is ≤0.1%, which is far lower than the false positive rate of 5%-10% of existing technologies; Good switching stability: voltage dip ≤10% and frequency fluctuation ≤±0.2Hz during grid-connected to islanded switching; islanded to grid-connected switching inrush current ≤1.2 times rated current, meeting GB / T 33593-2017 "Technical Requirements for Grid Connection of Distributed Power Sources"; Highly adaptable: It can be adapted to various types of microgrids including photovoltaics, energy storage, and micro gas turbines, with microgrid capacities ranging from 100kW to 10MW, and the fault threshold and control parameters can be adaptively adjusted; High reliability: fault handling time ≤300ms, power interruption time ≤500ms, improving the microgrid's ability to guarantee power supply to critical loads.
[0036] Example 2: In the multi-dimensional feature fusion algorithm of Example 1, a corresponding weight coefficient is matched for each fault feature parameter, and a quantitative adaptive weight adjustment method based on the dynamic development characteristics of faults is adopted. The specific implementation method is as follows: By analyzing electrical quantity data collected by the sensing layer in real time at the decision-making level, the fault development stage is quantitatively determined based on industry standards. The initial stage of the fault is categorized by sudden changes in characteristic parameters. Triggering occurs when a voltage change ≥ 5% of the rated voltage or a current change ≥ 10% of the rated current with a duration < 100ms. In grid-connected mode, the fault current amplitude I... fIn islanded mode, the weight of the frequency change rate df / dt is increased to 2-3 times that of other characteristic parameters, prioritizing the capture of instantaneous fault signals; in the fault development stage, parameter fluctuation stability is used as the criterion, and switching is performed when the sudden change amplitude drops to 1%-5% and the duration is 100-500ms, with the weight deviation of all characteristic parameters controlled within 5%, to collaboratively capture the fault evolution trend; in the fault stabilization stage, the steady state of parameters is used as the criterion, and activation is performed when the sudden change amplitude is <1% or the duration is >500ms, with the weight of total harmonic distortion rate THD in grid-connected mode and energy storage SOC change rate dSOC / dt in islanded mode increased to 2-3 times that of other parameters, accurately locking the steady-state characteristics of the fault; The weight adjustment interval is synchronized with the data acquisition cycle of the perception layer (10-20ms), and the IEEE 1588 precise time protocol is adapted to achieve time synchronization, ensuring that the weight switching and fault development are matched in real time.
[0037] The problem of ambiguous stage division is solved by quantitative criteria. The weights are allocated in a relative proportion to adapt to the characteristics of different microgrid topologies. No new hardware deployment is required. It can be achieved by optimizing the algorithm logic alone. Compared with the fixed weight method, the fault judgment misjudgment rate is reduced by more than 20%, the missed judgment rate is reduced by more than 25%, and the adaptability under complex operating conditions is improved.
[0038] Example 3: A microgrid integrated protection method and device based on multi-criteria adaptive identification, comprising: One or more processors; Memory, used to store one or more computer programs; When one or more programs are executed by one or more processors, the one or more processors perform the method in Embodiment 1 or Embodiment 2.
[0039] Example 4: A computer-readable storage medium having executable instructions stored thereon, which, when executed by a processor, cause the processor to perform the method in Example 1 or Example 2.
Claims
1. A microgrid integrated protection method based on multi-criteria adaptive identification, characterized in that... Includes the following steps: S1. In the microgrid, a three-layer control architecture of perception layer, decision layer and execution layer is constructed. The perception layer collects electrical quantity data of the microgrid in real time, including voltage, current, frequency and harmonic content. The decision layer performs pattern recognition and fault judgment based on electrical quantity data. The execution layer performs mode switching and fault handling operations. S2. Real-time collection of electrical quantity data of the microgrid using the sensing layer, and real-time identification of the current operating mode of the microgrid based on the power flow direction and frequency deviation at the grid connection point at the decision layer. The operating modes include grid connection mode and islanding mode. S3. At the decision-making level, extract fault feature parameters, match weight coefficients for each fault feature parameter, and use a multi-dimensional feature fusion algorithm to fuse fault feature parameters in grid-connected mode or islanded mode to determine the fault status. S4. The decision-making layer identifies the current operating mode of the microgrid and generates instructions to be sent to the execution layer. When the execution layer performs grid-connected to islanded switching, it first predicts the power demand of islanded operation, and after cutting off non-critical loads, it performs circuit breaker tripping and distributed power controller mode switching. When performing islanded to grid-connected switching, it first adjusts the microgrid voltage and frequency to synchronize with the grid, and then uses a soft grid-connected strategy to control the closing angle to complete the switching. After switching, it monitors faults and triggers protection mechanisms. S5. When a fault is detected, locate and isolate the fault location. For momentary faults, attempt to restore the circuit by reclosing. For permanent faults, adjust the operating strategy to ensure power supply to non-faulty areas.
2. The integrated microgrid protection method based on multi-criteria adaptive identification according to claim 1, characterized in that, In S1, the sensing layer specifically involves deploying one distributed detection unit at the outlet of each distributed power source and at the grid connection point where the microgrid connects to the power grid, to collect electrical quantity data in real time. Specifically, the decision-making layer consists of a central controller with a built-in fault feature database and control strategy library. Based on the data from the perception layer, it identifies the current operating mode of the microgrid in real time, makes fault judgments, and generates instructions to be sent to the execution layer. The execution layer includes grid-connected circuit breakers, distributed power controllers, and load switching switches, which receive instructions from the decision layer and execute operations.
3. The integrated microgrid protection method based on multi-criteria adaptive identification according to claim 1, characterized in that, In S2, based on the power flow direction and frequency deviation at the grid connection point, the microgrid grid-connected / islanded mode is identified in real time: if the total active power P at the grid connection point is ≥0 and the frequency deviation Δf is ≤±0.1Hz, the grid supplies power to the microgrid, which is determined to be grid-connected mode; if the total active power P at the grid connection point is <0 and the frequency deviation Δf is >±0.1Hz, the microgrid supplies power to the load, which is determined to be islanded mode; the rest are abnormal operating states.
4. The integrated microgrid protection method based on multi-criteria adaptive identification according to claim 2, characterized in that, In S3, fault characteristic parameters are extracted for both grid-connected and islanded modes. For grid-connected mode, the fault current amplitude I is extracted using wavelet transform modulus maxima detection. f The voltage drop depth ΔU is extracted by calculating the effective value, and the effective value of the nth harmonic is extracted by the fast Fourier transform to obtain the total harmonic distortion (THD). For islanded mode, the frequency change rate df / dt is extracted using the first-order difference method, where f is the frequency and t is the time. The voltage sag duration t is obtained based on the duration during which the voltage is below 85% of the rated value. drop The rate of change of energy storage SOC, dSOC / dt, is calculated by integrating the energy storage charging and discharging power. Will I f ΔU and THD are used as characteristic parameters of grid-connected faults, and df / dt and t are used as parameters of grid-connected faults. drop dSOC / dt are used as island fault characteristic parameters. Each fault characteristic parameter is standardized and a corresponding weight coefficient is matched for each fault characteristic parameter. The fault characteristic parameters are fused through a multi-dimensional feature fusion algorithm to obtain fault judgment indicators F1=Σ(grid-connected characteristic parameter × weight coefficient) and F2=Σ(island characteristic parameter × weight coefficient). Set the grid connection mode threshold T grid And the threshold T of the island mode island In grid-connected mode, when F1≥T grid When F2 ≥ T, it is determined to be a fault state; in islanded mode, when F2 ≥ T island When this occurs, it is determined to be a fault state.
5. The integrated microgrid protection method based on multi-criteria adaptive identification according to claim 4, characterized in that, T grid The method of obtaining it is: ; In the formula, This refers to the real-time operating capacity of the microgrid. This serves as a baseline value for microgrid capacity. To adjust the step size; T island The method of obtaining it is: 。 6. The integrated microgrid protection method based on multi-criteria adaptive identification according to claim 4, characterized in that, In S4, when a fault condition is determined to occur on the grid side in grid-connected mode, if the sensing layer collects I data in real time... f If the current is ≥1.5 times the rated current, the decision-making level will immediately trigger the prediction mechanism to calculate the minimum power P required for islanded operation. min =Total load power - Maximum output of distributed power source; If P min If P ≤ 0, it indicates that the distributed power source can meet the load demand, and a switchover can be performed directly; if P min If the value is greater than 0, a load shedding command will be issued first, prioritizing the shedding of non-critical loads, including streetlights and commercial loads, until P... min ≤0; Execution layer execution switching: Disconnect the grid connection point circuit breaker, with a tripping time of ≤20ms, and simultaneously switch the distributed power controller from PQ mode to V / f mode, and switch the grid connection mode to islanding mode; The decision-making level monitors energy storage overload faults in islanded mode in real time. If an energy storage overload fault with dSOC / dt < -5% / min is detected, the backup energy storage unit is immediately activated. Within 300ms after the switch is completed, the execution layer adjusts the output of the distributed power supply to restore the frequency to 50±0.02Hz and the voltage to the rated value ±1%.
7. The integrated microgrid protection method based on multi-criteria adaptive identification according to claim 4, characterized in that, In the aforementioned S4, under islanded mode, the sensing layer collects voltage and frequency data from the grid side, and the decision layer determines whether the grid meets the synchronization conditions. If not, the microgrid frequency is adjusted by energy storage charging and discharging, and the voltage deviation is compensated by the static var generator. If the conditions are met, a switching command is triggered. The execution layer adopts a soft grid connection strategy, controls the closing angle of the circuit breaker at the grid connection point, gradually increases from 0° to 180°, the closing time is 50ms, the inrush current is ≤1.2 times the rated current, the distributed power controller switches from V / f mode to PQ mode, operates according to the preset power command, and switches the island mode to grid connection mode. The protection mechanism is as follows: if ΔU ≥ 30% is detected on the grid-connected side within 1 second after switching, the circuit breaker at the grid connection point will be immediately disconnected, and the system will return to islanded mode.
8. The integrated microgrid protection method based on multi-criteria adaptive identification according to claim 1, characterized in that, In S5, fault isolation is performed by the decision-making layer locating the fault location based on fault characteristic parameters, and the execution layer isolating the fault area by opening the circuit breaker. The isolation time is ≤150ms. If it is an instantaneous fault, reclosing is attempted within 3s after fault isolation. After successful reclosing, normal operation is restored. In the case of a permanent fault, the decision-making level will adjust the operation strategy, including reducing the output of distributed power sources and disconnecting the load in the faulty area to ensure normal power supply to the non-faulty areas.
9. A microgrid integrated protection device based on multi-criteria adaptive identification, characterized in that, include: One or more processors; Memory, used to store one or more computer programs; Wherein, when one or more programs are executed by one or more processors, the one or more processors perform the method of any one of claims 1-8.
10. A computer-readable storage medium having executable instructions stored thereon, characterized in that, When the instruction is executed by the processor, it causes the processor to perform the method of any one of claims 1-8.
