Risk boundary-based provincial ac-dc grid scenario risk assessment method and system

Abstract

The application discloses a kind of based on risk boundary's provincial AC-DC power grid scene risk assessment method and system, method includes: S1, the city power grid subarea operation scene set of three types of operation scenes including normal operation scene, expected failure scene and extreme fluctuation scene is constructed, and weight is assigned to each scene;S2, the active power demand and reactive power demand under each operation scene of device are calculated, and the comprehensive active demand and comprehensive reactive demand of device are further obtained by weighted summation, and the rated capacity demand of device is determined;S3, the rated capacity demand is sequentially subjected to safety margin configuration and modularization correction, and the final rated capacity of device is obtained.The application can plan normal, failure, extreme multiple operation scenes, and realize the flexible interconnection device sizing method of comprehensive optimization of economy, safety, adaptability.

Description

Technical Field

[0001] This invention belongs to the field of power grid safe operation technology, and in particular relates to a risk assessment method and system for provincial AC / DC power grid scenarios based on risk boundaries. Background Technology

[0002] The safety and stability assessment of power system operation is a core element in ensuring the reliability and economy of the power grid. In current dispatching practices, the most mainstream safety assessment method is based on a deterministic criterion, namely the well-known N-1 criterion. This criterion requires that the power grid maintain stable operation and not cause power outages to users even when subjected to any single component failure. While the N-1 criterion has the advantages of simple judgment and strict enforcement, it essentially ignores the probabilistic differences in the occurrence of random events and the comprehensive consequences they cause, leading to overly conservative dispatching decisions and sacrificing significant economic operating margins. With the high proportion of renewable energy integration and the increasing complexity of AC / DC power grids, this safety criterion is no longer adequate for the modern power grid's need for an optimal trade-off between safety and economy.

[0003] To overcome the limitations of deterministic criteria, a method for quantitative security assessment based on risk values ​​is proposed. The core of this method is to transform power grid security threats into calculable economic losses, achieving a quantitative trade-off between security levels and economic costs. Currently, two main technical approaches exist in this field: one is the Monte Carlo simulation method, which simulates random events such as equipment failures and load fluctuations through random sampling, and combines power flow calculations and stability analysis to output risk indicators such as outage probability and load loss. However, it suffers from strong sampling blindness and slow convergence speed for multiple faults (Nk). The other is the analytical method, which derives the analytical relationship between fault probability and consequences by establishing mathematical models. This method has high computational efficiency, but when dealing with complex characteristics such as AC / DC coupling and the randomness of new energy sources, model simplification can easily lead to a decrease in assessment accuracy.

[0004] Furthermore, the increasing complexity of AC / DC power grids exacerbates the technical challenges of risk assessment. The randomness of power output from a high proportion of renewable energy sources intensifies fluctuations in grid operation, and power interaction between AC / DC systems increases the complexity of fault propagation. These factors all contribute to an exponential increase in the number of scenarios for occasional grid events. Traditional risk assessment methods either overlook some complex scenarios, leading to overly optimistic assessment results, or pursue full scenario coverage, resulting in excessively high computational costs. A particularly critical technical deficiency lies in the lack of a mechanism to quantify the maximum underestimation of risk from unassessed scenarios, thus failing to provide a confidence interval for the assessment results. This invention innovatively constructs a lower bound R for risk. L and risk upper bound R H This is to address this core pain point. Summary of the Invention

[0005] To address the problems mentioned in the background section, this invention discloses a risk assessment method and system for provincial AC / DC power grid scenarios based on risk boundaries. This method can comprehensively address normal, fault, and extreme operating scenarios, while simultaneously achieving refined and economical configuration of device capacity. To achieve the above objectives, the technical solution adopted by this invention is as follows: A risk assessment method for provincial AC / DC power grid scenarios based on risk boundaries, characterized by the following steps: S1, define scenario risk and construct control cost model; define scenario risk R(t) as the expected comprehensive consequence cost of all possible unforeseen events; S2, Constructing the risk boundary; dividing the entire set of incidental events S into precisely evaluated subsets. and unevaluated subset Define and calculate the lower bound of risk R. L and risk upper bound R H ,in: The lower bound of risk R L Based on the already accurately evaluated subset The minimum control cost C(s) and its probability of occurrence P(s) for each random event are calculated. The risk upper bound R H Based on the inaccurately evaluated subset The conservative control cost C for each random event H It is obtained by calculating its occurrence probability P(s); S3, based on the risk lower bound R L and risk upper bound R H Define risk accuracy γ = R L / R H This is used to quantify the confidence level of risk assessment results; S4, iteratively switching the precisely evaluated subset Gradually reduce the lower bound of the risk R L With risk upper bound R H The difference between the two is calculated until the risk accuracy γ meets the preset accuracy requirement, at which point the final risk assessment interval [R] is output. L R H ].

[0006] Furthermore, the scenario risk R(t) is precisely defined as follows:

[0007] Wherein, P(s) is the probability of occurrence of an unplanned event s, and C(s) is the cost of loss of an unplanned event s; the cost of loss of an unplanned event s, C(s), is the sum of the economic cost of load shedding, the cost of generator power adjustment, and the cost of DC power adjustment; the probability of occurrence of a single fault event, P(s), is calculated based on the equipment failure rate λ and the repair rate μ, and the probability of occurrence of Nk-type multiple fault events, P(s), is calculated using a joint probability model.

[0008] Furthermore, step S2 divides the entire set of random events S into precisely evaluated subsets. and unevaluated subset Specifically, it includes: In the initial state, N-1 fault events are selected as the accurately evaluated subset. ; In subsequent iterations, based on the risk contribution, failure events N-2, N-3, and up to Nk are gradually removed from the inaccurately evaluated subset. Move into the precisely evaluated subset .

[0009] Furthermore, in step S2, the lower bound of risk is calculated. The specific steps include: S21, for the precisely evaluated subset For each random event s in the process, construct a control model with the objective function of minimizing the costs of tripping, generator power adjustment, and DC power modulation; S22, Introduce a genetic algorithm to perform a global optimization search on the control model to obtain the optimal control strategy that satisfies the system safety and stability constraints and its corresponding minimum control cost C(s); S23. Verify the feasibility of the optimal control strategy through time-domain dynamic simulation. If the verification fails, apply a penalty term to the fitness function of the genetic algorithm to guide the algorithm to continue searching for feasible solutions. S24, according to formula Calculate the lower bound of risk .

[0010] Furthermore, the objective function of the control model is:

[0011] in, The economic cost of load shedding can be expressed as: ; The cost of adjusting generator power is expressed as: ; For incremental electricity pricing, To quote a reduced electricity price, For power increment, To reduce power; The cost of DC power regulation is expressed as: ; The cost of DC transmission power loss, The cost of DC regulation should be considered in terms of labor and other expenses.

[0012] Furthermore, the control model includes the following constraints: Load control constraints: ,in This represents the actual power of the load. Generator power regulation control constraints: ,in and These are the minimum and maximum adjustable power values ​​of the generator, respectively. DC power modulation constraints: ,in and These are the minimum and maximum power values ​​that can be adjusted for DC power. Transient frequency safety constraints: ,in This is the lowest frequency of the system after the fault. The lowest frequency threshold; Transient work angle stability constraints: ,in The maximum power angle difference between the generators; when When the angle is less than 180°, the system is considered transiently stable; when... When the angle is greater than 180°, the system is determined to be in transient power angle instability; Transient voltage safety constraints: 10 seconds after a DC blocking fault, the voltage at each load bus node must meet the following requirements: ,in Indicates the load bus node after disturbance voltage, Indicates the number of load busbar nodes; Power flow constraint: Power transmitted by each line on the transmission section , For the first section The transmission power of each line, For the first Maximum transmission power of each line The number of lines on the transmission line section; Operating constraints of DC modulation systems include: DC line current not exceeding limits and current interruption, and firing angle control constraints on the rectifier side and inverter side.

[0013] Furthermore, the specific execution steps of the genetic algorithm include: S221, Chromosome Encoding: The control strategy is encoded as a chromosome, where the generator switching control is represented by a single binary digit, and the generator power adjustment and DC power adjustment are represented by four binary digits. S222, Population Initialization: Construct an initial population with multiple chromosomes, each chromosome representing a control scheme. The length of the chromosome is determined by the controllable units to be optimized, the load, and the control ratio of the DC line. S223, Optimization Operation: Evaluate the merits of each chromosome in the population based on the control cost C(s), with smaller C(s) indicating better performance; Select individuals with better performance from the current population as parents; S224, Crossover operation: Randomly select two parent individuals with a preset crossover probability, exchange their different genes or gene segments, and generate offspring individuals; S225, Mutation operation: Randomly select some individuals in the population and randomly change the genes or gene segments of the individuals with a certain mutation probability. S226, Iteration Termination: When the preset maximum number of iterations is reached or there is no significant improvement in C(s) for multiple consecutive generations, the iteration stops and the current best individual is output as the optimal control cost C(s) for the desired scenario s.

[0014] Furthermore, the risk upper bound R is calculated. H The specific steps include: S25, regarding the inaccurately evaluated subset For each random event s, a conservative control cost strategy is adopted, and its control cost is set to the maximum conservative cost. ; S26, the maximum possible cost Defined as the maximum load shedding that the system may trigger, multiplied by the average node load loss cost factor, i.e. ,in This is the average node load loss cost coefficient. This represents the system's maximum load shedding capacity. S27, according to formula Calculate the upper bound of risk .

[0015] Furthermore, step S4 iteratively switches the already accurately evaluated subset. Specifically, it includes: S41, when the risk accuracy γ is lower than a preset threshold, according to the principle that the probability of an event occurring is proportional to the risk contribution, Nk (k≥2) fault events are gradually added to the process; S42 employs a simplified method of initial screening for short circuits at bus nodes and substitution using the mean of the Nk higher-order fault group to optimize the calculation of the lower bound of risk. Upper limit of risk and risk accuracy γ; S43, Repeat steps S41-S42 until the risk accuracy γ meets the preset accuracy requirements.

[0016] Furthermore, the characteristic feature is that step S42, which employs a bus node short-circuit preliminary screening, specifically includes the following steps: S421, Traverse all short circuits at all bus nodes and calculate their costs; S422, identify bus nodes where the short-circuit cost is not zero; S423, only check line short circuit scenarios connected to high short circuit cost bus nodes; The specific steps for replacing the mean of the Nk higher-order fault group include: S424, calculate the minimum sample size m representing the fault group Nk, using the following formula: Where Z is the reliability coefficient based on a preset information level. 2 Here, E represents the variance of the consequences, and E represents the allowable error. S425 randomly selects m samples from the Nk event set to form the event set S. m , for S m Perform time-domain dynamic simulation to obtain the consequence cost for each sample, and calculate the average consequence cost. ;

[0017] S426 calculates the total risk of fault group Nk.

[0018] , where P k (s) represents the total probability of occurrence of fault group Nk; S427 will Integrate to the lower risk bound At the same time, update the upper risk bound. of The sum of probabilities is used to iteratively update the risk boundary.

[0019] This invention also discloses a risk assessment system for provincial AC / DC power grid scenarios based on risk boundaries, characterized in that it includes: The scenario risk definition and control cost modeling module is used in step S1 above; The risk boundary construction module is used to perform step S2 above; The risk accuracy calculation module is used in step S3 above; The iterative convergence control module is used for step S4 above.

[0020] The present invention has the following beneficial effects: 1) This invention addresses the overly conservative nature of the traditional N-1 criterion by introducing a comprehensive consequence cost into the risk definition. The control model precisely quantifies the losses caused by system transient instability and the implementation costs of corresponding control measures, enabling a fundamental shift in power grid dispatching decisions from qualitative judgment to quantitative trade-offs based on economic benefits. By introducing a genetic algorithm for optimization, this method can efficiently search for and determine the minimum control cost strategy that satisfies all dynamic security and stability constraints. This maximizes the release of economic space for power grid operation while ensuring system transient security and stability, achieving optimal synergy between safety level and economic benefits.

[0021] 2) This invention achieves an effective trade-off between computational efficiency and evaluation accuracy by dividing the entire event set into subsets that require precise optimization and subsets that use conservative estimates. This hybrid strategy significantly reduces computational complexity and time cost. The introduction of an upper bound on risk effectively addresses the core pain points of underestimation of risk and lack of confidence in traditional simplified evaluation methods, clearly quantifying the maximum risk uncertainty caused by the inability to traverse all conservative subsets. This method transforms the evaluation objective from unattainable absolute accuracy to controllable accuracy by defining risk accuracy, and drives the convergence of risk boundaries by iteratively expanding the scale of the subsets that require precise optimization, thereby quantitatively guaranteeing the accuracy and confidence of the final evaluation results. At the same time, this evaluation system takes into account both N-1 basic safety and Nk higher-order risks, achieving effective coverage of the entire range of scenarios. Attached Figure Description

[0022] Figure 1 This is a flowchart of a risk assessment method for provincial AC / DC power grid scenarios based on risk boundaries, as described in a specific embodiment.

[0023] Figure 2 This is a schematic diagram illustrating the scenario division and risk boundary as described in a specific embodiment.

[0024] Figure 3 The risk value described in the specific embodiment varies with S L Scale relationship diagram.

[0025] Figure 4 The flowchart illustrates the minimum control cost based on a genetic algorithm in a specific embodiment.

[0026] Figure 5 The risk accuracy described in the specific embodiment varies with S L Scale relationship diagram.

[0027] Figure 6 This is a flowchart illustrating the risk assessment process in a specific embodiment. Detailed Implementation

[0028] To facilitate understanding by those skilled in the art, the present invention will be further described below in conjunction with embodiments and accompanying drawings.

[0029] This invention addresses the problems of heavy computational burden, long processing time, and lack of quantitative confidence in assessment accuracy caused by scenario explosion in traditional power grid risk assessment. It proposes a risk boundary-based method for provincial AC / DC power grid scenario risk assessment, innovatively constructing a risk boundary... This transforms the need for precise assessment of massive scenarios into an iterative convergence problem of risk uncertainty. This is achieved by dividing the set of random events S into subsets that have already been precisely assessed. and unevaluated subset Define and calculate the risk lower bound respectively. and upper bound of risk This provides a confidence interval for the total system risk R. Furthermore, a risk accuracy γ is defined to drive the evaluation subset. and Through iterative switching, a risk assessment result that meets the accuracy requirements is finally obtained. (See also...) Figure 1 The method in this embodiment includes the following steps: S1, Define scenario risks and construct a control cost model; the specific definition of scenario risks is as follows: The key to quantitatively assessing the operational safety of AC / DC power grids lies in establishing a comprehensive measurement system that reflects both the probability of risk and its combined consequences. This embodiment defines scenario risk as the expected combined consequences of all possible unforeseen events. It considers not only the probability of an unforeseen event P(s) but also introduces the combined consequences cost C(s), thus incorporating the losses suffered by the power grid and the economic costs of control measures into the scope of risk quantification. Based on this, the precise definition of scenario risk R(t) in this embodiment is as follows:

[0030] Where P(s) represents the probability of the occurrence of the sudden event s under consideration, and C(s) reflects the loss cost of all events caused by the sudden event before time t. Here, time t is set long enough so that the system can reach a new steady state.

[0031] All embodiments in this paper consider the failure probability under normal weather conditions. The probability P(s) of an occasional event s is determined based on the historical reliability data of the equipment operation. For a single failure event, its probability is calculated based on the equipment's failure rate λ and repair rate μ.

[0032]

[0033] For Nk-type multiple fault events, all k faults constituting event S are treated as independent events, and their probability P(s) is calculated using a joint probability model.

[0034]

[0035] The construction of the control cost model specifically includes: The loss cost C(s) of an unforeseen event s is the sum of the economic losses and control costs incurred in responding to the event. In scenarios that take into account AC and DC characteristics, it includes the economic cost of load shedding, generator power adjustment, and DC adjustment cost.

[0036]

[0037] Emergency load shedding; the economic cost of load shedding is expressed as:

[0038] Where, k i ∆P is the economic cost index of load loss at node i. load,i This represents the total load reduction at node i caused by event s, where n is the total number of nodes in the system.

[0039] The cost of adjusting generator power is expressed as follows:

[0040] Among them, a i For incremental electricity pricing, b i For reduced electricity pricing, ∆P G + Let ∆P be the generator power increment at node i. G - This represents the power reduction of the generator at node i, where n is the total number of nodes in the system.

[0041] The cost of DC power regulation can be expressed as follows:

[0042] Where C dc,loss (s) represents the cost of DC transmission power loss, C dc,control (s) is the cost of DC regulation considering labor and other expenses.

[0043] S2, constructing risk boundaries; specifically including: Define the initial S1 event set

[0044] The selection of the initial event set is a core prerequisite for evaluating efficiency and the accuracy of results. In view of the event probability characteristics and technical attributes, this invention selects N-1 fault events as the initial S1 event set.

[0045] From a risk contribution perspective, the quantitative calculation of power system fault risk R requires comprehensive consideration of the probability of event occurrence and the severity of consequences. Compared to multiple faults of N-2 or higher, N-1 events, i.e., single component failures or outages, have the highest probability of occurrence in actual power grid operation. Relevant statistics show that their frequency accounts for more than 80% of the total number of power grid faults. Therefore, it occupies a dominant risk contribution share in the total risk R and is a key influencing factor in determining the level of power grid security.

[0046] From a technical implementation perspective, N-1 faults typically manifest as clear scenarios such as single-line tripping or transformer disconnection, with well-defined boundary conditions and controllable impacts, providing a feasible path for rapid risk screening. Furthermore, the N-1 criterion, as a globally recognized power grid safety standard, explicitly requires that the system remain stable and avoid adverse consequences after a single component failure; this rigid requirement establishes its core assessment position. Therefore, all N-1 faults are used as the initial S1 event set.

[0047] If we consider the normal operation and fault states of all components in the system, it will result in 2 n If we further consider different power generation demands or the location of line faults, the number of scenarios will be enormous. However, due to the high computation time and economic cost of time-domain dynamic simulation, it is impossible to analyze such a large number of scenarios in a short period of time. Therefore, we adopt the method of introducing genetic algorithms and risk boundaries to quantify the risks of these scenarios.

[0048] This invention divides the entire set of random events S into a subset S1 where the optimal control cost has been solved using a genetic algorithm and a subset S2 where the cost has not been precisely calculated, as follows: Figure 2 As shown. Based on this division, R is defined under risk. L and risk upper bound R H This is used to define the range of the system's true total risk R.

[0049] By introducing a risk boundary, the evaluation objective is shifted from the absolute accuracy pursued by time-domain dynamic simulations, which are time-consuming and costly, to controllable accuracy and efficiency optimization under the quantitative constraints of risk uncertainty. R... H The introduction of is to quantify the degree of risk underestimation caused by the inability to fully traverse S2, thereby ensuring the confidence level of the assessment results.

[0050]

[0051] Among them, R L S1 is a subset of the data that has undergone optimization search based on a genetic algorithm and whose minimum control cost C(s) has been accurately calculated through time-domain simulation, serving as the lower bound for risk assessment. Therefore, risk assessment values ​​obtained solely from the S1 subset are always lower than the actual risk values, and thus serve as the lower bound for risk assessment.

[0052] For the subset of incidental events S2 that is not precisely calculated, due to the computational cost and time constraints of time-domain dynamic simulation, this method does not perform minimum cost optimization on its control strategy. To ensure the safety and confidence of the evaluation results, this invention adopts a conservative estimation strategy for S2, namely, calculating its maximum control cost C. H Maximum control cost C H Defined as the maximum control cost when all controllable resources are fully utilized under system operation and physical constraints. This conservative value is used to provide R... H The boundary is the biggest risk of underestimating the quantification of S2.

[0053]

[0054] Therefore, the following inequality exists:

[0055] This is the sum of the probabilities of all events in the subset S2. Calculating the probability for each individual event is tedious, and the sum of probabilities for the subset S1 has already been calculated. Since P(S1∪S2) = 1, the sum of the probabilities of all events in the subset S2 can be derived from the following formula:

[0056] like Figure 3 As shown, increasing the size of S1 will shorten R. H and R L This reduces the gap between them, thereby reducing risk and uncertainty.

[0057] Calculate the lower bound of risk R based on a genetic algorithm. L Specifically, it includes: Considering minimizing the costs of generator tripping, generator power adjustment, and DC power modulation, a control model is constructed that satisfies system safety and stability constraints. The objective function of the control model is to minimize the costs of generator tripping, generator power adjustment, and DC power modulation. Relevant formulas:

[0058] The costs are as shown in the section on constructing the control cost model in step S1.

[0059] The control model includes the following controllable resource constraints: The load is a switching resource, and its control constraints are as follows:

[0060] Where P loadi This represents the actual power of the load.

[0061] The generator is a continuously adjustable resource with step-size adjustment; therefore, it is considered as a continuously adjustable resource, and the control constraints are as follows:

[0062] Among them, P min- geni P max+ geni These are the maximum and minimum adjustable power values ​​of the generator, respectively.

[0063] DC power modulation is also a continuously adjustable resource, and its control constraints are:

[0064] Among them, P min- dc P max+ dc These are the maximum and minimum adjustable power values ​​for the DC power supply, respectively.

[0065] The control model also includes the following safety and stability constraints: The system minimum frequency is an important indicator reflecting the dynamic frequency security of the power system after a fault occurs. When the system frequency falls below the minimum frequency, it will trigger low-frequency load shedding and generator low-frequency protection actions. In this embodiment, the system minimum frequency is selected as the reference indicator for frequency security. The specific system transient frequency security constraints are as follows:

[0066] Among them, f min f represents the lowest frequency of the receiving-end system after a DC blockage fault. th min The system frequency is calculated according to the definition of the center frequency of inertia, based on the set minimum frequency threshold. Under normal operating conditions, the allowable deviation of the power supply frequency for systems with an installed capacity of 3GW or more is ±0.2Hz, and the allowable deviation for systems with an installed capacity of less than 3GW is ±0.5Hz.

[0067] Transient power angle stability of a power system refers to the ability of all generators to maintain synchronous operation after the system is subjected to a large disturbance. The maximum power angle difference between generators is an important criterion for judging whether the system has transient power angle instability. The specific constraints for system transient power angle stability are:

[0068] Where, △δ max This represents the maximum power angle difference between the generators. When it is less than 180°, the system is considered to be in transient power angle stability; conversely, when the power angle difference is greater than 180°, the system is considered to be in transient power angle instability.

[0069] After a power system fault occurs, the voltage at each bus node should remain within the allowable range. According to my country's practical engineering criteria for transient voltage stability, during the transient process following a large disturbance in the power system, if the load bus voltage can recover to above 0.8 pu within 10 seconds after the disturbance, the system is considered transiently stable. Therefore, the specific transient voltage safety constraint in this paper is that, 10 seconds after a DC blocking fault, each load bus node must satisfy the following:

[0070] Among them, u i This represents the voltage at node i of the load bus after the disturbance, n LB This indicates the number of load bus nodes.

[0071] High-power deficit incidents such as DC blocking can cause large-scale power flow shifts in the power grid, leading to power overruns at critical transmission sections. This paper imposes power flow constraints such that the transmission power of each line on the transmission section satisfies the following:

[0072] Among them, P i Let P be the transmission power of the i-th line on the cross section. i,max Let n be the upper limit of the transmission power of the i-th line. l This refers to the number of lines on the power transmission section.

[0073] The control model also includes the state-space expression for DC modulation and its security constraints.

[0074]

[0075]

[0076] Where E1 and E2 are the equivalent voltages of the AC system on the rectifier side and inverter side, respectively; L r L i These are the equivalent inductances of the rectifier and inverter sides of the AC system, respectively; U dcr U dci The DC line voltages for the rectifier and inverter sides are respectively; L dcr L dci The equivalent inductances on the rectifier side and inverter side of the DC line are respectively; C dc U is the equivalent capacitance of a DC line. dc This is the voltage across the DC line capacitor. dc ref U dc ref These are the command values ​​for the rectifier-side current and inverter-side voltage of the DC line; S cI S cU The integral outputs of the current and DC voltage controllers are respectively; K pIdc K ildc and K pUdcK IUdc These are the proportional and integral coefficients for the constant current controller and constant voltage controller, respectively; α and β are the rectifier-side delayed firing angle and the inverter-side advanced firing angle, respectively.

[0077] For DC power modulation systems, operational constraints include ensuring the DC line current does not exceed limits and preventing current interruption, as well as firing angle control constraints on the rectifier and inverter sides, namely:

[0078]

[0079] Calculate the lower bound of risk R based on a genetic algorithm. L Specifically, it includes: From the formula It can be seen that the lower bound of risk R L The accurate solution depends on the minimum comprehensive consequence cost C(s) for each scenario S in the evaluated subset S1. Since minimizing C(s) involves multiple control variable coupling, nonlinear dynamic constraints, and dependence on time-domain dynamic simulation, this embodiment introduces a genetic algorithm as an efficient global optimization solver. The introduction of the genetic algorithm transforms the problem of minimizing control costs into an optimization problem with safety constraints as hard penalties, which greatly improves the efficiency of searching for the optimal emergency control strategy in complex AC / DC power grid models.

[0080] Genetic algorithms draw upon natural selection and genetic mechanisms from biological evolution, searching for optimal solutions that satisfy dynamic constraints by simulating mechanisms such as crossover and mutation of chromosome genes during biological evolution. The specific steps include the following: Chromosome Encoding and Initial Population Generation: In the initial population generation stage, the algorithm constructs a certain number of chromosomes, each representing a control scheme. The length of the chromosome is determined by the controllable units to be optimized, the load, the DC line control ratio, etc. Gene values ​​on the chromosome are either 0 or 1. There are n control ratios for the controllable resources, represented by all binary numbers between 0 and n. Among them, generator switching control is a discrete control variable, which can be represented by a single gene, while generator power adjustment and DC power adjustment are continuous adjustment variables, represented by four binary numbers, i.e., four genes. Initial gene values ​​are determined by preset rules to form the initial emergency control scheme as the starting point for iteration. The control cost of the emergency control scheme represented by each chromosome is calculated, and this value serves as the standard for evaluating the quality of individuals in the population.

[0081] Optimization Operation: The selection operation aims to select individuals with better performance (i.e., lower control cost C(s)) from the current population based on the evaluation results of C(s), giving them the opportunity to pass on their superior traits to the next generation as parents. This process selects the best-performing set of n individuals from all current solutions as parents, preserving the low-cost control strategy in the population.

[0082] Crossover operation: Among the selected parents, the algorithm randomly selects two individuals with a preset crossover probability and exchanges their different genes or gene segments to generate new offspring. The crossover operation effectively combines the low-cost control characteristics of different parents and explores new and better combinations.

[0083] Mutation operation: The algorithm randomly selects a subset of individuals in the population and randomly alters their genes or gene segments with a certain mutation probability. Mutation provides an opportunity for the generation of new individuals, ensuring population diversity and preventing the algorithm from getting trapped in local optima.

[0084] Through selection, crossover, and mutation operations, the state of the current control measures is updated, and a new generation of control measure populations is generated.

[0085] The iteration termination condition is as follows: the algorithm will stop when the maximum number of iterations or the minimum C(s) has not shown significant improvement for several consecutive generations. The best individual in the current population is then identified as the optimal control cost C(s) for the desired scenario S.

[0086] The minimum control cost process based on genetic algorithms is as follows: Figure 4 As shown.

[0087] After the genetic algorithm generates the control strategy, the time-domain dynamic simulation module is invoked to run the control strategy to accurately verify whether dynamic safety and stability constraints such as power angle, frequency, and voltage are satisfied. If not, a large penalty term is applied to the fitness function to guide the algorithm to continue searching for feasible solutions.

[0088] The risk upper bound R is calculated based on the cost of conservative control. H Specifically, it includes: Risk cap R H The core idea is to adopt a conservative control approach, quantifying the risk uncertainty in the unassessed subset S2, thereby providing a safe and reliable upper limit for the risk value of the scenario. Related formula: .

[0089] Among them, C H It must be an upper limit value for ensuring a level of security confidence, and it needs to satisfy:

[0090] Depend on It can be seen that the control cost C(s) is the total cost of load shedding, generator power adjustment, and DC power adjustment. Since the fundamental purpose of all flexible control measures is to avoid or minimize load shedding, and the user-side load loss C load (s) is typically much higher than generator adjustment costs C. gen (s) and DC regulation cost C dc (s), therefore the maximum conservative cost C H It is defined as the maximum load shedding that the system may cause multiplied by the average node load loss cost factor.

[0091]

[0092] in, ΔP is the average nodal load loss cost factor. load,max This represents the maximum load shedding capacity of the system.

[0093] Risk upper bound R H The introduction of this method is key to achieving efficient risk assessment in the scenarios described in this invention. The strategy involves using algebraic calculations to determine the conservative control cost C. H This is used to quickly define the maximum potential consequences of loss in S2, thereby setting strict safety boundaries for the risk assessment results without the need for time-consuming dynamic simulation.

[0094] S3 defines risk accuracy, specifically including: Using risk upper and lower limits R L and R H Risk accuracy is defined by the ratio:

[0095] 1) Maximum control cost C H The smaller the difference between the expected value of the cost C(s) in subset S1 and the expected value of the cost C(s), the higher the risk accuracy.

[0096] 2) The higher the risk accuracy, the more events in subset S1 increase and the fewer events in S2 decrease.

[0097] Risk accuracy varies with S L Scale changes such as Figure 5 As shown.

[0098] From the above two characteristics, we can see that C H The risk accuracy should be as low as possible to achieve higher risk accuracy on the same subset S1. Feature 2 can be used for iterative risk calculations, terminating the calculation when the risk accuracy reaches the expected level. The iterative process is as follows: Figure 6 As shown.

[0099] S4, selecting the iterative event set and its simplification method, specifically including: See Figure 4 If the desired accuracy is not achieved after the first iteration, following the principle that the probability of an event occurring is proportional to its risk contribution, events N-2, N-3, ..., Nk will be gradually added to the subset S1 in subsequent iterations, and so on. However, the number of simulations required for risk assessment may quickly exceed computational capacity. Therefore, the following simplification method can be introduced to accelerate the computation process.

[0100] S41, Preliminary screening based on bus node short circuit

[0101] Since the severity of a fault occurring at the busbar is greater than the severity of a short circuit occurring in the line connected to that busbar node, if the cost of a short circuit fault at both busbar nodes connected to the line is 0 or very small, then the cost of a short circuit fault at that line is also likely to be 0.

[0102] Screening steps: Iterate through all short circuits at all bus nodes and calculate their costs.

[0103] Identify bus nodes where the short-circuit cost is not zero.

[0104] Only short-circuit scenarios involving lines connected to high short-circuit cost bus nodes are examined.

[0105] S42, using the mean to represent the cost of the nk fault group.

[0106] The high-order fault group Nk (N≥2) contains a huge number of random events. Performing time-consuming time-domain dynamic simulations for each scenario would result in an unbearable computational burden. Therefore, this method introduces a simplified strategy based on statistical sampling and inference to achieve a comprehensive assessment of the risk R of the entire Nk fault group. k Highly efficient evaluation.

[0107] First, the system no longer performs exhaustive simulation of all events in the Nk fault group, but instead calculates the minimum number of samples m that can represent the group based on statistical inference.

[0108]

[0109] Where Z is the reliability coefficient. The corresponding Z value is obtained by consulting the standard normal distribution table based on the preset confidence level. σ 2 Consequence variance is used to quantify the volatility of the consequences of all events in the Nk fault group. Variance σ 2 Typically, this is estimated by performing a small-scale pre-sampling of the fault group and calculating the squared standard deviation of the cost of the consequences of that pre-sampled sample. E is the allowable error, determined by the operator, and represents the average cost C calculated from the sample. k The maximum allowable deviation between the actual total cost and the average cost of the group.

[0110] Subsequently, an event set S is formed by randomly selecting m sample events from the Nk event set. m Only for this statistically representative subset S of random events m Perform accurate time-domain dynamic simulations to calculate the consequences and costs, and obtain the average consequence cost C. k Ultimately, the average cost C is adopted. k Multiplied by the total probability P of the entire Nk group k (s) is used to estimate the total risk R of the Nk fault group. k .

[0111]

[0112] After completing the statistical evaluation of the N−k fault groups, the R k This will be immediately integrated into the risk boundary update.

[0113] Risk lower bound update: R L The risk value is obtained by summing the N−k portions. Risk upper bound update: R H As fewer events were not evaluated in group S2, the total probability of those events decreased accordingly.

[0114] By iteratively calculating new risk upper and lower bounds by gradually adding Nk (k=2, 3, ..., N) event sets to the S1 subset, the risk value upper and lower bounds are driven to converge until the risk accuracy γ meets the precision requirements, and the final risk value range is output.

[0115] The above embodiments are merely illustrative of the technical concept of the present invention and should not be construed as limiting the scope of protection of the present invention. Any modifications made to the technical solution based on the technical concept proposed in this invention shall fall within the scope of protection of this invention.

Claims

1. A risk assessment method for provincial AC / DC power grid scenarios based on risk boundaries, characterized in that, The method includes the following steps: S1, define scenario risk and construct control cost model; define scenario risk R(t) as the expected comprehensive consequence cost of all possible unforeseen events; S2, Constructing the risk boundary; dividing the entire set of incidental events S into precisely evaluated subsets. and unevaluated subset Define and calculate the lower bound of risk R. L and risk upper bound R H ,in: The lower bound of risk R L Based on the already accurately evaluated subset The minimum control cost C(s) and its probability of occurrence P(s) for each random event are calculated. The risk upper bound R H Based on the inaccurately evaluated subset The conservative control cost C for each random event H It is obtained by calculating its occurrence probability P(s); S3, based on the risk lower bound R L and risk upper bound R H Define risk accuracy γ = R L / R H This is used to quantify the confidence level of risk assessment results; S4, iteratively switching the precisely evaluated subset Gradually reduce the lower bound of the risk R L With risk upper bound R H The difference between the two is calculated until the risk accuracy γ meets the preset accuracy requirement, at which point the final risk assessment interval [R] is output. L R H ].

2. The risk assessment method for provincial AC / DC power grid scenarios based on risk boundaries according to claim 1, characterized in that, The precise definition of the scenario risk R(t) is as follows: ， Wherein, P(s) is the probability of occurrence of an unplanned event s, and C(s) is the cost of loss of an unplanned event s; the cost of loss of an unplanned event s, C(s), is the sum of the economic cost of load shedding, the cost of generator power adjustment, and the cost of DC power adjustment; the probability of occurrence of a single fault event, P(s), is calculated based on the equipment failure rate λ and the repair rate μ, and the probability of occurrence of Nk-type multiple fault events, P(s), is calculated using a joint probability model.

3. The risk assessment method for provincial AC / DC power grid scenarios based on risk boundaries according to claim 1, characterized in that, Step S2 divides the entire set of random events S into precisely evaluated subsets. and unevaluated subset Specifically, it includes: In the initial state, N-1 fault events are selected as the accurately evaluated subset. ; In subsequent iterations, based on the risk contribution, failure events N-2, N-3, and up to Nk are gradually removed from the inaccurately evaluated subset. Move into the precisely evaluated subset .

4. The risk assessment method for provincial AC / DC power grid scenarios based on risk boundaries according to claim 1, characterized in that, In step S2, the risk lower bound is calculated. The specific steps include: S21, for the precisely evaluated subset For each random event s in the process, construct a control model with the objective function of minimizing the costs of tripping, generator power adjustment, and DC power modulation; S22, Introduce a genetic algorithm to perform a global optimization search on the control model to obtain the optimal control strategy that satisfies the system safety and stability constraints and its corresponding minimum control cost C(s); S23. Verify the feasibility of the optimal control strategy through time-domain dynamic simulation. If the verification fails, apply a penalty term to the fitness function of the genetic algorithm to guide the algorithm to continue searching for feasible solutions. S24, according to formula Calculate the lower bound of risk .

5. The risk assessment method for provincial AC / DC power grid scenarios based on risk boundaries according to claim 4, characterized in that, The objective function of the control model is: ， in, The economic cost of load shedding can be expressed as: ; The cost of adjusting generator power is expressed as: ; For incremental electricity pricing, To quote a reduced electricity price, For power increment, To reduce power; The cost of DC power regulation is expressed as: ; The cost of DC transmission power loss, The cost of DC regulation should be considered in terms of labor and other expenses.

6. The risk assessment method for provincial AC / DC power grid scenarios based on risk boundaries according to claim 4, characterized in that, The control model includes the following constraints: Load control constraints: ,in This represents the actual power of the load. Generator power regulation control constraints: ,in and These are the minimum and maximum adjustable power values ​​of the generator, respectively. DC power modulation constraints: ,in and These are the minimum and maximum power values ​​that can be adjusted for DC power. Transient frequency safety constraints: ,in This is the lowest frequency of the system after the fault. The lowest frequency threshold; Transient work angle stability constraints: ,in This represents the maximum power angle difference between the generators; when When the angle is less than 180°, the system is considered transiently stable; when... When the angle is greater than 180°, the system is determined to be in transient power angle instability; Transient voltage safety constraints: 10 seconds after a DC blocking fault, the voltage at each load bus node must meet the following requirements: ,in Indicates the load bus node after disturbance voltage, Indicates the number of load busbar nodes; Power flow constraint: Power transmitted by each line on the transmission section , For the first section The transmission power of each line, For the first Maximum transmission power of each line The number of lines on the transmission line section; Operating constraints of DC modulation systems include: DC line current not exceeding limits and current interruption, and firing angle control constraints on the rectifier side and inverter side.

7. The risk assessment method for provincial AC / DC power grid scenarios based on risk boundaries according to claim 4, characterized in that, The specific execution steps of the genetic algorithm include: S221, Chromosome Encoding: The control strategy is encoded as a chromosome, where the generator switching control is represented by a single binary digit, and the generator power adjustment and DC power adjustment are represented by four binary digits. S222, Population Initialization: Construct an initial population with multiple chromosomes, each chromosome representing a control scheme. The length of the chromosome is determined by the controllable units to be optimized, the load, and the control ratio of the DC line. S223, Optimization Operation: Evaluate the merits of each chromosome in the population based on the control cost C(s), with smaller C(s) indicating better performance; Select individuals with better performance from the current population as parents; S224, Crossover operation: Randomly select two parent individuals with a preset crossover probability, exchange their different genes or gene segments, and generate offspring individuals; S225, Mutation operation: Randomly select some individuals in the population and randomly change the genes or gene segments of the individuals with a certain mutation probability. S226, Iteration Termination: When the preset maximum number of iterations is reached or there is no significant improvement in C(s) for multiple consecutive generations, the iteration stops and the current best individual is output as the optimal control cost C(s) for the desired scenario s.

8. The risk assessment method for provincial AC / DC power grid scenarios based on risk boundaries according to claim 1 or 4, characterized in that, Calculate the risk upper bound R H The specific steps include: S25, regarding the inaccurately evaluated subset For each random event s, a conservative control cost strategy is adopted, and its control cost is set to the maximum conservative cost. ; S26, the maximum possible cost Defined as the maximum load shedding that the system may trigger, multiplied by the average node load loss cost factor, i.e. ,in This is the average node load loss cost coefficient. This represents the system's maximum load shedding capacity. S27, according to formula Calculate the upper bound of risk .

9. The risk assessment method for provincial AC / DC power grid scenarios based on risk boundaries according to claim 1, characterized in that, Step S4 iteratively switches the accurately evaluated subset. Specifically, it includes: S41, when the risk accuracy γ is lower than a preset threshold, according to the principle that the probability of an event occurring is proportional to the risk contribution, Nk (k≥2) fault events are gradually added to the process; S42 employs a simplified method of initial screening for short circuits at bus nodes and substitution using the mean of the Nk higher-order fault group to optimize the calculation of the lower bound of risk. Upper limit of risk And risk accuracy γ; the specific steps of step S42, which uses bus node short-circuit preliminary screening, include: S421, Traverse all short circuits at all bus nodes and calculate their costs; S422, identify bus nodes where the short-circuit cost is not zero; S423, only check line short circuit scenarios connected to high short circuit cost bus nodes; The specific steps for replacing the mean of the Nk higher-order fault group include: S424, calculate the minimum sample size m representing the fault group Nk, using the following formula: Where Z is the reliability coefficient based on a preset information level. 2 Here, E represents the variance of the consequences, and E represents the allowable error. S425 randomly selects m samples from the Nk event set to form the event set S. m , for S m Perform time-domain dynamic simulation to obtain the consequence cost for each sample, and calculate the average consequence cost. ; S426 calculates the total risk of fault group Nk. , where P k (s) represents the total probability of occurrence of fault group Nk; S427 will Integrate to the lower risk bound At the same time, update the upper risk bound. of The sum of probabilities completes the iterative update of the risk boundary. S43, Repeat steps S41-S42 until the risk accuracy γ meets the preset accuracy requirements.

10. A risk assessment system for provincial AC / DC power grid scenarios based on risk boundaries, characterized in that, include: The scenario risk definition and control cost modeling module is used to perform step S1 as described in claim 1; A risk boundary construction module is used to perform step S2 as described in claim 1; A risk accuracy calculation module is used to perform step S3 as described in claim 1; An iterative convergence control module is used to execute step S4 as described in claim 1.

Images