Relay protection setting method for distribution system considering new energy output fluctuation

Abstract

The application discloses a power distribution system relay protection setting method considering new energy output fluctuation, and relates to the technical field of power distribution systems, comprising: collecting power distribution network topology structure data and distributed new energy power prediction data in real time, and constructing a multi-scenario set reflecting output fluctuation characteristics; adopting interval fault calculation method to obtain amplitude boundary and direction characteristics of potential fault point short-circuit current; taking protection selectivity, sensitivity and rapidity as constraint conditions, establishing time-current constant value collaborative optimization model; adopting multi-agent collaborative calculation architecture to online calculate adaptive protection setting value group matched with current fluctuation working condition; and issuing adaptive protection setting value group to power distribution intelligent terminal through time-sensitive network and performing rolling update. The application realizes real-time tracking of new energy output fluctuation, maximizes protection coverage range, takes into account sensitivity and rapidity, and eliminates misoperation risk caused by value rigidity of traditional offline setting.

Description

Technical Field

[0001] This invention relates to the field of power distribution system technology, and more specifically to a method for setting relay protection in a power distribution system that takes into account the fluctuations in power output from new energy sources. Background Technology

[0002] With the deepening of the global energy structure transformation, distributed renewable energy generation technologies, represented by photovoltaics and wind power, have made significant progress and have been widely applied in power distribution network systems. Distributed renewable energy generation has outstanding advantages such as being clean, environmentally friendly, and renewable. Its large-scale grid connection effectively reduces carbon emissions and promotes the green and low-carbon transformation of the power system. However, the output characteristics of distributed renewable energy differ fundamentally from those of traditional synchronous generators. Its power generation is significantly intermittent and fluctuating due to the influence of natural resource conditions such as sunlight intensity and wind speed. This uncertainty brings new challenges to the safe and stable operation of the power distribution network.

[0003] Among them, the distribution network relay protection system, as the first line of defense for ensuring the safe operation of the power grid, has the core function of quickly and reliably isolating fault sections and minimizing the power outage area. Traditional distribution network relay protection setting methods are usually based on fixed maximum and minimum operating modes for offline setting, with protection settings remaining constant over a long period of time. Accurate identification and rapid isolation of various faults are achieved through the coordinated operation of selectivity, sensitivity, and speed. However, in distribution networks with a high proportion of distributed renewable energy integration, due to the intermittent fluctuations in renewable energy output, the system power flow distribution and the amplitude and direction of short-circuit currents are all dynamically changing. Traditional protection setting strategies based on fixed operating modes are difficult to adapt to such rapidly changing operating conditions.

[0004] Existing technologies have conducted extensive research and made some progress in relay protection for distributed renewable energy access, but shortcomings remain: First, traditional protection setting methods employ offline analysis, resulting in long protection setting update cycles and an inability to track real-time fluctuations in renewable energy output, leading to inaccurate protection range or loss of protection selectivity under certain operating conditions. Second, existing technologies often employ conservative maximum or minimum estimation methods to address uncertainties in renewable energy output, sacrificing the sensitivity and efficiency of the protection system while prioritizing system safety. Third, there is a lack of refined modeling methods to address the impact of renewable energy low-voltage ride-through characteristics on transient short-circuit currents, and the accuracy of existing fault calculation models is insufficient to meet the requirements of adaptive protection setting. Fourth, existing adaptive protection schemes are highly dependent on communication systems and master stations, with long calculation and decision-making cycles, making it difficult to achieve millisecond-level rapid responses. In summary, with the development trend of high proportion of distributed renewable energy being integrated into the distribution network, how to enable the relay protection system to follow the fluctuations in renewable energy output in real time and dynamically adjust the protection settings, and eliminate the risk of false tripping or failure to trip caused by rigid settings, has become a key technical problem that urgently needs to be solved to ensure the safe and reliable operation of the distribution system. Summary of the Invention

[0005] The present invention provides a method for setting relay protection in a power distribution system that takes into account the fluctuation of power output from new energy sources, which can solve the above-mentioned problems.

[0006] To solve the above problems, the technical solution adopted by the present invention is as follows:

[0007] A method for setting relay protection in a distribution system considering the fluctuation of renewable energy output includes: real-time acquisition of distribution network topology data and distributed renewable energy power prediction data to construct a multi-scenario set reflecting the fluctuation characteristics of renewable energy output; fault analysis of each scenario in the multi-scenario set using an interval fault calculation method to obtain the amplitude boundary and direction characteristics of the short-circuit current at each potential fault point in the distribution network; establishing a time-current setting collaborative optimization model with protection selectivity, sensitivity, and speed as constraints, and maximizing the protection coverage and minimizing the overall system protection action time as the objective function; based on the amplitude boundary and direction characteristics of the short-circuit current and the time-current setting collaborative optimization model, using a multi-agent collaborative computing architecture to calculate an adaptive protection setting group matching the current fluctuation condition online; and distributing the adaptive protection setting group to the distribution intelligent terminal through a time-sensitive network, and updating the protection setting on a rolling basis according to the changes in renewable energy output.

[0008] Furthermore, the method for constructing the multi-scenario set specifically includes: dividing the distributed renewable energy power prediction data into high-output scenarios, medium-output scenarios, and low-output scenarios, and determining the power fluctuation range corresponding to each scenario; using Latin hypercube sampling technology to extract sample points from each power fluctuation range, and mapping the sample points to the corresponding power output values ​​through the inverse transformation method, thereby generating the multi-scenario set covering the entire range of renewable energy output.

[0009] Furthermore, the method for determining the amplitude boundary and direction characteristics of the short-circuit current at each potential fault point in the distribution network specifically includes: based on the node admittance matrix of the distribution network, treating the voltage of the potential fault point as an unknown quantity, treating the injected current at the potential fault point as a disturbance quantity, representing the fluctuation range of the injected power from new energy sources as an interval number and substituting it into the injected current column vector of the node admittance matrix; applying boundary conditions according to the superposition principle and fault type, and solving the node voltage interval vector and injected current interval vector through chain propagation of interval mathematical operations to obtain the upper and lower bounds of the amplitude of the short-circuit current at each potential fault point and the phase angle variation range; calculating the distribution of positive-sequence current, negative-sequence current, and zero-sequence current in the network based on the sequence network graph decomposition method, identifying reverse power flow scenarios and bidirectional power flow scenarios, and establishing a short-circuit current direction characteristic matrix.

[0010] Furthermore, the constraints include:

[0011] Selective constraint: For any node that experiences a fault, the time difference between the action time of its main protection and the action time of its backup protection must satisfy a preset time level setting.

[0012] Sensitivity constraint: The operating current setting value of the protection device is less than or equal to the ratio of the minimum fault current lower limit value to the preset sensitivity coefficient.

[0013] Speed-of-action boundary constraint: Under no circumstances shall the protection action time exceed the limit clearing time allowed by the protected equipment or system.

[0014] Furthermore, the multi-agent collaborative computing architecture includes a master station agent, sub-station agents, and terminal agents. The online computing process is divided into a periodic decision-making mode and an event-driven decision-making mode. In the periodic decision-making mode, according to a preset time period, the master station agent sends a scenario update command to the sub-station agent. The sub-station agent performs local calculations and reports the results to the master station agent. The master station agent summarizes the calculation results of the entire network and performs a global optimization calculation, and sends the obtained adaptive protection setting group to the terminal agent through the sub-station agent. In the event-driven decision-making mode, when the rate of change or cumulative change of the aggregated output of new energy is detected to exceed a preset threshold, a fast optimization calculation is directly performed based on the stored short-circuit current boundary database.

[0015] Furthermore, the setting issuance command is encrypted using a combination of national cryptographic algorithms and employs a dual confirmation mechanism requiring the sending of a first confirmation frame and a second confirmation frame to ensure the reliability of the transmission. The rolling update adopts a sliding window model, evaluating the output fluctuation within the current window at each sliding moment. When the short-circuit current boundary change rate at any protection installation location exceeds a preset threshold, the setting of that protection and its cooperating protection group is triggered to be updated.

[0016] Furthermore, the method of the present invention also includes the step of incorporating the low voltage ride-through characteristics of new energy sources into the fault equivalent model: based on the low voltage ride-through capability curve and real-time voltage drop characteristics of distributed new energy power generation equipment, the active current injection and reactive current injection of the equipment under specific voltage drop conditions are calculated; the active current injection and reactive current injection are converted into additional injection terms and superimposed on the original deterministic injection current in the interval fault calculation model to form a corrected node injection current column vector, and the short-circuit current interval boundary is recalculated accordingly.

[0017] In the process of calculating active current injection and reactive current injection, a piecewise linear model is used to describe the photovoltaic inverter: when the system voltage drop depth is within the first preset range, the photovoltaic inverter outputs the active component of the rated current and the reactive component that can be additionally injected; when the system voltage drop depth is lower than the lower limit of the first preset range, the photovoltaic inverter only outputs reactive current.

[0018] This technical solution overcomes the limitations of traditional deterministic calculation methods by incorporating the low-voltage ride-through characteristics of distributed renewable energy power generation equipment into the fault equivalent model in a refined manner. It fully considers the active response characteristics and additional injected current of renewable energy equipment during faults, and corrects the calculation of transient short-circuit currents. This method makes the boundary calculation results of the fault current range more closely resemble real operating conditions, significantly improving the accuracy of protection setting.

[0019] Furthermore, the method of the present invention also includes the step of introducing an edge computing architecture: the distribution smart terminal is configured with a local computing unit and an energy storage unit, the local computing unit runs a simplified fault calculation algorithm using an equivalent simplified model and pre-calculation lookup table technology; when a fault occurs in the distribution network, the distribution smart terminals on both sides of the fault point simultaneously detect the fault signal, and the local computing unit completes fault identification, short-circuit current calculation and direction determination locally within a preset millisecond time. If the direction determination is positive, a trip command is output; if the direction determination is negative, a blocking signal is output.

[0020] This technical solution introduces an edge computing architecture, where the power distribution intelligent terminal performs fault identification, short-circuit current calculation, and direction determination locally. This mechanism decentralizes complex computing tasks, reducing reliance on centralized computing and communication networks at the master station. This enables the protection system to output tripping or blocking commands within milliseconds, perfectly meeting the extremely stringent requirements of the power distribution system for protection action speed.

[0021] Furthermore, the method of the present invention also includes a step of online evaluation of the effectiveness of the setting: based on the actual fault recording data and protection action records, the action accuracy rate, selective satisfaction rate and sensitivity index of the protection device action within the evaluation period are statistically analyzed; the action accuracy rate, selective satisfaction rate and sensitivity index are compared with their respective preset thresholds, and the adaptive adjustment of parameters in the time-current setting collaborative optimization model or the triggering of a local recalculation process is initiated based on the comparison results.

[0022] This technical solution introduces an online evaluation mechanism for the effectiveness of setpoints. It uses actual fault recording data to post-evaluate the accuracy of actions, selective satisfaction rate, and sensitivity indicators, and triggers adaptive adjustment or local recalculation of the optimization model parameters accordingly. This closed-loop self-evolution mechanism enables the optimization model to continuously improve based on actual operating experience, ensuring the safe and reliable operation of the distribution network under a high proportion of new energy access in the long term.

[0023] Compared with the prior art, the beneficial effects of the present invention are:

[0024] (1) This invention addresses the problems of traditional relay protection being based on fixed operating modes and having long update cycles through offline setting. It constructs a multi-scenario set by real-time collection of distribution network topology data and distributed renewable energy power prediction data, and combines a multi-agent collaborative computing architecture and a time-sensitive network to form a closed-loop rolling update mechanism. The protection settings can be adaptively adjusted in a timely manner according to the intermittent fluctuations in renewable energy output and changes in operating conditions, completely eliminating the risks of inaccurate protection range, false tripping, or failure to trip caused by rigid settings in traditional offline setting methods.

[0025] (2) Existing technologies for handling output uncertainty often employ conservative maximum or minimum value estimations, which significantly sacrifice sensitivity while ensuring safety. This invention constructs a time-current setpoint collaborative optimization model, incorporating time-level setting and current setpoint setting into a unified optimization framework. With protection selectivity, sensitivity, and speed as multiple constraints, it maximizes the protection coverage and minimizes the action time through multi-objective collaborative weighting. This achieves the comprehensive optimization of protection system sensitivity and action efficiency while ensuring system safety.

[0026] To make the above-mentioned objects, features and advantages of the present invention more apparent and understandable, embodiments of the present invention are described below in detail with reference to the accompanying drawings. Attached Figure Description

[0027] To more clearly illustrate the technical solutions of the embodiments of the present invention, the accompanying drawings used in the embodiments will be briefly introduced below. It should be understood that the following drawings only show some embodiments of the present invention and should not be regarded as a limitation on the scope. For those skilled in the art, other related drawings can be obtained based on these drawings without creative effort.

[0028] Figure 1 This is a flowchart of the relay protection setting method for a power distribution system that considers the fluctuation of new energy output, as described in an embodiment of the present invention;

[0029] Figure 2 This is a schematic diagram illustrating the core principle of the time-current setpoint collaborative optimization model in this embodiment of the invention;

[0030] Figure 3 This is a schematic diagram of the multi-level interaction relationship and data flow between the main station agent, sub-station agent and terminal agent in an embodiment of the present invention. Detailed Implementation

[0031] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are some embodiments of the present invention, but not all embodiments.

[0032] Example 1

[0033] This invention provides a method for setting relay protection in a distribution system that considers the fluctuations in renewable energy output. It revolves around five core aspects: real-time data acquisition, scenario-based modeling, interval calculation, optimization decision-making, and dynamic updating. Through systematic technical means, it achieves adaptive following of protection settings to the fluctuations in renewable energy output. The following is in conjunction with… Figure 1 The detailed implementation process of each step is explained.

[0034] Step 1: Establish a set of scenarios for fluctuation ranges in new energy power output

[0035] The implementation of Step 1 relies primarily on real-time acquisition devices and mechanisms for the distribution network's overall data. In the real-time acquisition phase of the distribution network topology data, the system collects fundamental parameter data reflecting the network topology structure through phasor measurement units, synchronous phasor measurement devices, and intelligent sensor groups with time synchronization capabilities deployed at substations and key nodes. Specifically, this includes: the reference voltage level, rated capacity, and equivalent impedance parameters of all buses; conductor type, impedance per unit length, capacitive reactance, allowable thermal stability current, and actual line length for each level of overhead lines; rated capacity, turns ratio, winding connection group, zero-sequence impedance parameters, and short-circuit voltage percentage for two-winding and three-winding transformers; and the current opening / closing status, action signals, and location feedback information of all circuit breakers and disconnectors in the entire network. This topology data is mapped into standardized data objects using the IEC 61850 communication protocol and synchronized to the data processing center at a 60Hz refresh rate, ensuring that the underlying network model used in subsequent calculations remains dynamically consistent with the actual system state.

[0036] In the real-time acquisition phase of distributed renewable energy power prediction data, the system establishes data interfaces with the monitoring and data acquisition systems of photovoltaic power plants, wind farm control systems, and distributed photovoltaic aggregator platforms to receive short-term and ultra-short-term power prediction curves from photovoltaic power plants and wind farms. The short-term power prediction curve has a prediction time domain of 15 minutes to 24 hours in the future, with a time resolution of no less than 15 minutes. It includes the active power prediction value output by the photovoltaic power plant based on the irradiance numerical weather prediction model and the active power prediction value output by the wind farm based on the wind speed probability density function and wind power curve conversion model. The ultra-short-term power prediction curve has a prediction time domain of 0 to 15 minutes in the future, with a time resolution of no less than 1 minute, used to capture minute-level power output fluctuations. The system simultaneously collects the current power output value, prediction error statistical characteristics, and power output confidence interval boundary data of each renewable energy plant.

[0037] In the real-time acquisition of electrical quantities at key system nodes, the phasor measurement unit synchronously acquires the positive-sequence component amplitude and phase angle information of each bus voltage at a sampling rate of 60 frames per second, with an accuracy requirement of 0.1 degree phase angle error and 0.5% amplitude error. The system synchronously acquires the active power, reactive power, apparent power, current amplitude, and phase angle of each line. The injected power of the distributed renewable energy generation unit is acquired through the secondary side signal of the current transformer of its grid-connected inverter, including active power injection value, reactive power injection value, and inverter DC side voltage and current values. After data cleaning and bad data identification, the above electrical quantity data are stored in a real-time database for subsequent processing.

[0038] Based on power prediction data and electricity quantities, step 1 employs a probabilistic interval analysis method to construct a multi-scenario set reflecting the power output fluctuation characteristics of new energy sources. This analysis process first divides the power output levels into hierarchical levels according to the confidence intervals of the power prediction curves. Specifically, for each new energy power station or aggregation area, based on the statistical distribution characteristics of its historical power output data and the time-series autocorrelation of ultra-short-term prediction errors, the high-output boundary value and low-output boundary value of its predicted power are determined. Using the per-unit value of the predicted power as the abscissa, the difference between the high-output boundary and the predicted power, and the difference between the low-output boundary and the predicted power, are used to determine the width of the high-output interval and the width of the low-output interval, respectively. Based on this, three baseline levels—high-output scenario, medium-output scenario, and low-output scenario—are defined. Each power output scenario corresponds to a power fluctuation interval, defined as the upper limit of the 95% confidence interval for the high-output scenario, the median of the 50% confidence interval for the medium-output scenario, and the lower limit of the 95% confidence interval for the low-output scenario. The system further refines the boundaries of the fluctuation range for each level, taking into account the uncertainty of power prediction and intraday fluctuation characteristics. The fluctuation range width for high-output scenarios is set to ±15% of the predicted power, the fluctuation range width for medium-output scenarios is set to ±10% of the predicted power, and the fluctuation range width for low-output scenarios is set to ±5% of the predicted power.

[0039] To generate scenario samples, the system uses Latin hypercube sampling to extract sample points from each fluctuation range. This sampling method constructs an N-dimensional parameter space using the power output of renewable energy power plants as random variables, where N is the number of renewable energy power plants or aggregation areas participating in scenario generation. The sampling process first divides the cumulative distribution function of each random variable into K equally probable intervals, where K is the preset total number of scenarios, typically set to an integer value between 100 and 500 to balance computational accuracy and efficiency. Then, within each interval, a sample point is randomly selected based on a uniform distribution, and the sample point is mapped to the corresponding power output value using an inverse transformation method. A correlation control matrix is ​​introduced during the sampling process to ensure that the sampling results are statistically consistent with the actual correlation structure of renewable energy power output. Finally, a scenario sample set covering the entire range of renewable energy power output is generated, with each sample containing the active power injection value, reactive power injection value, and corresponding probability weights of all renewable energy power plants.

[0040] Step 2: Calculate the short-circuit current interval boundaries

[0041] Step 2 employs the interval fault calculation method to perform fault analysis on each scenario in the scenario set generated in Step 1, in order to quickly determine the amplitude boundaries and directional characteristics of the short-circuit current under different scenarios. This calculation process first establishes an interval fault calculation model for the distribution network. This model represents the uncertainty of renewable energy output as an interval variable, and the injected power of traditional generators and loads as deterministic variables. Interval mathematical operations are then used to handle the impact of the uncertainty of renewable energy output on the short-circuit calculation results.

[0042] The interval fault calculation method analyzes faults at each potential fault point in the distribution network. The selection of potential fault points is based on the system topology, covering the midpoints, one-third points, two-thirds points of all lines, the bus side, and transformer ports. Fault types include two baseline types: three-phase short-circuit faults and single-phase-to-ground short-circuit faults. For each potential fault point and each fault type, the system executes the following calculation process: Based on the node admittance matrix of the distribution network, the voltage at the fault point is considered an unknown quantity, and the injected current at the fault point is considered a disturbance quantity; according to the superposition principle, the node voltage under fault conditions is equal to the superposition of the pre-fault state voltage and the fault component voltage; the fault component voltage is obtained by solving a system of linear equations, and its coefficient matrix is ​​a correction matrix of the node admittance matrix; corresponding boundary conditions are applied according to the fault type: during a three-phase short circuit, the three-phase voltage at the fault point is zero; during a single-phase-to-ground short circuit, the fault phase voltage at the fault point is zero, and the zero-sequence current satisfies the grounding impedance constraint.

[0043] The interval mathematical operation introduces an uncertainty handling mechanism based on the aforementioned deterministic calculation process. It substitutes the fluctuation range of the injected power from new energy sources into the fault calculation process, replacing the fixed injection value in traditional methods. In the injected current column vector of the node admittance matrix, the elements corresponding to the new energy power station are represented in interval number form. The addition, subtraction, multiplication, and division operations of the interval numbers adopt the interval endpoint propagation rule. The interval division operation transforms the nonlinear problem into a linear interval problem by introducing auxiliary variables. Through the chain propagation of interval mathematical operations, the interval representation of the short-circuit current at each fault point is finally obtained, including the upper limit of the amplitude, the lower limit of the amplitude, and the phase angle variation range. The upper limit of the amplitude corresponds to the maximum short-circuit current that may occur at the fault point under the scenario of high new energy output, and the lower limit of the amplitude corresponds to the minimum short-circuit current that may occur at the fault point under the scenario of low new energy output.

[0044] The specific mathematical expression for interval fault calculation is as follows, establishing the nodal admittance matrix equation containing interval variables:

[0045]

[0046] In the formula, The corrected node admittance matrix taking into account network parameters and fault boundary conditions; For unknown node voltage interval vectors; Inject current interval vectors into nodes.

[0047] For nodes containing new energy power stations Its injected current elements are represented in interval number form:

[0048]

[0049] In the formula, and These correspond to the lower and upper bounds of the injection current in the low-output and high-output new energy scenarios defined in step 1, respectively. Through chain propagation of interval mathematical operations to invert or iteratively solve, the short-circuit current interval representation for each potential fault point k is finally obtained. and the range of phase angle variation .

[0050] Step 2 further analyzes the directional characteristics of the short-circuit current, based on the sequence network diagram decomposition method. This method decomposes the fault network into three independent networks: a positive-sequence network, a negative-sequence network, and a zero-sequence network. For each scenario sample, the system calculates the distribution of positive-sequence, negative-sequence, and zero-sequence currents within the network. The calculation process employs the symmetrical component method, converting the three-phase currents into three sequence components: positive-sequence, negative-sequence, and zero-sequence. The calculation of the positive-sequence current is based on the network equations of the positive-sequence network, while the calculations of the negative-sequence and zero-sequence currents are based on the network equations of the negative-sequence and zero-sequence networks, respectively. The network parameters are dynamically updated according to the system's operating state before the fault, taking into account the impact of fluctuations in renewable energy output on the equivalent impedance of each sequence network.

[0051] The system identifies reverse power flow scenarios and bidirectional power flow scenarios based on the distribution patterns of positive-sequence, negative-sequence, and zero-sequence currents. Reverse power flow scenarios refer to situations where the power flow direction in some branches of the system is opposite to the normal direction. In this case, the directional characteristics of the short-circuit current may conflict with the judgment results of the protection directional elements based on the power flow direction setting. The system establishes a short-circuit current directional feature matrix. The row indices of this matrix correspond to the installation locations of each protection unit, and the column indices correspond to each fault scenario. Matrix elements record the phase angle range of the current flowing through the protection unit's installation location relative to the positive direction of the protection setting under a specific fault scenario. When the phase angle range crosses the setting threshold of the direction criterion, the matrix element is marked as bidirectional. The directional feature matrix is ​​used for adaptive adjustment of the directional element setting values ​​in the subsequent optimization model.

[0052] Step 3: Construct a time-current setpoint co-optimization model

[0053] Combination Figure 2 As shown, step 3 establishes a time-current setting co-optimization model with protection selectivity, sensitivity, and speed as constraints and maximizing protection coverage as the objective function. The core design concept of this model is to incorporate time-level setting and current setting into a unified optimization framework, achieving a comprehensive improvement in protection performance through co-optimization.

[0054] The model uses protection selectivity as the primary constraint. Selectivity requires that the protection device closest to the fault point operates first when a fault occurs, and the tripping range is controlled within the minimum fault interval. The constraint is that for any adjacent protection coordination pair, the operating time of the protection device must satisfy a time difference constraint with the operating time of the coordinating protection. This time difference value is determined according to the protection type; for coordination between instantaneous overcurrent protection and overcurrent protection, the time difference is set to 0.3 to 0.5 seconds; for coordination between overcurrent protection and the next-level overcurrent protection, the time difference is set to 0.2 to 0.3 seconds. The time difference constraint ensures that the upstream protection will not operate beyond its cascade when a fault occurs, and that the downstream protection can trip before the upstream protection operates.

[0055] The model uses sensitivity as an auxiliary constraint. Sensitivity requires the protection device to have reliable operating sensitivity to the minimum fault current. The sensitivity constraint is that the operating current setting of the protection must be less than or equal to the ratio of the minimum fault current to the sensitivity coefficient. The value of the sensitivity coefficient is determined according to the protection type and the importance of the protected equipment. For instantaneous overcurrent protection, the sensitivity coefficient is set to 1.2 to 1.5; for overcurrent protection, the sensitivity coefficient is set to 1.5 to 2.0. This constraint ensures that the protection device can still operate reliably when a metallic short circuit occurs under the minimum operating conditions of the system.

[0056] The model constructs a multi-objective comprehensive optimization function that balances coverage and speed. Specifically, the optimization objectives include two dimensions: first, maximizing the protection coverage (defined as the percentage of the total line length that the protection device can protect); and second, pursuing optimal speed (i.e., minimizing the overall protection action time of the system). During the optimization process, weighting coefficients are introduced to synergistically weight these two objectives, and the system synchronously adjusts the current setting and time setting parameters of each protection device to achieve comprehensive optimal protection performance across the short-circuit current range of all scenario samples. The introduction of this multi-objective optimization mechanism overcomes the shortcomings of traditional conservative estimation methods that rely on reducing the protection range to ensure selectivity, while also avoiding unnecessary extensions in protection action time.

[0057] In addition, the model adds speed boundary constraints to ensure that the protection action time in any fault scenario will not exceed the limit tolerance time of the equipment or system.

[0058] This objective is achieved by adjusting the parameters of the time-current matching curve family. The matching curves adopt inverse time-limit characteristic curves, and their mathematical expression is:

[0059]

[0060] in, Indicates the duration of the action. Indicates fault current. This indicates the current setting reference value. Represents the time constant. and The parameter is the curve shape. The advantage of this inverse-time characteristic curve is that it can adaptively adjust the action time according to the magnitude of the fault current: the larger the fault current, the shorter the action time, achieving rapid clearing of near-end faults; the smaller the fault current, the longer the action time, with more coordination stages, ensuring selectivity.

[0061] Based on the above concept of collaborative optimization, the following multi-objective mathematical optimization model is established:

[0062] (1) Objective function:

[0063] To balance maximizing protection coverage with minimizing protection action time (speed objective), a multi-objective comprehensive optimization function is constructed:

[0064]

[0065] In the formula, To comprehensively optimize the objective function; A collection of power distribution network protection devices; A set of preset fault scenarios; For protection devices The actual length of the line that can be protected; For protection devices The total length of the route; For protection devices In failure scenarios The action time is determined by the inverse time-limit curve equation. and These are the weighting coefficients.

[0066] (2) Constraints:

[0067] Selective constraint (time difference): For any node that experiences a fault, the main protection action time... Action time of backup protection (superior protection) The following conditions must be met:

[0068]

[0069] In the formula, The preset time difference setting value is (0.3s~0.5s for instantaneous overcurrent protection and overcurrent protection combined, and 0.2s~0.3s for overcurrent protection combined).

[0070] Sensitivity constraints:

[0071]

[0072] In the formula, For protection devices The operating current setting value; The minimum fault current lower bound value is obtained in combination with the low output scenario of new energy (derived from the interval result in step 2). The specified sensitivity coefficient (e.g., 1.2~1.5 for instantaneous overcurrent).

[0073] Motionality boundary constraints:

[0074] Under no circumstances shall the protection action time exceed the limit withstand time for equipment thermal stability or power grid transient stability.

[0075]

[0076] In the formula, The maximum allowable cut-off time for the protected equipment or system.

[0077] Step 4: Online decision-making for adaptive protection setting groups

[0078] Step 4, based on the short-circuit current interval boundary database established in Step 2 and the time-current setting collaborative optimization model established in Step 3, calculates online the adaptive protection setting group that matches the current fluctuating operating conditions. This decision-making process adopts a multi-agent collaborative computing architecture to achieve distributed processing and collaborative decision-making of computing tasks.

[0079] like Figure 3 As shown, the multi-agent collaborative computing architecture comprises three levels of agent units. The master agent is deployed in the distribution network dispatch and control center, configured with a high-performance computing server cluster, and is responsible for global optimization calculation tasks. The master agent receives data aggregation results from substation agents and terminal agents, and executes the global solution of the time-current setpoint collaborative optimization model. Substation agents are deployed in the integrated automation systems of each substation, configured with multi-core processors and dedicated digital signal processing chips, and are responsible for local setpoint calculation tasks. Substation agents, based on the protection device configuration and regional network topology within their respective substations, execute the solution of the local optimization model and coordinate boundaries with agents of adjacent substations. Terminal agents are deployed in distribution intelligent terminals, including line protection measurement and control devices, ring main unit intelligent units, and pole-mounted switch controllers, and are responsible for real-time data acquisition and setpoint issuance command execution.

[0080] Data interaction between agents is achieved through a high-speed communication network. Communication between the master agent and sub-agents uses a custom application layer protocol built on top of the TCP / IP protocol stack, with a data transmission cycle of 1 second, including network topology update data, protection setting data, and optimization result upload data. Communication between sub-agents and terminal agents uses a lightweight message queue protocol developed based on languages ​​such as Golang, with a data transmission cycle of 100 milliseconds, including real-time electrical quantity acquisition data and setting confirmation data. Communication between terminal agents and sensors uses the IEC 61850 MMS protocol, with a data sampling rate set to 4.8kHz, i.e., 96 samples per cycle.

[0081] The online decision-making process is divided into two modes: periodic decision-making and event-driven decision-making. Periodic decision-making executes according to a preset time period, typically set to once every 15 minutes, consistent with the update cycle of ultra-short-term power prediction. The execution flow of periodic decision-making is as follows: the master station agent sends a scenario update command to the sub-station agent; upon receiving the command, the sub-station agent triggers partial execution of steps 1 and 2, and reports the calculation results to the master station agent; the master station agent aggregates the calculation results across the entire network and triggers global optimization calculation in step 3, distributing the optimized setpoint group to the terminal agent through the sub-station agent; the terminal agent performs the setpoint switching operation and reports the execution results. Event-driven decision-making is triggered when a sudden change in renewable energy output is detected. The mutation criterion is that the rate of change in renewable energy aggregate output exceeds 5% per minute or the cumulative change exceeds 20% of the predicted output. Event-driven decision-making employs an accelerated calculation process, skipping the partial optimization steps and directly performing rapid optimization calculations based on the stored short-circuit current boundary database, compressing the setpoint switching cycle to within 30 seconds.

[0082] Step 5: Issue and continuously update protection settings

[0083] Step 5 involves sending the adaptive protection setting group calculated in Step 4 to the distribution smart terminal via the communication network, and updating the protection setting on a rolling basis according to the changes in the output of new energy sources, forming a closed-loop mechanism of prediction-calculation-decision-update.

[0084] The communication network adopts a Time-Sensitive Networking Protocol (TSN) based on Industrial Ethernet. This protocol introduces a time synchronization mechanism and a time-aware scheduling algorithm on top of standard Ethernet to ensure that the end-to-end latency of data transmission meets preset real-time requirements. The synchronization accuracy of the TSN is better than 1 microsecond, and the time-aware scheduling algorithm allocates a dedicated transmission window for the protection setpoint traffic, with a window scheduling period set to 1 millisecond. The key performance indicators of the communication network are: end-to-end transmission latency of the setpoint command not exceeding 5 milliseconds, packet loss rate not exceeding 0.001%, and jitter not exceeding 0.1 milliseconds.

[0085] The fixed-value issuance command employs encrypted transmission and a dual-confirmation mechanism to ensure the reliability and security of transmission. Encrypted transmission uses a combination of the Chinese national cryptographic algorithms SM2 / SM3 / SM4. SM2 is used for digital signature and key negotiation of the fixed-value data, SM3 for message integrity verification, and SM4 for symmetric encryption of the fixed-value data. The dual-confirmation mechanism requires the terminal agent to send a first confirmation frame upon receiving the fixed-value issuance command, containing the verification result of the fixed-value data; and a second confirmation frame after completing the fixed-value writing operation, containing the read echo data from the fixed-value writing area. Both confirmation frames are sent to the substation agent via different transmission paths. The absence of either confirmation frame or verification failure triggers a fixed-value retransmission process.

[0086] The rolling update mechanism dynamically adjusts the update frequency and range of protection settings based on changes in renewable energy output. The rolling update strategy employs a sliding window model with a window width of 15 minutes and a sliding step size of 1 minute. At each sliding moment, the system assesses the output fluctuation within the current window. When the fluctuation exceeds a preset threshold, a setting update is triggered. The evaluation metric for fluctuation is the rate of change of the short-circuit current boundary. When the rate of change of the short-circuit current boundary at any protection installation location exceeds 10%, it is determined that the settings of that protection and its cooperating protection group need to be updated. During the rolling update process, the system employs seamless switching technology to ensure uninterrupted protection function during setting switching. Specifically, this is achieved through alternating writing to dual buffers and hot-standby setting area switching.

[0087] In a preferred embodiment, the present invention further includes the step of incorporating the low-voltage ride-through characteristics of new energy sources into the fault equivalent model. Based on the low-voltage ride-through capability curve of distributed new energy power generation equipment, a fault equivalent circuit for new energy sources considering low-voltage ride-through characteristics is established to correct and calculate the transient short-circuit current, thereby obtaining a more accurate fault current waveform and effective value and improving the setting accuracy.

[0088] Low-voltage ride-through parameters are acquired through a data interface established with the renewable energy power plant monitoring system. The system collects low-voltage ride-through capability curves for each distributed renewable energy power generation device, which describe the device's ability to maintain grid connection during grid voltage dips. The curve data includes: voltage dip threshold, defined as the per-unit system voltage at which the device begins to enter low-voltage ride-through mode, typically set to 0.9 times the rated voltage; voltage dip depth, defined as the ratio of the lowest point of the system voltage during a fault to the rated voltage; grid disconnection delay, defined as the time interval from the occurrence of a voltage dip to the device detecting grid disconnection conditions, typically set to 0.1 to 2 seconds; and power recovery time, defined as the time it takes for the device to recover from low-power operation to rated power after the fault is cleared, typically set to 0.5 to 5 seconds.

[0089] The calculation of fault ride-through current injection characteristics is based on low-voltage ride-through parameters and real-time voltage sag characteristics. The calculation process first determines the voltage sag depth and duration based on the voltage sag curve at the protection installation point during a fault. Then, it determines the active and reactive current injection of the equipment under a given voltage sag condition based on the low-voltage ride-through capability curve. The low-voltage ride-through characteristics of mainstream photovoltaic inverters adopt a piecewise linear model with a constant current region and a constant power region: when the voltage sag depth is between 0.9 and 0.45 times the rated voltage, the inverter outputs the active component of the rated current and can inject an additional 0.4 times the rated current of reactive current; when the voltage sag depth is below 0.45 times the rated voltage, the inverter only outputs reactive current, the magnitude of which is calculated based on the voltage sag depth. The low-voltage ride-through characteristics of wind turbine doubly-fed induction generators employ a control strategy coordinated with rotor Crowbar protection. During a fault, the rotor-side converter is locked, and the generator operates as an induction motor. The attenuation characteristics of the injected short-circuit current are determined based on the rotor resistance and time constant.

[0090] Specifically, the short-circuit current injection characteristics (including active current) of mainstream photovoltaic inverters during faults and reactive current The following piecewise model is used to describe this:

[0091]

[0092]

[0093] In the formula, This represents the actual system voltage per unit value (i.e., voltage drop depth). This is the inverter's rated current; and These are the rated active current and initial reactive current before the fault, respectively; The reactive current support function is determined according to the grid connection guidelines under deep drop conditions. This represents the maximum allowable overload current of the inverter. The above analytical solution is converted into additional injection terms and superimposed on the interval fault calculation model.

[0094] The specific implementation of superimposing the fault ride-through current injection characteristics onto the traditional fault equivalent model is as follows: During the interval fault calculation in step 2, an additional injection term representing the fault ride-through current of the new energy equipment is added to the node injection current column vector. The calculation of the additional injection term uses the real-time updated voltage drop curve and low-voltage ride-through parameters, and the actual injection current of the new energy equipment during the fault period is determined through iterative calculation. This injection current is superimposed on the original deterministic injection current to form a corrected node injection current column vector, based on which the fault current interval boundary is recalculated. The corrected transient current calculation results can reflect the characteristic of the new energy equipment actively adjusting the output current according to the grid voltage change during the fault period, making the setting results more consistent with the actual operating conditions.

[0095] In a preferred embodiment, the present invention further includes the step of introducing an edge computing architecture, in which the power distribution terminal completes part of the computing tasks locally, reducing the communication and computing dependence on the master station and achieving a response speed that meets preset requirements.

[0096] The edge computing architecture adopts a hierarchical computing model, distributing computing tasks to different levels of processing units according to real-time requirements and computational complexity. The power distribution terminal, as the core node of the edge computing, is configured with local computing units and energy storage units. The local computing unit uses an embedded processor based on the ARM Cortex-A72 architecture or a high-performance FPGA chip, with a main frequency of no less than 1.5GHz and at least four computing cores, supporting floating-point operations and parallel computing. The local computing unit runs a simplified fault calculation algorithm, which, while maintaining computational accuracy, compresses the computation time to the millisecond level through model simplification and pre-calculation lookup table techniques. The implementation strategies for the simplified algorithm include: using an equivalent simplified model of the power distribution network to simplify the complex multi-branch network into an equivalent two-port network; pre-calculating the short-circuit current boundaries of each fault point under different output scenarios and storing them in a local database, directly obtaining the results through table lookup and linear interpolation during fault occurrence; and using approximate numerical methods to replace precise iterative calculations, such as using a simplified estimation formula for Thevenin equivalent impedance.

[0097] The energy storage unit provides uninterrupted power to the terminal, using lithium iron phosphate battery packs as the energy storage medium. Its capacity configuration is determined based on the terminal's rated power consumption and backup power time requirements. During normal operation, the energy storage unit is in float charge mode. When an external power interruption is detected, it automatically switches to discharge mode to provide continuous power to the local computing unit and communication module. The capacity configuration of the energy storage unit must meet the requirement that the terminal can still independently perform basic protection functions in the event of a communication interruption. These basic protection functions include fault detection, fault direction determination, and backup delayed tripping. The backup power time of the energy storage unit is set to 2 hours, during which the terminal can maintain protection functions at a reduced sampling rate and calculation frequency.

[0098] The introduction of edge computing architecture enables millisecond-level response. When a fault occurs in the distribution network, the fault signal is transmitted to the distribution terminal via sensors and merging units. The local computing unit of the terminal completes fault identification within 5 milliseconds of receiving the fault signal, short-circuit current calculation and direction determination within 10 milliseconds, and outputs a trip command within 15 milliseconds. This response time meets the stringent requirements of the distribution system for protection action speed and is superior to centralized architectures that rely entirely on master station calculations.

[0099] In a preferred embodiment, the present invention further includes an online evaluation step of the setting validity. Based on actual fault waveform data and protection action records, the issued adaptive protection settings are post-evaluated to analyze the action accuracy, selectivity and sensitivity indicators of the settings under real fault scenarios, and the optimization model parameters are adaptively adjusted based on the evaluation results.

[0100] The data sources for online assessment include the operation reports of the fault recording device and the operation records of the protection device. The fault recording device records waveform data of various electrical quantities during the fault at a sampling rate of not less than 6.4kHz. The recorded content includes three-phase voltage, three-phase current, and switch status change information for 10 cycles before and after the fault and 20 cycles after the fault. The operation records of the protection device include protection start time, operation time, tripping phase, fault location results, and intermediate quantity status information within the protection device.

[0101] The method for evaluating the accuracy of protective actions is as follows: All fault events that occurred during the evaluation period are statistically analyzed, and the relationship between the protection's action determination and the actual fault location is compared for each event. Events where the action determination matches the actual fault location are counted as correct actions, and events where the action determination does not match the actual fault location are counted as incorrect actions. The formula for calculating the accuracy of protective actions is: the number of correct actions divided by the total number of actions. The evaluation period is typically set to one month or 100 action events, whichever comes first.

[0102] The selection index is evaluated as follows: for each fault event, analyze whether the protection device closest to the fault point operates first, and whether the upstream protection device operates beyond its cascade. The selection index is expressed as the selection satisfaction rate, which is the number of fault events that meet the selection requirements divided by the total number of fault events. The sensitivity index is evaluated as follows: calculate the actual fault current during the fault period based on the fault recording data, compare it with the operating current setting value of the protection device, and evaluate the protection's response capability to the minimum fault current.

[0103] The specific quantitative calculation formula for the effectiveness assessment of the set value is as follows:

[0104] Action accuracy :

[0105]

[0106] In the formula, The number of correct action events that are determined to be consistent with the actual fault location within the evaluation period; This represents the total number of activation events of the protection device.

[0107] Selective satisfaction rate :

[0108]

[0109] In the formula, To meet the requirement of the number of fault events for which the nearest protection device takes priority and the superior protection does not escalate to the next higher level; This represents the total number of failure events. Based on the comparison results of the above indicators with preset thresholds, the system triggers adaptive adjustments or partial recalculations of the optimization model parameters.

[0110] The evaluation results are used to adaptively adjust the model parameters. When the action accuracy rate is lower than a preset threshold, the system analyzes the cause. If the cause is identified as model parameter deviation, a parameter correction process is triggered; if the cause is identified as insufficient scenario coverage, a scenario set expansion process is triggered. When the selectivity satisfaction rate is lower than a preset threshold, the system analyzes the action time difference between the protection pairs. If the difference is insufficient, the optimized model is re-solved. If the sensitivity index decreases, the system analyzes whether the fault current boundary database needs to be updated. If so, the partial recalculation process in step 2 is triggered. The parameter adaptive adjustment mechanism ensures that the optimized model can be continuously improved based on actual operating experience, thereby improving the accuracy and reliability of protection settings.

[0111] The technical solution of this invention achieves adaptive following of protection settings to fluctuations in renewable energy output through the above steps. In step 1, the system collects real-time panoramic data of the distribution network and uses Latin hypercube sampling technology to construct a scene set covering the entire range of renewable energy output, ensuring the completeness and representativeness of the input data for subsequent calculations. In step 2, the system uses the interval fault calculation method to obtain the amplitude boundary and direction characteristics of the short-circuit current, incorporating the uncertainty of renewable energy output into the fault analysis framework, breaking through the limitations of traditional deterministic calculation methods. In step 3, the system constructs a time-current setting collaborative optimization model, unifying protection selectivity, sensitivity, and speed into the optimization objectives, achieving a comprehensive improvement in protection performance. In step 4, the system adopts a multi-agent collaborative computing architecture to achieve distributed processing of computing tasks, ensuring the real-time nature of setting updates through a combination of periodic decision-making and event-driven decision-making. In step 5, the system ensures the reliability and security of setting transmission through time-sensitive networking and a double confirmation mechanism, and achieves dynamic optimization of the settings through a rolling update mechanism. Furthermore, by incorporating the low-voltage ride-through characteristics of new energy sources into the fault equivalent model, the transient current calculation was corrected, further improving the setting accuracy. By introducing an edge computing architecture, millisecond-level response was achieved, meeting the stringent requirements of the power distribution system for protection action speed. Through the online evaluation mechanism for setting effectiveness, continuous improvement of optimized model parameters was realized.

[0112] Example 2

[0113] This embodiment describes a specific application example of 10kV feeder protection setting in a distribution network based on the method of the present invention.

[0114] A certain 10kV distribution network comprises three power substations, 12 main feeders, and several branch lines. The main feeders are overhead insulated conductors, with a total length of approximately 200km. The protection devices for the main feeders employ digital current protection and control devices, configured with a three-stage current protection function. The first stage is instantaneous overcurrent protection, with the operating current set to avoid the maximum three-phase short-circuit current at the end of the line, and the operating time set to instantaneous. The second stage is time-limited instantaneous overcurrent protection, with the operating current set to coordinate with the instantaneous overcurrent protection of adjacent lines, and the operating time set to 0.3 seconds. The third stage is overcurrent protection, with the operating current set to avoid the maximum load current, and the operating time set to be adjustable from 1.0 second to 2.0 seconds. This distribution network connects to two 10MW photovoltaic power plants and one 5MW wind farm, with the total installed capacity of new energy sources accounting for approximately 30% of the total load of the distribution network.

[0115] The process of implementing the method of this invention for this distribution network is as follows. First, in step 1, the system collects 10kV bus voltage, line power flow distribution, and switch status information of three substations through the distribution network dispatch automation system. Both photovoltaic power plants and wind farms are equipped with power prediction systems. The system receives their short-term prediction curves and ultra-short-term prediction curves, with a time resolution set to 5 minutes. Based on historical operating data, the confidence interval width for photovoltaic power output is determined to be ±15% of the predicted value, and the confidence interval width for wind power output is ±20% of the predicted value. Latin hypercube sampling technology is used to generate 200 scenario samples, covering high, medium, and low output operating conditions for both photovoltaic and wind power.

[0116] Then, in step 2, the system performs interval fault calculations for each potential fault point on the 12 main feeders, with a fault point spacing of 1 km, totaling approximately 2400 fault points. The system performs calculations for two fault types: three-phase short circuit and single-phase ground fault, obtaining the upper and lower bounds of the short-circuit current amplitude and the phase angle variation range for each fault point. Through sequence network diagram decomposition analysis, the directional feature matrix is ​​obtained, identifying that when photovoltaic and wind power output is high, reverse power flow exists in local areas of the distribution network, and the short-circuit current direction may be opposite to the positive direction of the protection setting.

[0117] Next, in step 3, the system constructs a time-current setting collaborative optimization model for the protection devices of each main feeder. The model's constraints are: the operating current setting value of the first protection segment is no greater than the ratio of the minimum short-circuit current to 1.5, and the time difference between adjacent protection pairs is no less than 0.3 seconds. The model's optimization objective is to maximize the protection range of each protection segment, where the target protection range value for the first protection segment is 80% of the total line length, and the target protection range value for the second protection segment is 100% of the total line length. Through optimization, the adaptive protection setting set for each feeder is obtained, including the operating current setting value and time setting parameters for each protection segment.

[0118] Then, in step 4, the system employs a multi-agent collaborative computing architecture to execute online decisions. The master agent is deployed at the distribution network dispatch center, the sub-agents are deployed at three substations, and the terminal agents are deployed at the protection devices of 12 feeders. The decision cycle is set to 15 minutes, and the event-driven threshold is set to a change rate of over 5% per minute in the aggregated output of new energy sources. Upon detecting a sudden change in photovoltaic output, the system triggers event-driven decision-making, completes the setting update within 30 seconds, and sends the new protection settings to the protection devices of the relevant feeders.

[0119] Finally, in step 5, the settings are distributed to each protection device via an industrial Ethernet network based on time-sensitive networking. The measured delay for setting distribution is 3.2 milliseconds, and the packet loss rate is 0%. After the settings are updated, the system performs a post-evaluation of the setting validity. Based on the statistical data from the past 6 months of operation, the protection action accuracy rate is 99.2%, the selectivity satisfaction rate is 98.7%, and the sensitivity index meets the requirements. Based on the evaluation results, the parameters of the optimization model were partially adjusted, further improving the protection performance.

[0120] Example 3

[0121] This embodiment describes a distribution network relay protection system architecture and its setting method that includes edge computing nodes.

[0122] This power distribution network introduces an edge computing architecture based on Example 2, deploying edge computing terminals at key nodes. Each edge computing terminal is equipped with a local computing unit and an energy storage unit. The local computing unit uses a quad-core ARM Cortex-A72 processor with a clock speed of 1.8GHz, 2GB of RAM, and a 16GB solid-state drive. The local computing unit comes pre-installed with a simplified fault calculation algorithm, which uses an equivalent simplified model and pre-calculated lookup table technology to control the calculation time for a single short-circuit current to within 2 milliseconds. The energy storage unit uses a 48V 50Ah lithium iron phosphate battery pack, which can provide uninterrupted power to the terminal for 2 hours when the external power supply is interrupted.

[0123] The local computing tasks of the edge computing terminal include: real-time acquisition of secondary side signals from the current transformer and voltage transformer at the installation location, and calculation of the instantaneous and effective values ​​of the local current; performing preliminary fault diagnosis, and determining whether a fault has occurred based on the current surge and zero-sequence current criteria; calculating the phase angle difference in the fault direction, and outputting a direction determination signal based on the comparison result between the phase angle difference and the setting threshold; and performing local setting adjustment or backup protection tripping based on the locally stored settings when communication is interrupted or the master station fails.

[0124] The protection operation process of this architecture is as follows: When a fault occurs in the distribution network, the edge computing terminals on both sides of the fault point simultaneously detect the fault signal. Terminal A's local computing unit completes fault current calculation and direction determination within 5 milliseconds after the fault occurs, and outputs a forward trip command. Terminal B's local computing unit completes fault current calculation and direction determination within 6 milliseconds after the fault occurs. Since the direction determination result is reversed, Terminal B outputs a blocking signal. This process ensures that only the protection action near the fault point trips, while the protection action far from the fault point remains closed, achieving selective protection. The introduction of the edge computing architecture allows the entire protection operation process to be completed independently without the participation of the master station, effectively reducing the impact of communication failures on the reliability of the protection system.

[0125] Through the detailed description of the above embodiments, this invention fully discloses all the technical details of the relay protection setting method for a power distribution system that takes into account the fluctuations in renewable energy output. Those skilled in the art can implement the technical solution of this invention based on these disclosures.

[0126] The above description is merely a preferred embodiment of the present invention and is not intended to limit the invention. Various modifications and variations can be made to the present invention by those skilled in the art. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the scope of protection of the present invention.

Claims

1. A method for setting relay protection in a power distribution system that considers the fluctuation of new energy output, characterized in that, include: Real-time acquisition of distribution network topology data and distributed renewable energy power prediction data; construction of a multi-scenario set reflecting the fluctuation characteristics of renewable energy output. An interval fault calculation method is used to analyze the faults in each scenario of the multi-scenario set, and the amplitude boundary and direction characteristics of the short-circuit current at each potential fault point in the distribution network are obtained. Using protection selectivity, sensitivity, and speed as constraints, and maximizing protection coverage and minimizing the overall system protection action time as the objective function, a time-current setting collaborative optimization model is established. Based on the amplitude boundary and direction characteristics of the short-circuit current and the time-current setting collaborative optimization model, a multi-agent collaborative computing architecture is used to calculate an adaptive protection setting group that matches the current fluctuating operating conditions online. The adaptive protection setting group is distributed to the distribution intelligent terminal through a time-sensitive network, and the protection settings are updated on a rolling basis according to changes in renewable energy output. The multi-agent collaborative computing architecture includes a master station agent, a sub-station agent, and a terminal agent. The online computing process is divided into a periodic decision-making mode and an event-driven decision-making mode. The method for constructing the multi-scenario set specifically includes: dividing the distributed new energy power prediction data into high-output scenarios, medium-output scenarios, and low-output scenarios, and determining the power fluctuation range corresponding to each scenario; using Latin hypercube sampling technology to extract sample points from each power fluctuation range, and mapping the sample points to the corresponding power output values ​​through the inverse transformation method, thereby generating the multi-scenario set covering the entire range of new energy power output; The method for determining the amplitude boundary and direction characteristics of the short-circuit current at each potential fault point in a distribution network specifically includes: based on the node admittance matrix of the distribution network, treating the voltage of the potential fault point as an unknown quantity, treating the injected current of the potential fault point as a disturbance quantity, and representing the fluctuation range of the injected power from new energy sources as an interval number and substituting it into the injected current column vector of the node admittance matrix; applying boundary conditions according to the superposition principle and fault type, and solving the node voltage interval vector and injected current interval vector through chain propagation of interval mathematical operations to obtain the upper and lower bounds of the amplitude of the short-circuit current at each potential fault point, as well as the phase angle variation range; calculating the distribution of positive-sequence current, negative-sequence current, and zero-sequence current in the network based on the sequence network graph decomposition method, identifying reverse power flow scenarios and bidirectional power flow scenarios, and establishing a short-circuit current direction characteristic matrix; the upper bound of the amplitude corresponds to the maximum short-circuit current at the fault point under the scenario of high output from new energy sources, and the lower bound of the amplitude corresponds to the minimum short-circuit current at the fault point under the scenario of low output from new energy sources.

2. The method for setting relay protection of a power distribution system considering the fluctuation of new energy output as described in claim 1, characterized in that, The constraints include: Selective constraint: For any node that experiences a fault, the time difference between the action time of its main protection and the action time of its backup protection must satisfy a preset time level setting. Sensitivity constraint: The operating current setting value of the protection device is less than or equal to the ratio of the minimum fault current lower limit value to the preset sensitivity coefficient. Speed-of-action boundary constraint: Under no circumstances shall the protection action time exceed the limit clearing time allowed by the protected equipment or system.

3. The method for setting relay protection of a power distribution system considering the fluctuation of new energy output as described in claim 1, characterized in that, In the periodic decision-making mode, according to the preset time period, the master station agent sends a scene update instruction to the sub-station agent. The sub-station agent performs local calculations and reports the results to the master station agent. The master station agent summarizes the calculation results of the entire network and performs global optimization calculations. The resulting adaptive protection setting group is then sent to the terminal agent through the sub-station agent. In the event-driven decision-making mode, when the rate of change or cumulative change in the aggregated output of new energy is detected to exceed a preset threshold, a fast optimization calculation is performed directly based on the stored short-circuit current boundary database.

4. The method for setting relay protection of a power distribution system considering the fluctuation of new energy output according to claim 1, characterized in that, The setting issuance command is encrypted and transmitted using a combination of national cryptographic algorithms, and a dual confirmation mechanism requiring the sending of a first confirmation frame and a second confirmation frame is adopted to ensure the reliability of the transmission. The rolling update adopts a sliding window model, which evaluates the output fluctuation within the current window at each sliding moment. When the short-circuit current boundary change rate at any protection installation location exceeds a preset threshold, the setting of the protection and its cooperating protection group is triggered to be updated.

5. The method for setting relay protection of a power distribution system considering the fluctuation of new energy output as described in claim 1, characterized in that, The method also includes the step of incorporating the low-voltage ride-through characteristics of new energy sources into the fault equivalent model: based on the low-voltage ride-through capability curve and real-time voltage drop characteristics of distributed new energy power generation equipment, the active current injection and reactive current injection of the equipment under specific voltage drop conditions are calculated. The active current injection and reactive current injection are converted into additional injection terms and superimposed on the original deterministic injection current in the interval fault calculation model to form a corrected node injection current column vector, based on which the short-circuit current interval boundary is recalculated.

6. The method for setting relay protection of a power distribution system considering the fluctuation of new energy output according to claim 5, characterized in that, In the process of calculating active current injection and reactive current injection, a piecewise linear model is used to describe the photovoltaic inverter: when the system voltage drop depth is within the first preset range, the active component of the rated current output by the photovoltaic inverter and the reactive component that can be additionally injected. When the system voltage drop depth is lower than the lower limit of the first preset range, the photovoltaic inverter only outputs reactive current.

7. The method for setting relay protection of a power distribution system considering the fluctuation of new energy output according to claim 1, characterized in that, The method further includes the step of introducing an edge computing architecture: the distribution smart terminal is configured with a local computing unit and an energy storage unit, and the local computing unit runs a simplified fault calculation algorithm using an equivalent simplified model and pre-calculation lookup table technology; when a fault occurs in the distribution network, the distribution smart terminals on both sides of the fault point simultaneously detect the fault signal, and the local computing unit completes fault identification, short-circuit current calculation and direction determination locally within a preset millisecond time. If the direction determination is positive, a trip command is output; if the direction determination is negative, a blocking signal is output.

8. The method for setting relay protection of a power distribution system considering the fluctuation of new energy output according to claim 1, characterized in that, The method further includes an online evaluation step of the setting validity: based on actual fault recording data and protection action records, the action accuracy rate, selective satisfaction rate and sensitivity index of the protection device action within the evaluation period are statistically analyzed; the action accuracy rate, selective satisfaction rate and sensitivity index are compared with their respective preset thresholds, and the adaptive adjustment of parameters in the time-current setting collaborative optimization model or the triggering of a local recalculation process is initiated based on the comparison results.

Images