Intelligent power grid electric energy quality monitoring system and method and storage medium

Abstract

The invention relates to the technical field of smart power grids, and discloses a smart power grid electric energy quality monitoring system and method and a storage medium. Constructing a power grid graph according to the power grid topological graph of the transformer area; obtaining vertex characteristics of the first vertex according to the power grid operation state information code, and obtaining vertex characteristics of a corresponding second vertex according to the operation state information code of the flexible power distribution equipment; inputting the vertex features of the first vertex and the second vertex into a multi-type decision network, and outputting control parameters of the flexible power distribution equipment; controlling the flexible power distribution equipment in the regional power grid according to the control parameters of the flexible power distribution equipment; according to the invention, the negative interaction of multiple DFCATS devices can be reduced, and the effects of voltage stabilization and frequency stabilization can be maximized through combined control of flexible power distribution devices in a transformer area.

Description

Technical Field

[0001] The present invention relates to the field of smart grid technology, and more specifically, to a smart grid power quality monitoring system, method and storage medium. Background Art

[0002] The introduction of distributed power sources into the power grid, the extensive use of power electronic equipment, and nonlinear loads may all lead to a decline in the power quality of the distribution network. Therefore, improving the power quality of the distribution network is an important part of improving the safe and stable operation of the distribution network and ensuring the safe use of electricity by users. There are many ways to improve the power quality of the power grid, among which various types of DFACTS devices have attracted the attention of domestic and foreign scholars for their excellent performance and flexible control methods.

[0003] Flexible distribution technology is an extension of Flexible AC transmission System (FACTS) technology in distribution network, referred to as DFACTS (Distribution FACTS).

[0004] FACTS uses power electronics and control technology to flexibly and quickly adjust the impedance, voltage, phase and other basic parameters of the AC transmission system, and then flexibly control the active and reactive power flows of the system to improve the transmission capacity and stability of the transmission system. As an important technical means to improve the safety, stability and operating efficiency of the power system, FACTS technology has been widely used in power systems. Experiments have shown that after multiple DFCATS devices are connected, there will be mutual influence. Negative interactions cannot be avoided simply by task allocation. Therefore, they need to be coupled to the same control system to avoid negative interactions, and different control parameters of multiple DFCATS devices need to be considered. Summary of the invention

[0005] The present invention provides a smart grid power quality monitoring system, method and storage medium to solve technical problems in related technologies.

[0006] The present invention provides a method for monitoring power quality of a smart grid, comprising:

[0007] Step 100, collecting observation information; the observation information includes power grid operation status information and operation status information of each flexible power distribution device;

[0008] Step 200, constructing a power grid diagram according to the power grid topology diagram of the substation, the power grid diagram including a first vertex and a second vertex, the first vertex represents the power grid of the substation, and a second vertex represents a flexible power distribution device;

[0009] There are edges between the first vertex and all second vertices, and there are edges between the second vertices corresponding to the flexible power distribution devices directly connected on the power grid topology graph;

[0010] Step 300, obtaining a vertex feature of a first vertex according to the power grid operation status information code, and obtaining a vertex feature of a corresponding second vertex according to the operation status information code of a flexible power distribution device;

[0011] Step 400, inputting vertex features of the first vertex and the second vertex into a multi-type decision network, the multi-type decision network comprising a shared layer and a plurality of task-specific layers, wherein the shared layer inputs vertex features of the first vertex and the second vertex and outputs hidden features of the second vertex;

[0012] A task-specific layer inputs the hidden features of the second vertex and outputs a type of control parameter of a flexible power distribution device, and the flexible power distribution device represented by the input second vertex is the same type as the flexible power distribution device corresponding to the output control parameter;

[0013] Step 500: Control the flexible power distribution equipment in the control area power grid according to the control parameters of the flexible power distribution equipment.

[0014] Furthermore, the calculation formula of the shared layer is as follows:

[0015]

[0016] Where E i represents the hidden feature of the i-th vertex, W represents the weight parameter of the shared layer, and X i and X j Represent the vertex features of the i-th and j-th vertices, respectively, T i represents the set of vertices directly connected to the i-th vertex, M represents T i The total number of vertices in ;

[0017] R i,j =tanh(δ*X i T *X j +1

[0018] tanh represents the hyperbolic tangent function, δ represents an adjustable parameter; the default value of δ is 1 / N, where N is the dimension of the vertex feature.

[0019] Furthermore, the calculation formula of the task-specific layer is as follows:

[0020]

[0021] Where W c and b cThey represent the weight parameter and bias parameter of the cth task-specific layer, σ represents the sigmoid function, G represents the control parameter matrix of the c-th task-specific layer corresponding to the i-th flexible power distribution device of the c-th type, and the element in the m-th row and n-th column represents the probability of the m-th discrete value of the n-th control parameter of the flexible power distribution device; i Represents the hidden features of the second vertex corresponding to the i-th flexible power distribution device of the c-th type.

[0022] Furthermore, there is an edge between the second vertices corresponding to the flexible power distribution devices connected to the same variable.

[0023] Furthermore, the flexible distribution equipment in the substation includes the following types: intelligent general transformer, dynamic voltage restorer, dynamic uninterruptible power supply, static synchronous compensator, and static reactive power compensation device.

[0024] Furthermore, the multi-type decision network uses the Q-Learning method to train the decision model, and the state of Q-Learning is defined as the input of the multi-type decision network; the action space of Q-Learning is a combination of discrete values ​​of the control parameters of the flexible distribution equipment, and an action contains a discrete value of all control parameters of the flexible distribution equipment.

[0025] Furthermore, the Q value of an action is the average value of the probability values ​​of the control parameter matrix corresponding to the discrete values ​​corresponding to the action.

[0026] Furthermore, the reward function of Q-Learning is:

[0027]

[0028] in Indicates the maximum value of the grid voltage deviation value of the p-th flexible distribution equipment grid connection point from time t to t+1;

[0029] V b Indicates the voltage standard value, f t Indicates the maximum value of the grid frequency deviation from time t to time t+1, f b represents the standard value of frequency, and Pe represents the collection of flexible distribution equipment.

[0030] A smart grid power quality monitoring system, comprising:

[0031] An information collection module collects observation information; the observation information includes the operation status information of the power grid and the operation status information of each flexible distribution device;

[0032] The flexible distribution equipment in the substation area includes the following types: intelligent general transformer, dynamic voltage restorer, dynamic uninterruptible power supply, static synchronous compensator, and static reactive power compensation device.

[0033] The operating parameters and control parameters of each type of flexible distribution equipment are different.

[0034] A power grid diagram generation module is used to construct a power grid diagram according to a power grid topology diagram of a substation, wherein the power grid diagram includes a first vertex and a second vertex, wherein the first vertex represents a power grid of the substation, and a second vertex represents a flexible power distribution device;

[0035] There are edges between the first vertex and all second vertices, and there are edges between the second vertices corresponding to the flexible power distribution devices directly connected on the power grid topology graph;

[0036] Or there is an edge between the second vertices corresponding to the flexible distribution devices connected to the same variable.

[0037] An encoding module, which obtains vertex features of the first vertex according to the power grid operation status information encoding, and obtains vertex features of the corresponding second vertex according to the operation status information encoding of the flexible power distribution device;

[0038] A decision module, inputting vertex features of the first vertex and the second vertex into a multi-type decision network, the multi-type decision network comprising a shared layer and a plurality of task-specific layers, wherein the shared layer inputs the vertex features of the first vertex and the second vertex and outputs a hidden feature of the second vertex;

[0039] A task-specific layer inputs the hidden features of the second vertex and outputs a type of control parameter of a flexible power distribution device, and the flexible power distribution device represented by the input second vertex is the same type as the flexible power distribution device corresponding to the output control parameter;

[0040] The execution control module controls the flexible power distribution equipment in the control area power grid according to the control parameters of the flexible power distribution equipment.

[0041] The present invention discloses a storage medium, wherein the medium stores a computer program, and when the program is executed by a processor, the processor executes the steps of the aforementioned smart grid power quality monitoring method.

[0042] The beneficial effects of the present invention are as follows: the present invention can reduce the negative interactions among multiple DFCATS devices, and can realize the joint control of the flexible power distribution equipment in the substation to maximize the effects of voltage and frequency stabilization. BRIEF DESCRIPTION OF THE DRAWINGS

[0043] Figure 1 It is a flow chart of the present invention. DETAILED DESCRIPTION

[0044] The subject matter described herein will now be discussed with reference to example implementations. It should be understood that the discussion of these implementations is only to enable those skilled in the art to better understand and implement the subject matter described herein, and the functions and arrangements of the elements discussed may be changed without departing from the scope of protection of the present specification. Various examples may omit, replace, or add various processes or components as needed. In addition, the features described in some examples may also be combined in other examples.

[0045] At least one embodiment of the present invention discloses a method for monitoring power quality of a smart grid. Figure 1 As shown, the following steps are included:

[0046] Step 100, collecting observation information; the observation information includes power grid operation status information and operation status information of each flexible power distribution device;

[0047] The flexible distribution equipment in the substation area includes the following types: intelligent general transformer, dynamic voltage restorer, dynamic uninterruptible power supply, static synchronous compensator, and static reactive power compensation device.

[0048] The operating parameters and control parameters of each type of flexible distribution equipment are different.

[0049] Step 200, constructing a power grid diagram according to the power grid topology diagram of the substation, the power grid diagram including a first vertex and a second vertex, the first vertex represents the power grid of the substation, and a second vertex represents a flexible power distribution device;

[0050] There are edges between the first vertex and all second vertices, and there are edges between the second vertices corresponding to the flexible power distribution devices directly connected on the power grid topology graph;

[0051] Or there is an edge between the second vertices corresponding to the flexible distribution devices connected to the same variable.

[0052] Step 300, obtaining a vertex feature of a first vertex according to the power grid operation status information code, and obtaining a vertex feature of a corresponding second vertex according to the operation status information code of a flexible power distribution device;

[0053] Step 400, inputting vertex features of the first vertex and the second vertex into a multi-type decision network, the multi-type decision network comprising shared layers and a plurality of task-specific layers, wherein the shared layer inputs the vertex features of the first vertex and the second vertex and outputs the hidden features of the second vertex;

[0054] A task-specific layer inputs the hidden features of the second vertex and outputs a type of control parameter of a flexible power distribution device, and the flexible power distribution device represented by the input second vertex is the same type as the flexible power distribution device corresponding to the output control parameter;

[0055] The calculation formula of the shared layer is as follows:

[0056]

[0057] Where E i represents the hidden feature of the i-th vertex, W represents the weight parameter of the shared layer, and X i and X j Represent the vertex features of the i-th and j-th vertices, respectively, T i represents the set of vertices directly connected to the i-th vertex, M represents T i The total number of vertices in ;

[0058] R i,j =tanh(δ*X i T *X j +1

[0059] tanh represents the hyperbolic tangent function, δ represents an adjustable parameter; the default value of δ is 1 / N, where N is the dimension of the vertex feature;

[0060] The calculation formula of the task-specific layer is as follows:

[0061]

[0062] Where W c and b c They represent the weight parameter and bias parameter of the cth task-specific layer, σ represents the sigmoid function, G represents the control parameter matrix of the c-th task-specific layer corresponding to the i-th flexible power distribution device of the c-th type, and the element in the m-th row and n-th column represents the probability of the m-th discrete value of the n-th control parameter of the flexible power distribution device; i Represents the hidden features of the second vertex corresponding to the i-th flexible power distribution device of the c-th type.

[0063] The multi-type decision network uses the Q-Learning method to train the decision model and defines the state of Q-Learning as the input of the multi-type decision network. The action space of Q-Learning is a combination of discrete values ​​of the control parameters of the flexible distribution equipment, and an action contains a discrete value of all the control parameters of the flexible distribution equipment.

[0064] The Q value of an action is the average of the probability values ​​of the control parameter matrix corresponding to the discrete values ​​corresponding to the action;

[0065] The reward function of Q-Learning is:

[0066]

[0067] in represents the maximum value of the grid voltage deviation value of the pth flexible distribution equipment grid connection point from time t to t+1, V b Indicates the voltage standard value, f t Indicates the maximum value of the grid frequency deviation from time t to time t+1, f b represents the standard value of frequency, and Pe represents the collection of flexible distribution equipment.

[0068] Through Q-Learning, multi-type decision-making networks can realize the joint control of flexible distribution equipment within the substation to maximize the effects of voltage and frequency stabilization.

[0069] Step 500: Control the flexible power distribution equipment in the control area power grid according to the control parameters of the flexible power distribution equipment.

[0070] In one embodiment of the present invention, the power grid operation status observation information includes:

[0071] Voltage: Monitor the voltage level on the output side of the distribution transformer and the voltage at each vertex of the low-voltage line.

[0072] Current: Monitors current levels in low voltage lines, especially load currents in distribution transformers.

[0073] Frequency: The frequency of the power grid, usually 50Hz or 60Hz.

[0074] Power factor: reflects the ratio of active power to apparent power in the system.

[0075] Reactive power: reactive power consumption in the substation area.

[0076] Active power: active power consumption in the substation area.

[0077] Apparent power: the vector sum of reactive power and active power.

[0078] Line fault information (fault type, location, time, etc.).

[0079] Smart transformer operating status information:

[0080] Voltage level: The input and output voltage level of the transformer.

[0081] Capacity: The maximum output power of a transformer, usually expressed in kilovolt-amperes (kVA) or megavolt-amperes (MVA).

[0082] Connection group: The connection method between windings, such as star connection (Y) or triangle connection (△).

[0083] Impedance voltage (Uk%): It indicates the percentage drop in voltage when the transformer is short-circuited, reflecting the internal impedance of the transformer;

[0084] Current: The current value of the primary and secondary sides.

[0085] Voltage: The voltage value of the primary and secondary sides.

[0086] Power Factor (PF): A parameter that reflects the nature of the load and is ideally close to 1.

[0087] Load factor: The ratio of the current load to the rated capacity of the transformer, used to evaluate operating efficiency.

[0088] Dynamic Voltage Restorer (DVR) operating status information:

[0089] Power Status:

[0090] Input voltage: monitors the voltage level supplied by the power grid to the DVR.

[0091] Output voltage: Monitor the output voltage after DVR compensation to ensure that it meets the load requirements.

[0092] Voltage drop detection: When voltage drop is detected, the DVR starts the compensation mechanism.

[0093] Compensation capability:

[0094] Compensation voltage: The compensation voltage generated by the DVR.

[0095] Compensation current: The compensation current injected into or extracted from the grid by the DVR.

[0096] Compensation time: The time from detecting voltage drop to starting compensation. Usually the DVR can react within a few milliseconds.

[0097] Energy storage status:

[0098] DC bus voltage: the charging status of the DVR's internal energy storage unit (such as a battery or supercapacitor).

[0099] Energy storage unit SOC (State of Charge): The state of charge of the energy storage unit, used to understand the remaining available energy.

[0100] Charge and discharge status: Displays whether the energy storage unit is in the charging or discharging state.

[0101] Inverter status:

[0102] Working mode: The operating mode of the DVR, such as normal power supply mode or inverter power supply mode.

[0103] Temperature: The operating temperature of key components inside the inverter.

[0104] Fault record: record any abnormal situation of the inverter.

[0105] Load status:

[0106] Load voltage: The voltage actually received by the load.

[0107] Load current: The current drawn by the load from the grid.

[0108] Power Factor: The power factor of a load, which indicates how effectively the load uses electricity.

[0109] The operating status information of the uninterruptible power supply (UPS) includes:

[0110] Output status:

[0111] Output voltage: The stable voltage provided by the UPS to the load.

[0112] Output Current: The current provided by the UPS to the load.

[0113] Output power: The actual power provided by the UPS to the load, usually expressed in kilowatts (kW) or kilovolt-amperes (kVA).

[0114] Load percentage: The percentage of current load relative to the rated capacity of the UPS.

[0115] Battery status:

[0116] Battery Voltage: The voltage level of the battery pack.

[0117] Battery Capacity: The remaining battery capacity, usually expressed as a percentage.

[0118] Battery charge status: Whether the battery is charging or discharging.

[0119] Battery health status: The health status of the battery, such as whether it is aging or needs replacement.

[0120] Bypass status:

[0121] Bypass switch status: Whether the bypass switch is closed.

[0122] Bypass voltage: The voltage level when bypass power is supplied.

[0123] Bypass frequency: The frequency of bypass power supply.

[0124] Inverter status:

[0125] Inverter output voltage: The voltage provided by the inverter.

[0126] Inverter output frequency: The frequency provided by the inverter.

[0127] Inverter Temperature: The temperature inside the inverter to prevent overheating.

[0128] Inverter fault status: whether there is a fault alarm or error message.

[0129] The operating status information of the static VAR compensation device (SVC) includes:

[0130] Electrical parameters

[0131] Reactive Power: The reactive power level provided by the SVC, measured in kilovars (kVar).

[0132] Active Power: Although SVCs are primarily concerned with reactive power, small amounts of active power may be generated in certain circumstances.

[0133] Apparent Power: The vector sum of reactive power and real power, expressed in kilovolt-amperes (kVA).

[0134] Control parameters

[0135] Compensation current: The compensation current injected into the grid by the SVC to offset the reactive current in the system.

[0136] Firing angle: If the SVC contains a thyristor controlled reactor (TCR), the firing angle of the thyristor will be monitored.

[0137] Capacitor switching status: If the SVC contains fixed or controllable capacitor banks, the switching on and off status of these capacitors is monitored.

[0138] The operating status information of the static synchronous compensator (STATCOM) includes:

[0139] Electrical parameters

[0140] Reactive power: The reactive power level provided by the STATCOM, in kilovars (kVar).

[0141] Active power: Although STATCOM is mainly used for reactive power compensation, it may also generate a small amount of active power in some cases.

[0142] Apparent power: The vector sum of reactive power and active power, expressed in kilovolt-amperes (kVA).

[0143] Power factor: reflects the ratio of active power to apparent power.

[0144] Control parameters

[0145] Compensation current: The compensation current injected into the grid by STATCOM to offset the reactive current in the system.

[0146] Firing angle: If a thyristor controlled reactor (TCR) is used, the firing angle of the thyristor is an important control parameter.

[0147] Capacitor bank status: For STATCOM containing capacitor banks, the switching status of capacitors is also an important control parameter.

[0148] The operation status information of the flexible power distribution equipment includes operation parameters. The general encoding method for the operation parameters is to directly combine them as vectors, which will lose type information and characteristic. In one embodiment of the present invention, a method for encoding the operation status information of the flexible power distribution equipment is provided, including:

[0149] The operation parameters in the operation status information are combined into actual measurement values ​​(vectors) in order, and one dimension of the actual measurement value represents one operation parameter;

[0150] The actual measurement value is input into the unscented Kalman filter, and the state estimation value (vector) is output as the vertex feature of the second vertex.

[0151] The calculation formula of the unscented Kalman filter is as follows:

[0152] Calculate Sigma points:

[0153]

[0154] λ=α 2 (n+κ)-n

[0155] Where α and k are adjustment parameters, and n is the state dimension;

[0156] Time update:

[0157] Propagate the Sigma point through the state equation: k|k-1 =f(χ k-1 ,u k-1 )

[0158] Predicted state mean:

[0159] Prior state covariance matrix:

[0160] Measurement update (correction):

[0161] Propagate the Sigma point through the measurement equation: Y k|k-1 =h(χ k|k-1)

[0162] Predict the measured mean: Predicted measurement covariance matrix:

[0163] The cross-covariance matrix of states and measurements:

[0164] Kalman gain: Update the state estimate: Update the prior state covariance matrix: Symbol definition:

[0165] The estimated state value at time k;

[0166] P k|k-1 : The prior state covariance matrix at time k;

[0167] χ k : Sigma point set at time k;

[0168] u k-1 : Control input at time k;

[0169] n: state dimension;

[0170] λ, α, κ: scaling parameters used to calculate Sigma points;

[0171] f(·): state transfer function (state equation);

[0172] Q: process noise covariance matrix;

[0173] Y k : The value of the Sigma point propagated through the measurement equation at time k;

[0174] The measured predicted value at time k;

[0175] h(·): measurement function (measurement equation);

[0176] R: measurement noise covariance matrix;

[0177] : Covariance matrix of measurement prediction values;

[0178] : The cross-covariance matrix between states and measurements;

[0179] K k : Kalman gain at time k;

[0180] y k: The actual measured value at time k;

[0181] : The mean weight and variance weight of the i-th Sigma point;

[0182] The subscript k|k-1 indicates that the information at time k is predicted using the information at time k-1, and T indicates matrix transpose.

[0183] It should be noted that initialization is required when the filter is cold started:

[0184] Initial state estimate:

[0185] Initial prior state covariance matrix:

[0186] At least one embodiment of the present invention discloses a smart grid power quality monitoring system, comprising:

[0187] An information collection module collects observation information; the observation information includes the operation status information of the power grid and the operation status information of each flexible distribution device;

[0188] The flexible distribution equipment in the substation area includes the following types: intelligent general transformer, dynamic voltage restorer, dynamic uninterruptible power supply, static synchronous compensator, and static reactive power compensation device.

[0189] The operating parameters and control parameters of each type of flexible distribution equipment are different.

[0190] A power grid diagram generation module is used to construct a power grid diagram according to a power grid topology diagram of a substation, wherein the power grid diagram includes a first vertex and a second vertex, wherein the first vertex represents a power grid of the substation, and a second vertex represents a flexible power distribution device;

[0191] There are edges between the first vertex and all second vertices, and there are edges between the second vertices corresponding to the flexible power distribution devices directly connected on the power grid topology graph;

[0192] Or there is an edge between the second vertices corresponding to the flexible distribution devices connected to the same variable.

[0193] An encoding module, which obtains vertex features of the first vertex according to the power grid operation status information encoding, and obtains vertex features of the corresponding second vertex according to the operation status information encoding of the flexible power distribution device;

[0194] A decision module, inputting vertex features of the first vertex and the second vertex into a multi-type decision network, the multi-type decision network comprising shared layers and a plurality of task-specific layers, wherein the shared layer inputs the vertex features of the first vertex and the second vertex and outputs the hidden features of the second vertex;

[0195] A task-specific layer inputs the hidden features of the second vertex and outputs a type of control parameter of a flexible power distribution device, and the flexible power distribution device represented by the input second vertex is the same type as the flexible power distribution device corresponding to the output control parameter;

[0196] The execution control module controls the flexible power distribution equipment in the control area power grid according to the control parameters of the flexible power distribution equipment.

[0197] At least one embodiment of the present invention discloses a storage medium storing a computer program. When the program is executed by a processor, the processor executes the steps of a smart grid power quality monitoring method.

[0198] The above describes an embodiment of the present invention, but this embodiment is not limited to the above-mentioned specific implementation mode. The above-mentioned specific implementation mode is merely illustrative and not restrictive. Under the guidance of this embodiment, ordinary technicians in this field can also make more forms of equivalent embodiments, all of which are within the protection of this embodiment.

Claims

1. A method for monitoring power quality of a smart grid, characterized in that: include: Step 100, collecting observation information; The observation information includes the operation status information of the power grid and the operation status information of each flexible distribution device; Step 200, constructing a power grid diagram according to the power grid topology diagram of the substation, the power grid diagram including a first vertex and a second vertex, the first vertex represents the power grid of the substation, and a second vertex represents a flexible power distribution device; There are edges between the first vertex and all second vertices, and there are edges between the second vertices corresponding to the flexible power distribution devices directly connected on the power grid topology graph; Step 300, obtaining a vertex feature of a first vertex according to the power grid operation status information code, and obtaining a vertex feature of a corresponding second vertex according to the operation status information code of a flexible power distribution device; Step 400, inputting vertex features of the first vertex and the second vertex into a multi-type decision network, the multi-type decision network comprising a shared layer and a plurality of task-specific layers, wherein the shared layer inputs vertex features of the first vertex and the second vertex and outputs hidden features of the second vertex; A task-specific layer inputs the hidden features of the second vertex and outputs a type of control parameter of a flexible power distribution device, and the flexible power distribution device represented by the input second vertex is the same type as the flexible power distribution device corresponding to the output control parameter; Step 500: Control the flexible power distribution equipment in the control area power grid according to the control parameters of the flexible power distribution equipment.

2. A smart grid power quality monitoring method according to claim 1, characterized in that: The calculation formula of the shared layer is as follows: Where E i represents the hidden feature of the i-th vertex, W represents the weight parameter of the shared layer, and X i and X j Represent the vertex features of the i-th and j-th vertices, respectively, T i represents the set of vertices directly connected to the i-th vertex, M represents T i The total number of vertices in ; R i,j =tanh(δ*X i T *X j +1 tanh represents the hyperbolic tangent function, δ represents an adjustable parameter; the default value of δ is 1 / N, where N is the dimension of the vertex feature.

3. A smart grid power quality monitoring method according to claim 1, characterized in that: The calculation formula of the task-specific layer is as follows: Where W c and b c They represent the weight parameter and bias parameter of the cth task-specific layer, σ represents the sigmoid function, G represents the control parameter matrix of the c-th task-specific layer corresponding to the i-th flexible power distribution device of the c-th type, and the element in the m-th row and n-th column represents the probability of the m-th discrete value of the n-th control parameter of the flexible power distribution device; i Represents the hidden features of the second vertex corresponding to the i-th flexible power distribution device of the c-th type.

4. A smart grid power quality monitoring method according to claim 1, characterized in that: There is an edge between the second vertices corresponding to the flexible distribution devices connected to the same variable.

5. A smart grid power quality monitoring method according to claim 1, characterized in that: The flexible distribution equipment in the substation area includes the following types: intelligent general transformer, dynamic voltage restorer, dynamic uninterruptible power supply, static synchronous compensator, and static reactive power compensation device.

6. A smart grid power quality monitoring method according to claim 1, characterized in that: The multi-type decision network uses the Q-Learning method to train the decision model and defines the state of Q-Learning as the input of the multi-type decision network. The action space of Q-Learning is a combination of discrete values ​​of the control parameters of the flexible distribution equipment, and an action contains a discrete value of all the control parameters of the flexible distribution equipment.

7. A smart grid power quality monitoring method according to claim 6, characterized in that: The Q value of an action is the average of the probability values ​​of the control parameter matrix corresponding to the discrete values ​​corresponding to the action.

8. A smart grid power quality monitoring method according to claim 6, characterized in that: The reward function of Q-Learning is: Where V t p Indicates the maximum value of the grid voltage deviation value of the p-th flexible distribution equipment grid connection point from time t to t+1; V b Indicates the voltage standard value, f t Indicates the maximum value of the grid frequency deviation from time t to time t+1, f b represents the standard value of frequency, and Pe represents the collection of flexible distribution equipment.

9. A smart grid power quality monitoring system, characterized in that: include: Information collection module, collecting observation information; The observation information includes the operation status information of the power grid and the operation status information of each flexible distribution device; The flexible power distribution equipment in the substation area includes the following types: intelligent general transformer, dynamic voltage restorer, dynamic uninterruptible power supply, static synchronous compensator, static reactive power compensation device; A power grid diagram generation module is used to construct a power grid diagram according to a power grid topology diagram of a substation, wherein the power grid diagram includes a first vertex and a second vertex, wherein the first vertex represents a power grid of the substation, and a second vertex represents a flexible power distribution device; There are edges between the first vertex and all second vertices, and there are edges between the second vertices corresponding to the flexible power distribution devices directly connected on the power grid topology graph; An encoding module, which obtains vertex features of the first vertex according to the power grid operation status information encoding, and obtains vertex features of the corresponding second vertex according to the operation status information encoding of the flexible power distribution device; A decision module, inputting vertex features of the first vertex and the second vertex into a multi-type decision network, the multi-type decision network comprising a shared layer and a plurality of task-specific layers, wherein the shared layer inputs the vertex features of the first vertex and the second vertex and outputs a hidden feature of the second vertex; A task-specific layer inputs the hidden features of the second vertex and outputs a type of control parameter of a flexible power distribution device, and the flexible power distribution device represented by the input second vertex is the same type as the flexible power distribution device corresponding to the output control parameter; The execution control module controls the flexible power distribution equipment in the control area power grid according to the control parameters of the flexible power distribution equipment.

10. A storage medium, characterized in that: The medium stores a computer program, and when the program is executed by a processor, the processor executes the steps of a smart grid power quality monitoring method as described in any one of claims 1-8.

Images