Method and system for automatically checking relay protection setting value of power plant

Abstract

The invention relates to the technical field of relay protection, in particular to an automatic checking method and system for a relay protection setting value of a power plant, and the method comprises the steps: obtaining relay protection data, including obtaining electrical parameters, equipment parameters and power topological structure information; constructing a topological graph model of the power network by using the power topological structure information, constructing an electrical model based on elements according to the electrical parameters and the equipment parameters, and combining the topological graph model with the electrical model by using node parameter association to obtain a power system integrated model; and performing constant value check calculation based on the power system integrated model, and generating a report according to a constant value check calculation result. Hidden layers are set for different fault types by adopting the multi-branch structure network, so that the neural network can perform specialized learning for different types of fault features, and the adaptability of the system to different fault scenes is enhanced.

Description

Technical Field

[0001] The present invention relates to the technical field of relay protection, and in particular to a method and system for automatically checking relay protection constants in a power plant. Background Art

[0002] In the modern power industry, power plants are crucial sources of electric energy, and the safe and stable operation of their power systems plays a fundamental and critical role in the reliability of the entire power grid. Relay protection devices, the safety guardians of power systems, can quickly operate in the event of a system failure or abnormal operating condition, isolating the faulty component and preventing the incident from escalating, thereby ensuring the safety of power equipment and personnel, and ensuring the continuity of power supply.

[0003] However, the accuracy and rationality of relay protection settings are key elements in ensuring the proper functioning of relay protection devices. Traditional relay protection setting verification relies primarily on manual operation and empirical judgment, which presents numerous challenges. First, with the continuous expansion of power plant scale and the increasing complexity of power systems, the number and types of relay protection devices involved are numerous and complex, and the setting parameters are numerous and interrelated. Manual verification and calculation is enormous and inefficient, making it difficult to meet the rapidly evolving power system operation and maintenance requirements. Second, manual verification is complex and susceptible to subjective factors such as miscalculation, misunderstanding of procedures, and fatigue. This leads to a high error rate in setting verification, which can cause protection devices to malfunction or fail to operate during faults, posing a significant threat to the safe and stable operation of the power system. Furthermore, the complex and ever-changing power system operating environment, with frequent grid structure adjustments and constantly changing operating modes, requires relay protection settings to adapt to these changes and be verified and updated accordingly. However, traditional manual calibration methods are difficult to achieve rapid and dynamic set value adaptability assessment and adjustment, and cannot effectively cope with the dynamic changes in the power system. At this stage, there is a need for an automatic calibration method and system for power plant relay protection set values. Summary of the Invention

[0004] In order to solve the problems of low calibration efficiency, insufficient accuracy and adaptability in traditional fixed value calibration methods, the present invention provides a method and system for automatic calibration of relay protection fixed values in a power plant.

[0005] In a first aspect, the present invention provides a method for automatically checking relay protection settings in a power plant, which adopts the following technical solutions:

[0006] A method for automatically checking relay protection settings in a power plant, comprising:

[0007] Obtain relay protection data, including electrical parameters, equipment parameters, and power topology information;

[0008] Using power topology information to construct a topological graph model of the power network, including using a graph theory-based approach to set electrical equipment in power plants as nodes and lines connecting the equipment as edges;

[0009] Construct component-based electrical models based on electrical parameters and device parameters, including using Parker transformations and T-type equivalent circuits to construct electrical models;

[0010] The topology model and the electrical model are combined by using node parameter association to obtain an integrated power system model.

[0011] Perform fixed value verification calculations based on an integrated power system model, including building a multi-layer neural network to predict fault types and introducing a distributed computing framework for fixed value verification calculations;

[0012] Generate reports based on the fixed value verification calculation results.

[0013] Furthermore, the component-based electrical model is constructed based on the electrical parameters and the device parameters, including obtaining the generator parameters in the device parameters, converting the electrical quantities of the stator three-phase winding to the dp0 coordinate system according to the Park transformation formula, and constructing the electrical model of the generator based on the voltage equation in the dp0 coordinate system. The voltage equation is expressed as:

[0014]

[0015] Among them, R a is the stator resistance, U d and U q are the direct-axis and quadrature-axis components of the stator three-phase winding voltage in the rotating coordinate system, I d and I q are the direct axis and quadrature axis components of the stator current in the dq0 coordinate system, ψ d and ψ q They are the direct-axis and quadrature-axis flux linkages in the dq0 coordinate system, U0 and I0 are the zero-sequence voltage component and zero-sequence current component, respectively, and ψ0 is the zero-sequence symmetrical component of the three-phase stator winding flux linkage.

[0016] Furthermore, the construction of the component-based electrical model based on the electrical parameters and the device parameters also includes obtaining the transformer parameters in the device parameters, calculating the primary side loop equation and the secondary side loop equation respectively according to the transformer parameters, generating a T-type equivalent circuit according to the side loop equations, and then obtaining the line length, and calculating the series impedance and parallel admittance of the line. According to the distributed parameter model equation of the line, the voltage and current transmission characteristics of the line at different positions are calculated to construct the electrical model of the line.

[0017] Furthermore, the topology model and the electrical model are combined by using node parameter association, including determining the electrical connection relationship between each node through the topology model, transferring the series impedance and parallel admittance in the line to the electrical model of the node connected to the line according to the connection relationship, and establishing the relationship equation between the node voltage and the branch current based on Kirchhoff's law.

[0018] Furthermore, the construction of a multi-layer neural network to predict the fault type includes extracting features based on the acquired relay protection data, determining the number of input layer nodes based on the selected features, setting the number of output layer nodes based on the fault type, and setting hidden layers for different fault types through a multi-branch structure network.

[0019] Furthermore, the construction of a multi-layer neural network to predict fault types also includes using an integrated power system model to generate virtual fault data, setting different types of faults in the physical model, calculating corresponding electrical parameter change data, and using the electrical parameter change data as a training set to train the multi-layer neural network.

[0020] Furthermore, the introduction of a distributed computing framework for constant value verification calculation includes calling the power system integration model based on the fault type obtained by intelligent prediction, and combining the fault point and power supply connection relationship determined by the topology model, calculating the current protection constant according to the short-circuit current calculation principle, and then calculating the measured impedance from the fault point to the protection installation, comparing the measured impedance with the distance protection constant, and finally calculating the differential current and braking current based on the current on both sides of the protected equipment.

[0021] In a second aspect, a power plant relay protection setting automatic calibration system includes:

[0022] The data acquisition module is configured to: acquire relay protection data, including electrical parameters, device parameters and power topology information;

[0023] A topology module is configured to: construct a topology model of the power network using the power topology structure information, including setting the electrical equipment of the power plant as nodes and the lines connecting the equipment as edges using a graph theory-based method;

[0024] An electrical model module is configured to: construct a component-based electrical model based on electrical parameters and device parameters, including constructing the electrical model using Park transformation and T-type equivalent circuit;

[0025] The conversion module is configured to: combine the topology model with the electrical model by using node parameter association to obtain an integrated power system model;

[0026] The fixed value verification module is configured to perform fixed value verification calculations based on the integrated power system model, including building a multi-layer neural network to predict fault types and introducing a distributed computing framework for fixed value verification calculations;

[0027] The output module is configured to generate a report based on the fixed value verification calculation results.

[0028] In a third aspect, the present invention provides a computer-readable storage medium storing a plurality of instructions, wherein the instructions are suitable for being loaded and executed by a processor of a terminal device, for example, a method for automatic calibration of relay protection settings of a power plant.

[0029] In a fourth aspect, the present invention provides a terminal device comprising a processor and a computer-readable storage medium, wherein the processor is used to implement various instructions; the computer-readable storage medium is used to store multiple instructions, and the instructions are suitable for being loaded and executed by the processor to implement the method for automatic calibration of relay protection constants of a power plant.

[0030] In summary, the present invention has the following beneficial technical effects:

[0031] 1. The present invention uses Park transformation to construct the dq0 coordinate system electrical model of the generator, and constructs a T-type equivalent circuit model and a line distributed parameter model based on transformer parameters. It can accurately simulate the electrical characteristics of various components of the power system under different working conditions, so that the actual operating status of the power system can be more accurately reflected during the constant value verification calculation, thereby effectively improving the accuracy of the relay protection constant value verification, reducing the constant value misjudgment caused by model errors, and ensuring the safe and stable operation of the power system.

[0032] 2. The present invention constructs a multi-layer neural network to predict the fault type and uses the prediction results to perform constant value verification calculations. By learning a large amount of relay protection data, the neural network can accurately identify the characteristic patterns of different fault types and predict the possible fault types and related electrical parameter change trends in advance.

[0033] 3. The present invention introduces a distributed computing framework for constant value verification calculations, decomposing complex computing tasks into multiple computing nodes for parallel processing. When calculating key parameters for constant value verification such as short-circuit current, measurement impedance, and differential current of large-scale power systems, computing tasks in different regions or of different types can be assigned to multiple computing nodes for simultaneous execution, greatly shortening the calculation time and improving the calculation efficiency of constant value verification.

[0034] 4. In the process of constructing an integrated power system model, the present invention improves the efficiency of model construction and data processing by constructing a topological graph model based on a graph theory method and optimizing the node parameter association method. The topological graph model concisely and intuitively represents the structure of the power network, facilitates the rapid determination of the connection relationship between each component and the electrical parameter transmission path, and reduces the time complexity of model construction and data query.

[0035] 5. The present invention adopts a multi-branch structure network to set hidden layers for different fault types, so that the neural network can perform specialized learning on different types of fault characteristics. The targeted learning and processing methods enable the constant value verification system to better adapt to the actual operating conditions of the power system where multiple fault types coexist and are complex and changeable, thereby enhancing the system's adaptability to different fault scenarios.

[0036] 6. The present invention obtains comprehensive relay protection data, including electrical parameters, equipment parameters and power topology information, and organically combines the topology model constructed based on these data with the electrical model to form an integrated power system model. This data fusion and model integration method fully utilizes information resources from all aspects of the power system, so that the constant value verification calculation is no longer limited to the analysis of a single component or local network, but can be comprehensively evaluated and calculated from a macro perspective of the entire power system. BRIEF DESCRIPTION OF THE DRAWINGS

[0037] Figure 1 The present invention is a schematic diagram of the overall process of a method for automatically checking relay protection settings in a power plant. DETAILED DESCRIPTION

[0038] The present invention will be further described in detail below with reference to the accompanying drawings.

[0039] Example 1

[0040] Reference Figure 1 A method for automatically checking relay protection settings in a power plant according to this embodiment includes:

[0041] Obtain relay protection data, including electrical parameters, equipment parameters, and power topology information;

[0042] Using power topology information to construct a topological graph model of the power network, including using a graph theory-based approach to set electrical equipment in power plants as nodes and lines connecting the equipment as edges;

[0043] Construct component-based electrical models based on electrical parameters and device parameters, including using Parker transformations and T-type equivalent circuits to construct electrical models;

[0044] The topology model and the electrical model are combined by using node parameter association to obtain an integrated power system model.

[0045] Perform fixed value verification calculations based on an integrated power system model, including building a multi-layer neural network to predict fault types and introducing a distributed computing framework for fixed value verification calculations;

[0046] Generate reports based on the fixed value verification calculation results.

[0047] Specifically, the following steps are included:

[0048] like Figure 1 As shown, S1, obtaining relay protection data, including obtaining electrical parameters, equipment parameters and power topology information;

[0049] Responsible for collecting data related to relay protection from various equipment and systems in the power plant, including electrical parameters of the power system (such as voltage, current, power), equipment parameters (transformer capacity, line impedance), set relay protection settings, and topological structure information of the power system. Among them, the topological structure information of the power system includes the connection method and connection path between various electrical equipment through the power line, including the starting device node and the ending device node of the line. For complex power networks, there may be multiple lines connecting the same pair of device nodes or a line branch connecting multiple device nodes. The specific details of the connection, such as the location of the branch point, the phase sequence of the line connection, and other information must be accurately recorded to fully present the line connection topology of the power system.

[0050] S2. constructing a topological graph model of the power network using the power topology information, including using a graph theory-based method to set electrical equipment in power plants as nodes and lines connecting the equipment as edges;

[0051] First, the electrical equipment in the power plant is classified according to its functions, characteristics and voltage levels, and generators, transformers, busbars, circuit breakers, reactors, capacitors and various power lines are identified separately. For each type of equipment, node identifiers are added. For example, for generators, they are marked as "G1", "G2", etc. according to the unit number; transformers are marked as "T1", "T2", etc. according to their installation location or capacity; busbars are marked as "B1", "B2", etc. according to their location and voltage level. This can quickly locate specific equipment nodes when constructing topological models and conducting various power system analyses in the subsequent construction, providing a clear basic framework for model construction.

[0052] When defining nodes, the key parameters of the corresponding equipment are associated with the node information. For the generator node "G1", in addition to basic identification information, detailed electrical parameters such as its rated capacity, rated voltage, synchronous reactance, transient reactance, and damping winding parameters are also attached; transformer nodes such as "T1" cover parameters such as transformation ratio, primary winding resistance and leakage reactance, secondary winding resistance and leakage reactance, excitation resistance and excitation reactance; busbar node "B1" records its voltage level, short-circuit capacity and other information; as a connecting device, the line has its length, unit length resistance, inductance, capacitance, conductance and other line characteristic parameters indicated in the node information at both ends. This allows subsequent electrical calculations, fault analysis, and relay protection setting verification based on the model to directly obtain the required parameters, greatly improving work efficiency and ensuring the integrity and consistency of information.

[0053] Afterwards, the power lines connecting each electrical device are accurately mapped to edges in the topology model, and the starting and ending nodes of each edge are clearly defined, that is, the node identifiers corresponding to the two devices connected by the line. For example, if there is a transmission line connecting the output terminal of generator "G1" to bus "B1", an edge pointing from node "G1" to node "B1" is created in the topology graph. For complex power networks, there are multiple lines connected in parallel, branch lines connecting multiple devices, or ring connections. The connection path of each edge must be accurately depicted to ensure that the topology graph fully reflects the actual connection architecture of the power system.

[0054] In addition to determining the connectivity of edges, each edge must also be assigned corresponding electrical properties. These properties are derived from the physical characteristics of the line itself and its electrical performance during power system operation. For example, the resistance property of an edge is calculated from the line's length and resistance per unit length, reflecting the resistance loss during current transmission. The inductance property is determined by the line's inductance per unit length and length, affecting the electromagnetic induction effect when current changes in the line, as well as the phase relationship between voltage and current. The capacitance property considers the capacitance effect formed between the line and the surrounding environment, affecting the transmission of reactive power and voltage distribution. The conductivity property reflects the insulation performance of the line and reflects leakage.

[0055] Finally, the adjacency matrix method in graph theory is used to store the connection information of the topological graph model. A two-dimensional matrix is constructed. The number of rows and columns of the matrix is equal to the total number of nodes. If there is a connection edge between node i and node j, the element in the i-th row and j-th column of the matrix (and the j-th row and i-th column for undirected graphs) is set to 1, otherwise it is set to 0. For weighted graphs, if the electrical properties of the edge are considered, the position element can be set to the corresponding electrical parameter value (such as resistance, inductance, etc.). The adjacency matrix can represent the connection relationship between nodes in the topological graph, which is convenient for computer storage and fast retrieval. The shortest path algorithm, such as the Dijkstra algorithm or the Floyd-Warshall algorithm, is used to determine the shortest electrical path between any two nodes in the power network. When analyzing short-circuit faults, the shortest path algorithm can quickly determine the shortest flow path of the fault current from the power source to the fault point. Then, combined with the electrical parameters of the line and equipment, the short-circuit current size and distribution can be accurately calculated, providing a key basis for the accurate calibration of relay protection settings. In power dispatching, the shortest path algorithm is used to optimize the power transmission path, reduce line losses, and improve the operating efficiency of the power system.

[0056] S3. Construct component-based electrical models based on electrical parameters and device parameters, including using Parker transformation and T-type equivalent circuit to construct electrical models;

[0057] According to the data obtained by S1 and the collected parameters are cleaned and verified, the rationality between the parameters is checked using the built-in verification module of the professional power system analysis software. For example, based on the rated power, rated voltage and rated current of the generator, verify whether its power factor is within a reasonable range; based on the theoretical relationship between synchronous reactance, transient reactance and excitation winding reactance, check for parameter entry errors and ensure the accuracy of each parameter, laying the foundation for subsequent accurate modeling based on Park transformation. Based on the collected real-time monitoring values of the stator three-phase winding voltage and current (assuming U a 、U b 、U c and,I a , I b , I c , use the Parker transformation formula to transform it to the dq0 coordinate system. The Parker transformation formula is as follows:

[0058]

[0059] Where θ is the rotor position angle, which is calculated based on the generator speed ω1 and time t U a 、U b and U c They represent the real-time monitoring values of the phase voltages of the three-phase stator windings (phase a, phase b, and phase c), and I a , Ib and I c They are the real-time monitoring values of the phase currents of the three-phase stator windings. The real-time updated θ value ensures that the Park transformation can dynamically track the operating status of the generator and accurately reflect the changes in the direct-axis and quadrature-axis components of the three-phase winding electrical quantities in the rotating coordinate system.

[0060] The dynamic electrical model of the generator is constructed with the voltage equation in the dp0 coordinate system as the core. The voltage equation is expressed as:

[0061]

[0062] Among them, R a is the stator resistance, U d and U q are the direct-axis and quadrature-axis components of the stator three-phase winding voltage in the rotating coordinate system, I d and I q are the direct axis and quadrature axis components of the stator current in the dq0 coordinate system, ψ d and ψ q They are the direct-axis and quadrature-axis flux linkages in the dq0 coordinate system, U0 and I0 are the zero-sequence voltage component and zero-sequence current component, respectively, and ψ0 is the zero-sequence symmetrical component of the three-phase stator winding flux linkage.

[0063] In this model, the magnetic flux ψ d , ψ q There is a complex coupling relationship between ψ0 and current and voltage. By introducing the reactance parameters of the generator (such as synchronous reactance, transient reactance, etc.) and the excitation system model (describing the relationship between excitation current and magnetic flux), the electrical characteristics of the generator under different operating conditions (such as steady-state operation, transient disturbance, and fault state) are simulated. For example, at the moment of a three-phase short circuit fault, the model combined with the operating parameters before the short circuit can quickly predict the sudden change of the generator output current, providing a key basis for the accurate calibration of the relay protection setting.

[0064] Detailed transformer parameters are also collected from multiple data sources at the power plant. These parameters include the transformation ratio, which determines the conversion ratio between the primary and secondary voltages of the transformer; the primary winding resistance and leakage reactance, which reflect the resistance loss of the primary winding when current passes through it and the reactance characteristics caused by leakage magnetic phenomena; the secondary winding also has corresponding resistance and leakage reactance parameters, and its function is similar to that of the primary winding, except that it is located in a different circuit position; there are also excitation resistance and excitation reactance. The excitation resistance mainly reflects the equivalent resistance characteristics of the transformer core loss, while the excitation reactance characterizes the reactance characteristics required to establish the core magnetic field.

[0065] After obtaining these parameters, a T-type equivalent circuit model is constructed. In this model, the primary side loop equation is: U1=I1(R1+jX1)+U′ 2, where U1 represents the primary side voltage, I1 is the primary side current, R1+jX1 is the impedance of the primary winding, and U ′ 2 is the voltage converted from the secondary side to the primary side, X1 is the leakage reactance of the primary winding, j represents the imaginary unit, and R1 is the resistance of the transformer primary winding. For the secondary side circuit equation, U2 = I2(R2+jX2)+U′2, where U2 is the actual voltage on the secondary side, I2 is the secondary side current, R2+jX2 is the impedance of the secondary winding, X2 represents the leakage reactance of the secondary winding, and R2 is the resistance of the transformer secondary winding. Among them, n represents the transformation ratio, the excitation branch current It reflects the current in the excitation branch and depends on the primary side voltage U1 and the excitation resistance R m and magnetizing reactance X m The primary side current I1 is converted from the secondary side current I2 and the excitation branch current I m superposition, that is, I1=I′2+I m , and the secondary side current Through such a set of interrelated equations, a complete T-type equivalent circuit electrical model of the transformer is constructed. This model can accurately simulate the electrical characteristics of the transformer under different operating conditions (such as different loads, different input voltages, etc.), providing a reliable transformer electrical behavior simulation basis for subsequent power system analysis work such as relay protection setting verification.

[0066] Finally, to construct the line distributed parameter model, we first need to read the line-related parameters from the temporary database. The line length is a key parameter that directly affects the overall electrical characteristics of the line. At the same time, we also need to obtain parameters such as unit length resistance, unit length inductance, unit length capacitance, and unit length conductance. These parameters describe the basic electrical properties of the line per unit length.

[0067] For a line with a length of L, its series impedance Z = r0L + jωl0L, where r0L is the total resistance part of the line, which determines the resistance loss when the line current passes through, jωl0L is the total inductive reactance part of the line, ω is the angular frequency, L0 is the inductance per unit length of the line, the inductive reactance will hinder the current change in the line and affect the phase relationship between voltage and current, and the parallel admittance Y = g0L + jωc0L, where g0L represents the total conductance of the line, which reflects the leakage characteristics of the line, jωc0L is the total capacitive susceptance part of the line, the capacitive susceptance will affect the reactive power characteristics and voltage distribution of the line, g0 is the conductance per unit length of the line, and c0 is the capacitance per unit length of the line, which is the capacitance formed between the conductor of the line and the surrounding ground or other conductors.

[0068] Based on the distributed parameter model equations of these lines, the voltage and current transmission characteristics of the lines at different locations are deeply calculated. Through equivalent circuit analysis of different positions of the lines, combined with Kirchhoff's law circuit analysis method, the distribution changes of current along the line and the voltage drop at different nodes can be determined, thereby constructing a complete line electrical model. The construction of the entire electrical model is completed by combining the dynamic electrical model of the generator and the T-type equivalent circuit electrical model of the transformer.

[0069] S4. Combining the topology model with the electrical model using node parameter association to obtain an integrated power system model;

[0070] The power network topology model, constructed based on graph theory, uses electrical equipment as nodes and connecting lines as edges. It intuitively presents the physical architecture of the power system, showing the location of each device in the system and how they are connected. In the topology, node identifiers clearly identify the connection between the generator and the step-up transformer, and the connection layout between the busbar and multiple outgoing lines, providing a visual framework for accurately locating electrical connections. The electrical connections between nodes determined in the topology, and the start and end nodes corresponding to each connecting edge, directly correspond to the energy transmission paths between different components in the electrical model, allowing the relevant electrical parameters to be transferred along these paths to construct a complete electrical description of the system.

[0071] As the link for power transmission, the series impedance and shunt admittance of a line are important factors affecting the characteristics of power transmission. Based on the connection relationship revealed by the topological model, the series impedance of the line, Z = r0L + x0L, is transferred to the electrical model of the node connected to the line. This means that the impact of the line's resistive losses and inductive reactance on the flow of current into and out of the node is considered at the node. For example, for a transmission line connecting a generator and a transformer, its series impedance is transferred to the electrical model of the generator node. When the generator outputs power, it needs to overcome the voltage drop caused by this part of the line impedance, which affects the generator's terminal voltage and output current characteristics. Transferring it to the transformer's primary node model will change the phase relationship and amplitude of the voltage and current input to the transformer.

[0072] Similarly, the line's shunt admittance (Y = g0L + jωc0L) is transferred to the node electrical model, accounting for the impact of line leakage and capacitive reactive characteristics on the node voltage. Taking a busbar node as an example, the combined shunt admittance of multiple lines connected to the busbar will alter the busbar's reactive power balance, thereby affecting the busbar voltage level. This transfer eliminates the need for isolated node electrical models, fully accounting for the electrical characteristics of connected lines and better aligning them with actual power system operation.

[0073] Then, based on Kirchhoff's current law, for any node in the power system, the sum of the currents flowing into the node is equal to the sum of the currents flowing out of the node. In the process of combining the topological model with the electrical model, by clarifying the electrical connection relationship between nodes, the node current equation is established based on KCL. For example, at a substation bus node with multiple generators, transformers, and multiple outgoing lines, all the branches connected to it are determined according to the topological map, and then the current balance equation of the bus node can be written using KCL, that is, ∑ k I k,in =∑ m I m,out , where I k,in is the branch current flowing into the busbar, I m,out is the current of each branch flowing out of the bus, k represents the branch used to mark the branch flowing into the node, and m represents the branch used to mark the branch flowing out of the node. This equation links the current of each branch (whose size is related to the electrical model parameters and operating status of the electrical components connected to the branch), reflects the current conservation relationship at the node, and provides a constraint condition for solving the operating status of the power system.

[0074] According to Kirchhoff's voltage law, in any closed loop, the sum of the voltage drops across each component is equal to the power source electromotive force. When constructing an integrated power system model, the KVL is used to establish node voltage equations based on the loop information (a closed path consisting of nodes and connecting edges) in the topology model. For example, a simple closed loop containing a generator, a transformer, a transmission line, and a load is selected. Following the loop direction, based on the electrical models of each component (such as the voltage equation of the generator, the loop equation of the transformer, and the series impedance characteristics of the line), ∑ n U n =E, where, U n is the voltage drop across each component in the circuit, E is the source electromotive force, and n denotes the number of components. This equation takes into account the voltage characteristics of each component in the circuit. Combined with the KCL equations, by solving them simultaneously, we can determine electrical parameters such as the voltage at each node in the system and the current in each branch, enabling accurate analysis and simulation of the entire power system's operating status.

[0075] S5. Perform fixed value verification calculations based on the integrated power system model, including building a multi-layer neural network to predict fault types and introducing a distributed computing framework for fixed value verification calculations;

[0076] First, the collected raw relay protection data is rigorously cleaned to remove data points with obvious errors or anomalies. This includes voltage and current data that exceeds reasonable ranges (e.g., voltage values that suddenly jump to extreme values and do not conform to the normal operation logic of the power system), as well as outliers caused by sensor failures or data transmission errors. Furthermore, a subset of data highly relevant to fault type prediction is selected, and redundant data with no significant impact on fault diagnosis is removed to improve data quality and reduce the waste of computing resources, ensuring the accuracy, reliability, and effectiveness of the data used for neural network training.

[0077] Based on the physical principles of power systems and fault diagnosis experience, representative and distinguishing features are extracted from the cleaned and filtered relay protection data. For example, the voltage unbalance feature is calculated by the formula

[0078]

[0079] Among them, U a 、U b 、U c They are three-phase voltage, U avg The three-phase voltage average value is used to measure the degree of imbalance of the three-phase voltage. This feature often changes significantly in fault types such as single-phase grounding faults. The current harmonic content feature is extracted and the harmonic components in the current signal are analyzed through signal processing methods such as Fourier transform. Different types of faults (such as nonlinear load faults, power electronic equipment faults, etc.) may lead to the increase of specific frequency harmonics. The power factor change rate feature can also be calculated, that is, Among them, PF t is the power factor at the current moment, PF t―1 is the power factor at the previous moment, Δt is the time interval, and abnormal changes in the power factor may indicate reactive power imbalance or equipment failure in the power system.

[0080] A feature evaluation algorithm is used to evaluate the importance of the numerous extracted features. This embodiment employs methods such as the chi-square test to calculate the degree of correlation between each feature and the fault type. Based on the evaluation results, key features that contribute significantly to fault type prediction are selected, while features with weaker correlation or redundancy are discarded. Assuming that, after evaluation, features such as voltage imbalance, current harmonic content, power factor change rate, and spectral energy in specific frequency bands are identified as key features, these features will serve as input features for the neural network. These features can accurately reflect the operating status of the power system and effectively distinguish different types of faults, thereby improving the prediction accuracy and efficiency of the neural network.

[0081] The number of input layer nodes is determined based on the selected key features. Since features such as voltage imbalance, current harmonic content, power factor change rate, and spectral energy in specific frequency bands are selected as input, the number of input layer nodes is four. Each input node corresponds to a specific eigenvalue. The preprocessed relay protection data features are input into the neural network according to the corresponding node, providing a data foundation for neural network learning.

[0082] The number of output layer nodes is set according to the fault type. Hidden layers are set for different fault types through a multi-branch structure network. Each branch has an independent hidden layer structure. Taking the common three-layer hidden layer as an example, in the first hidden layer of the short-circuit fault branch, assuming that the number of neurons is n, the input feature vector is:

[0083] x=[ΔI sc ,ΔU sc ,ρ z ],

[0084] Among them, ΔI sc Short-circuit current mutation, ΔU sc Short-circuit voltage drop, ρ Z The line impedance change rate is calculated by linearly combining the input feature vector with the weight matrix and adding the bias vector to obtain an intermediate vector. This introduces nonlinear factors, enabling the network to learn complex relationships. The intermediate vector is activated using the ReLU activation function to obtain the output of the first hidden layer. Subsequent hidden layers proceed in this manner, continuously abstracting and integrating features to gradually extract higher-level, more discriminative feature representations, until the output layer provides information that can be used for fault diagnosis.

[0085] Before the output layer of the network, the results of each branch need to be fused. A common and effective method is weighted summation. Assume that there are m branches in total, and the output of the i-th branch before the output layer is o i , the corresponding weight is w i The formula for calculating the fused output is: When there are short-circuit fault branches and voltage abnormality fault branches, the short-circuit fault branch output o1 contributes more to the short-circuit fault probability estimation and is assigned a larger weight w1; the voltage abnormality fault branch output o2. Similarly, through weighted summation, the fault feature information learned by each branch is comprehensively considered, so that the final output O can fully reflect the various possible fault type tendencies of the power system. At the output layer, by comparing with the threshold or further probabilistic processing, an accurate fault type prediction is given.

[0086] Finally, the output layer nodes are set. The number of output layer nodes is set according to the fault type that needs to be predicted. If four states, such as three-phase short circuit, two-phase short circuit, single-phase ground short circuit, and normal operation, are to be predicted, four nodes are set in the output layer. The output value of each output node represents the probability of the corresponding fault type. By analyzing and judging the probability values of the output layer nodes, the most likely fault type of the power system can be determined.

[0087] 2. According to the fault type and possible short-circuit current change trend obtained by intelligent prediction, the generator Parker transformation model, transformer T-type equivalent circuit model and line distribution parameter model in the power system integration model are used, combined with the connection relationship between the fault point and the power source determined by the power network topology model. Based on the power system short-circuit current calculation principle, the symmetrical component method is used to carry out short-circuit current calculation for different fault types (three-phase short circuit, two-phase short circuit, single-phase ground short circuit). For the three-phase short-circuit current calculation, the formula is used. Among them, E s Expressed as the power supply electromotive force, Z s Expressed as the system equivalent impedance, Z k Expressed as the impedance from the fault point to the power supply, E s 、Z s and Z k The determination of is adjusted according to the prediction results. For example, if it is predicted that the output voltage of the generator may fluctuate to a certain extent under the current working conditions, the value of will be corrected according to the predicted fluctuation range.

[0088] Two-phase short-circuit current calculation formula And the calculation formula of single-phase ground short-circuit current The relevant parameters in are also adjusted in the same way, where Z0 represents the zero-sequence impedance, Z1 represents the positive-sequence impedance, and Z2 represents the negative-sequence impedance.

[0089] Then, according to the action principle of current protection, combined with the calculated short-circuit current and the set current protection value, determine whether the value meets the sensitivity and selectivity requirements. The sensitivity requirement Among them, I k.min is the minimum short-circuit current at the end of the protection zone, I op Expressed as operating current, K sen.set The sensitivity setting value is used. If it is not satisfied, an alarm message will be issued. Based on the predicted fault type, the measured impedance from the fault point to the protection installation is calculated. Among them, U m Indicates the voltage measured at the protection installation location, I m Indicates the measured current at the protection installation, and for three-phase short-circuit faults, the measured impedance. mand I m When considering the predicted trend of electrical parameter changes, for example, if the line voltage drop near the fault point is large, then U m The calculation will be corrected by combining the line distributed parameter model and the predicted voltage drop.

[0090] The calculated measured impedance is compared with the fixed value of the distance protection (such as the operating impedance and operating time), while taking into account the influence of the transition resistance. A correction method based on the fault distance measurement algorithm is used to compensate for the influence of the transition resistance on the measured impedance. This embodiment adopts a fault distance measurement algorithm based on the traveling wave method. According to the predicted fault type and electrical parameter changes, the calculation of parameters such as the traveling wave propagation speed is adjusted to more accurately determine the influence of the transition resistance on the measured impedance and make corrections. If the measured impedance falls within the operating zone of the distance protection and the operating time meets the selectivity requirements, the fixed value check passes; otherwise, an alarm is issued.

[0091] According to the intelligent predicted fault conditions, the differential current and braking current are calculated. The differential current calculation formula is I d =|I1―I2|, braking current calculation formula Among them, I1 and I2 are the currents on both sides of the protected equipment. During the calculation, accurate calculation is performed based on the transformer and generator models in the power system integration model and the impact of the predicted fault on the currents on both sides of the equipment.

[0092] Then, according to the action characteristic curve of the differential protection (usually the ratio braking characteristic curve, the curve formula is expressed as I d =K r I r +I op0 ) to determine whether the calculated differential current and braking current meet the action conditions. If the differential current exceeds the action threshold under the given braking current, there may be a problem with the set value and an alarm message will be issued.

[0093] The entire set value verification calculation task is decomposed according to the area or equipment type of the power system. For example, the set value verification tasks of different voltage level bus areas of the power plant are assigned to different computing nodes. Taking computing node i as an example, the task assigned to it can be expressed as Task i , which contains the model data of the power system part related to the task (such as the topological structure information of the area, equipment parameters, etc.) and the relay protection setting information that needs to be calibrated.

[0094] S6. Generate a report based on the calculation results of the fixed value verification;

[0095] After completing the calibration calculations for all types of relay protection settings, the results are output and managed. First, all result data stored during the calibration calculation process is read from the result database, including the calibration results (pass or fail) for each protection setting, the detailed reasons for failure (such as insufficient sensitivity, improper action time coordination, measurement impedance deviation, etc.), and related calculation data (short-circuit current, measurement impedance, differential current, etc.). Then, this data is organized and a detailed report is generated. The report content covers all the above information so that operators can quickly and accurately adjust and optimize the relay protection settings based on these suggestions, ultimately completing the entire power plant relay protection setting automatic calibration process to ensure the safe and stable operation of the power system.

[0096] Example 2

[0097] This embodiment differs from the first embodiment in that it provides a power plant relay protection setting automatic calibration system, including:

[0098] The data acquisition module is configured to: acquire relay protection data, including electrical parameters, device parameters and power topology information;

[0099] A topology module is configured to: construct a topology model of the power network using the power topology structure information, including setting the electrical equipment of the power plant as nodes and the lines connecting the equipment as edges using a graph theory-based method;

[0100] An electrical model module is configured to: construct a component-based electrical model based on electrical parameters and device parameters, including constructing the electrical model using Park transformation and T-type equivalent circuit;

[0101] The conversion module is configured to: combine the topology model with the electrical model by using node parameter association to obtain an integrated power system model;

[0102] The fixed value verification module is configured to perform fixed value verification calculations based on the integrated power system model, including building a multi-layer neural network to predict fault types and introducing a distributed computing framework for fixed value verification calculations;

[0103] The output module is configured to generate a report based on the fixed value verification calculation results.

[0104] A computer-readable storage medium stores a plurality of instructions, wherein the instructions are suitable for being loaded and executed by a processor of a terminal device, and a method for automatically checking relay protection settings of a power plant is disclosed.

[0105] A terminal device includes a processor and a computer-readable storage medium, wherein the processor is used to implement various instructions; the computer-readable storage medium is used to store multiple instructions, and the instructions are suitable for being loaded and executed by the processor to implement the method for automatically calibrating the relay protection setting of a power plant.

[0106] The above are all preferred embodiments of the present invention, and are not intended to limit the scope of protection of the present invention. Therefore, any equivalent changes made based on the structure, shape, and principle of the present invention should be included in the scope of protection of the present invention.

Claims

1. A method for automatically checking relay protection settings in a power plant, characterized in that: include: Obtain relay protection data, including electrical parameters, equipment parameters, and power topology information; Using power topology information to construct a topological graph model of the power network, including using a graph theory-based approach to set electrical equipment in power plants as nodes and lines connecting the equipment as edges; Construct component-based electrical models based on electrical parameters and device parameters, including using Parker transformations and T-type equivalent circuits to construct electrical models; The topology model and the electrical model are combined by using node parameter association to obtain an integrated power system model. Perform fixed value verification calculations based on an integrated power system model, including building a multi-layer neural network to predict fault types and introducing a distributed computing framework for fixed value verification calculations; Generate reports based on the fixed value verification calculation results.

2. The method for automatic calibration of relay protection settings in a power plant according to claim 1, characterized in that: The component-based electrical model is constructed according to the electrical parameters and device parameters, including obtaining the generator parameters in the device parameters, converting the electrical quantities of the stator three-phase winding to the dp0 coordinate system according to the Park transformation formula, and constructing the electrical model of the generator based on the voltage equation in the dp0 coordinate system.

3. The method for automatic calibration of relay protection settings in a power plant according to claim 1, characterized in that: The method of constructing a component-based electrical model based on electrical parameters and device parameters further includes obtaining transformer parameters from the device parameters, calculating the primary-side loop equation and the secondary-side loop equation respectively based on the transformer parameters, generating a T-type equivalent circuit based on the side loop equations, then obtaining the line length, and calculating the series impedance and parallel admittance of the line. Based on the distributed parameter model equation of the line, the voltage and current transmission characteristics of the line at different positions are calculated to construct an electrical model of the line.

4. The method for automatic calibration of relay protection settings in a power plant according to claim 1, characterized in that: The method of combining the topology model and the electrical model by associating node parameters includes determining the electrical connection relationship between each node through the topology model, transferring the series impedance and parallel admittance in the line to the electrical model of the node connected to the line according to the connection relationship, and establishing the relationship equation between the node voltage and the branch current based on Kirchhoff's law.

5. The method for automatic calibration of relay protection settings in a power plant according to claim 4, characterized in that: The method of constructing a multi-layer neural network to predict the fault type includes extracting features based on the acquired relay protection data, determining the number of input layer nodes based on the selected features, setting the number of output layer nodes based on the fault type, and setting hidden layers for different fault types through a multi-branch structure network.

6. A method for automatic calibration of relay protection settings in a power plant according to claim 5, characterized in that: The construction of a multi-layer neural network to predict fault types also includes using an integrated power system model to generate virtual fault data, setting different types of faults in the physical model, calculating corresponding electrical parameter change data, and using the electrical parameter change data as a training set to train the multi-layer neural network.

7. A method for automatic calibration of relay protection settings in a power plant according to claim 6, characterized in that: The introduction of a distributed computing framework for constant value verification calculation includes calling the power system integration model based on the fault type obtained by intelligent prediction, and combining the fault point and power supply connection relationship determined by the topology model, calculating the current protection constant value according to the short-circuit current calculation principle, then calculating the measured impedance from the fault point to the protection installation, comparing the measured impedance with the distance protection constant value, and finally calculating the differential current and braking current based on the current on both sides of the protected equipment.

8. A power plant relay protection setting automatic calibration system, which executes the method according to claim 1, characterized in that: include: The data acquisition module is configured to: acquire relay protection data, including electrical parameters, device parameters and power topology information; A topology module is configured to: construct a topology model of the power network using the power topology structure information, including setting the electrical equipment of the power plant as nodes and the lines connecting the equipment as edges using a graph theory-based method; An electrical model module is configured to: construct a component-based electrical model based on electrical parameters and device parameters, including constructing the electrical model using Park transformation and T-type equivalent circuit; The conversion module is configured to: combine the topology model with the electrical model by using node parameter association to obtain an integrated power system model; The fixed value verification module is configured to perform fixed value verification calculations based on the integrated power system model, including building a multi-layer neural network to predict fault types and introducing a distributed computing framework for fixed value verification calculations; The output module is configured to generate a report based on the fixed value verification calculation results.

9. A computer-readable storage medium storing a plurality of instructions, characterized in that: The instructions are suitable for being loaded by a processor of a terminal device and executed by a method for automatically checking relay protection settings of a power plant as claimed in claim 1.

10. A terminal device comprising a processor and a computer-readable storage medium, wherein the processor is configured to implement various instructions; and the computer-readable storage medium is configured to store a plurality of instructions, wherein: The instructions are suitable for being loaded by a processor and executed by a method for automatically checking relay protection settings of a power plant as claimed in claim 1.

Images