A multi-user power load aggregation response scheduling method and system

Abstract

The present application relates to the technical field of smart grid dispatch, in particular to a multi-user power consumption load aggregation response dispatch method and system, wherein the dispatch method comprises the following steps: S1, constructing an instantaneous trajectory probability distribution; S2, calculating a dispersion index and a model correction index; S3, generating a comprehensive risk index; S4, triggering emergency stability control; S5, generating a dispatch instruction; by constructing the instantaneous trajectory probability distribution and calculating the state trajectory dispersion and the model correction double-dimension index, the potential risk caused by unknown coupling effect can be autonomously found; on this basis, the double-layer defense mechanism of emergency stability control and adaptive optimization dispatch is constructed, the efficient cooperation of response rapidity and dispatch accuracy is realized, the protection misoperation chain caused by the mismatch between control logic and actual physical law is avoided, and the resilience and security protection capability of large-scale power grid under complex disturbance are significantly improved.

Description

Technical Field

[0001] This invention relates to the technical field of smart grid dispatching, and in particular to a multi-user electricity load aggregation response dispatching method and system. Background Technology

[0002] With the large-scale integration of distributed power sources and diverse loads, the operating characteristics of modern power systems have undergone profound changes. When disturbances such as line faults and natural disasters cause local power grids to form unexpected islands, the system exhibits characteristics such as low inertia and weak voltage support due to the loss of strong support from the main power grid. The traditional assumptions of active power-frequency (Pf) and reactive power-voltage (QV) decoupling control are difficult to hold, and the electrical quantities within the system exhibit significant strong coupling characteristics.

[0003] Existing multi-user load aggregation response scheduling methods are mostly based on decoupled model designs under large power grid conditions. Their control logic follows the independent response principle of shedding loads when the frequency is low and generating reactive power when the voltage is low. However, under weakly connected conditions such as islanding, such decoupled control can trigger catastrophic cascading failures. For example, when a batch of resistive loads is shelved to cope with a frequency drop, due to the strong coupling effect of active power and voltage (PV), the sudden decrease in active power may lead to an abnormal increase in local voltage, which in turn triggers the overvoltage protection of distributed photovoltaic inverters, causing large-scale photovoltaic disconnection. Photovoltaic disconnection further exacerbates the active power deficit of the system, ultimately leading to frequency collapse and islanding instability. This chain of malfunctions in compliance protection caused by the mismatch between control logic and actual physical laws is a technical problem that urgently needs to be solved to improve the power grid's security capabilities and enhance system resilience. Summary of the Invention

[0004] To address the technical problems existing in the background art, this invention proposes a multi-user electricity load aggregation response scheduling method and system, the specific scheme of which is as follows: A multi-user electricity load aggregation response scheduling method includes the following steps: S1. Obtain real-time measurement values ​​of multiple key state variables, and construct an instantaneous trajectory probability distribution based on the real-time measurement values; S2. Based on the instantaneous trajectory probability distribution, calculate the dispersion index reflecting the degree of power system disorder, and calculate the model correction index reflecting the degree of power system behavior deviating from the historical normal pattern. S3. Combine the aforementioned dispersion index and model correction index to generate a comprehensive risk index; S4. When the comprehensive risk index meets the first preset condition, emergency stabilization control is triggered, the adjustable load is controlled to provide enhanced frequency damping support and voltage damping support, and flexible load reduction is performed. S5. When the comprehensive risk index meets the second preset condition, the abnormal coupling mode is located based on the model correction degree index, and a penalty term is introduced into the preset optimized scheduling model for correction to generate scheduling instructions.

[0005] Furthermore, in S1, constructing an instantaneous trajectory probability distribution based on the real-time measurement values ​​includes: The multidimensional state space composed of multiple key state variables is discretized into a preset number of state cells. Within a preset sliding time window, the frequency of the power system state trajectory falling into each of the state cells is counted; The instantaneous trajectory probability distribution is constructed based on the frequency; Among them, the multiple key state variables include at least: power system frequency, key node voltage, aggregated load total active power, total reactive power, and distributed generation output.

[0006] Furthermore, in S2, the dispersion index reflecting the degree of power system turbulence is calculated, including: Based on the instantaneous trajectory probability distribution, the state trajectory dispersion index is calculated. The state trajectory dispersion index is the Shannon entropy. When the probability of a certain state cell in the instantaneous trajectory probability distribution is zero, the Shannon entropy term corresponding to the state cell takes the value of zero. Calculate the time rate of change of the dispersion index of the state trajectory to obtain the dispersion growth rate, which is used as the dispersion index.

[0007] Furthermore, in S2, a model correction index reflecting the degree to which power system behavior deviates from historical normal patterns is calculated, including: Acquire historical normal operation data and train and construct a baseline operation characteristic model based on the historical normal operation data; By utilizing real-time measurements of multiple key state variables, the parameters or state probability distributions of the baseline operating characteristic model are updated based on Bayesian inference, thereby obtaining an instantaneous operating characteristic model that can fit the current power system behavior. Calculate the KL divergence between the real-time operating characteristic model and the baseline operating characteristic model to obtain the model correction index, which is used as the model correction index.

[0008] Furthermore, in S3, the dispersion index and the model correction index are combined to generate a comprehensive risk index, including: The dispersion index and the model correction index are normalized respectively; The normalized dispersion index and the model correction index are weighted and summed according to preset weights to obtain the comprehensive risk index.

[0009] Furthermore, in S4, the first preset condition is that the comprehensive risk index exceeds a preset first threshold.

[0010] Furthermore, in S4, the controllable load provides enhanced frequency damping support and voltage damping support, including: The adjustable load is controlled to adjust the power. The power adjustment amount is the negative value of the sum of the product of the frequency deviation and the preset frequency damping coefficient and the product of the voltage deviation and the preset voltage damping coefficient, wherein the frequency damping coefficient and the voltage damping coefficient are respectively greater than the corresponding damping coefficients in the normal operation mode. The implementation of flexible load reduction includes controlling lower-priority loads to perform gradual power reduction operations.

[0011] Furthermore, in S5, the second preset condition is that the comprehensive risk index is between a preset first threshold and a preset second threshold, or falls from above the first threshold to between the first threshold and the second threshold, wherein the first threshold is higher than the second threshold; A penalty term is introduced into the preset optimized scheduling model for correction in order to generate scheduling instructions, including: the preset optimized scheduling model is an optimized scheduling model with the goal of minimizing scheduling cost, and the optimized scheduling model includes a scheduling cost function; Based on the identified abnormal coupling pattern, a risk coupling penalty term specifically designed for that abnormal coupling pattern is added to the scheduling cost function to obtain the corrected optimized scheduling model. The scheduling instructions are generated based on the revised optimized scheduling model.

[0012] Furthermore, in S5, the abnormal coupling mode is located based on the model correction index, including: For the state transition sequence that causes the model correction index to increase, by comparing the parameter differences of the state transition sequence in the baseline operating characteristic model and the real-time operating characteristic model, the abnormal coupling mode in which the coupling relationship between state variables is abnormally enhanced is located. When the identified abnormal coupling mode is a strong coupling between a decrease in active power and an abnormal increase in voltage, and the abnormal coupling mode causes the model correction index to increase, the risk coupling penalty term is constructed as follows: the product of the penalty coefficient and the sum of the absolute values ​​of the active loads to be cut off for each load, multiplied by an indicator function, wherein the indicator function takes a value of 1 when the voltage change predicted based on the real-time operating characteristic model is greater than a preset voltage safety threshold, and otherwise takes a value of 0, and the penalty coefficient is a coefficient positively correlated with the model correction index.

[0013] A multi-user electricity load aggregation response dispatching system includes: The trajectory modeling module is used to acquire real-time measurement values ​​of multiple key state variables and construct an instantaneous trajectory probability distribution based on the real-time measurement values. The dual-dimensional evaluation module is used to calculate a dispersion index reflecting the degree of power system disorder based on the instantaneous trajectory probability distribution, and to calculate a model correction index reflecting the degree of deviation of power system behavior from historical normal patterns. The risk fusion module is used to fuse the dispersion index and the model correction index to generate a comprehensive risk index. The emergency control module is used to trigger emergency stabilization control when the comprehensive risk index meets the first preset condition, control the adjustable load to provide enhanced frequency damping support and voltage damping support, and perform flexible load reduction; The adaptive scheduling module is used to locate the abnormal coupling mode based on the model correction index when the comprehensive risk index meets the second preset condition, and accordingly introduce a penalty term into the preset optimized scheduling model for correction to generate scheduling instructions.

[0014] Compared with the prior art, the present invention can achieve at least the following beneficial effects: 1. This invention constructs an instantaneous trajectory probability distribution and calculates two-dimensional indices: state trajectory dispersion and model correction degree. This enables the autonomous discovery of potential risks caused by unknown coupling effects without relying on precise physical models or pre-set fault rule bases. Based on this, a two-layer defense mechanism of emergency stability control and adaptive optimal scheduling is constructed: in high-risk situations, local autonomous enhanced damping support and flexible load reduction quickly curb power system deterioration; in medium-risk situations, abnormal coupling patterns are located and targeted penalty terms are introduced into the optimal scheduling model to avoid triggering paths of cascading faults at the decision-making level. This achieves efficient synergy between rapid response and precise scheduling, fundamentally avoiding protection malfunction chains caused by mismatch between control logic and actual physical laws, and significantly improving the resilience and security of large-scale power grids under complex disturbances.

[0015] 2. This invention constructs an instantaneous trajectory probability distribution and calculates a dispersion index reflecting the degree of power system disorder and a model correction index reflecting the degree of behavioral deviation. Without relying on a precise physical model or a pre-set fault rule base, it can autonomously discover potential risks caused by unknown coupling effects or abnormal behavior patterns in an unsupervised manner, and has stronger adaptability and robustness to events that exceed historical experience. By integrating the dispersion index and the model correction index into a comprehensive risk index, and using this index for tiered response, a two-tiered defense mechanism of emergency stability control and adaptive optimization scheduling is constructed. When the risk is high, emergency stability control quickly curbs system deterioration and prevents the triggering and spread of cascading failures by providing enhanced frequency and voltage damping support and flexible load reduction. When the risk is at a moderate level or has been initially contained, adaptive optimization scheduling identifies abnormal coupling patterns and introduces targeted penalty terms into the optimization scheduling model to avoid control behaviors that may trigger cascading failures at the decision-making level, thus eradicating the source of the failure. This tiered response mechanism balances the speed of response with the accuracy of scheduling. By introducing a risk coupling penalty term corresponding to the abnormal coupling mode into the scheduling cost function, control commands that may exacerbate system risks can be automatically avoided during the optimized scheduling process. This generates a safe scheduling strategy that matches the actual physical characteristics of the current power system, thereby avoiding the compliance protection malfunction chain caused by the mismatch between control logic and physical laws, and significantly improving the resilience and safety margin of large-scale power grids under weak connection conditions. Attached Figure Description

[0016] To more clearly illustrate the technical solutions of the embodiments of the present invention, the accompanying drawings used in the description of the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort. Wherein: Figure 1 This is a flowchart of the method of the present invention.

[0017] Figure 2 This is a system principle block diagram of the present invention. Detailed Implementation

[0018] Embodiments of the present invention are described in detail below. Examples of these embodiments are illustrated in the accompanying drawings, wherein the same or similar symbols denote the same or similar elements or elements having the same or similar functions throughout. The embodiments described below with reference to the accompanying drawings are exemplary and are only used to explain the present invention, and should not be construed as limiting the present invention.

[0019] Example 1, please refer to Figure 1 This invention provides a multi-user electricity load aggregation response scheduling method, comprising the following steps: S1. Obtain real-time measurement values ​​of multiple key state variables, and construct an instantaneous trajectory probability distribution based on the real-time measurement values.

[0020] In an optional embodiment, in S1, constructing an instantaneous trajectory probability distribution based on the real-time measurement values ​​includes: The multidimensional state space composed of multiple key state variables is discretized into a preset number of state cells. Within a preset sliding time window, the frequency of the power system state trajectory falling into each of the state cells is counted; The instantaneous trajectory probability distribution is constructed based on the frequency; Among them, the multiple key state variables include at least: power system frequency, key node voltage, aggregated load total active power, total reactive power, and distributed generation output.

[0021] It should be noted that the aforementioned key state variables are core electrical parameters used to characterize the real-time operating status of the power system. Among them, the power system frequency reflects the active power balance of the system and is the primary indicator for measuring system stability; the key node voltage reflects the voltage level of important buses in the system, and is typically monitored at grid connection points or load centers that have a significant impact on system stability; the aggregated load's total active power and total reactive power reflect the total active and reactive power consumed by the load side, respectively, and are fundamental data for assessing load response potential; the distributed generation output reflects the real-time power generation of distributed sources such as photovoltaics and wind power in the system, and is an important reference for assessing the system's inertia level and regulation capability. These key state variables are collected in real-time by measurement devices deployed at each user side or key node and reported to the load aggregator side for aggregation and processing via a communication network.

[0022] It should be noted that the multidimensional state space refers to a high-dimensional space spanned by the aforementioned key state variables, where each point represents the complete operating state of the system at a certain moment. Discretizing the multidimensional state space into a predetermined number of state cells means dividing the value range of each key state variable uniformly or non-uniformly into several intervals. The Cartesian product of these intervals forms a hypercube-shaped state cell. For example, the frequency dimension can be divided into several intervals with a width of 0.01 Hz, the voltage per unit dimension into several intervals with a width of 0.01 pu, and so on. The number of discretized state cells depends on the precision of the division of each dimension and the total number of dimensions. The specific number can be preset based on the computational resources and risk identification accuracy requirements of the actual system.

[0023] It should be noted that the preset sliding time window refers to an observation window with a fixed duration that slides forward over time, such as the most recent 5 seconds. Within this sliding time window, the system state trajectory refers to the path of the system's operating state as it evolves in the multidimensional state space over time. The frequency of falling into each state cell is counted, that is, the number of times the system state trajectory visits each state cell within this time window. Constructing an instantaneous trajectory probability distribution based on this frequency involves dividing the frequency of each state cell's visit by the total number of samples within the time window to obtain an estimated probability that the system state falls into each state cell. This instantaneous trajectory probability distribution reflects the system's operating characteristics within the current time window. When the system is running stably, the state trajectory is usually concentrated in a few state cells, and the probability distribution is relatively concentrated; when the system is disturbed or tends to become unstable, the state trajectory moves rapidly over a wider area, and the probability distribution tends to be more dispersed. This change in distribution characteristics provides a basis for calculating the dispersion index and assessing the degree of system turbulence in subsequent steps.

[0024] S2. Based on the instantaneous trajectory probability distribution, calculate the dispersion index reflecting the degree of power system disorder, and calculate the model correction index reflecting the degree of power system behavior deviating from the historical normal pattern.

[0025] In an optional embodiment, in S2, the dispersion index reflecting the degree of power system turbulence is calculated, including: Based on the instantaneous trajectory probability distribution, the state trajectory dispersion index is calculated. The state trajectory dispersion index is the Shannon entropy. When the probability of a certain state cell in the instantaneous trajectory probability distribution is zero, the Shannon entropy term corresponding to the state cell takes the value of zero. Calculate the time rate of change of the dispersion index of the state trajectory to obtain the dispersion growth rate, which is used as the dispersion index.

[0026] It should be noted that the dispersion index quantifies the degree of operational disorder in a system from the perspective of the disorder of its state trajectory. Its core idea is that when a power system is in a stable operating state, the measured values ​​of key state variables fluctuate slightly around the rated operating point, and the system state trajectory is concentrated in a few state cells in the multidimensional state space, resulting in a highly concentrated instantaneous trajectory probability distribution. When the system is disturbed or tends towards instability, key state variables deviate significantly, and the state trajectory moves rapidly among a wider range of state cells, leading to a more uniform and dispersed instantaneous trajectory probability distribution. The state trajectory dispersion index uses the mathematical form of Shannon entropy to measure the degree of concentration or dispersion of this distribution. A lower entropy value indicates a more ordered system operation, while a higher entropy value indicates a more disordered system operation. It is stipulated that when the probability of a certain state cell is zero, the corresponding Shannon entropy term is zero to ensure mathematical rigor and avoid problems with the domain of the logarithmic function. The dispersion growth rate is the difference between the state trajectory dispersion index and adjacent calculation periods, characterizing the changing trend of the system's disorder. When the dispersion growth rate is significantly positive, it indicates that the system is rapidly transitioning from an ordered state to a disordered state, and is identified as a first-type risk indication signal.

[0027] In an optional embodiment, in S2, a model correction index reflecting the degree to which the power system behavior deviates from historical normal patterns is calculated, including: Acquire historical normal operation data and train and construct a baseline operation characteristic model based on the historical normal operation data; By utilizing real-time measurements of multiple key state variables, the parameters or state probability distributions of the baseline operating characteristic model are updated based on Bayesian inference, thereby obtaining an instantaneous operating characteristic model that can fit the current power system behavior. Calculate the KL divergence between the real-time operating characteristic model and the baseline operating characteristic model to obtain the model correction index, which is used as the model correction index.

[0028] It should be noted that the model correction index quantifies the degree of system anomaly from the perspective of the deviation between the system's behavioral pattern and historical experience. The baseline operating characteristic model is a statistical model trained based on long-term historical normal operating data of the power system, used to characterize the transition patterns between state variables under normal operating conditions. The baseline operating characteristic model can be implemented using a dynamic Bayesian network probabilistic graphical model, whose parameters encode the conditional dependencies and state transition probabilities between each state variable under normal operating conditions. For example, under normal operating conditions, the conditional probability between "cutting off active load" and "significant voltage increase" is usually extremely low because active-voltage coupling is weak under large power grid conditions. Updating the baseline model based on Bayesian inference using real-time measurements refers to recursively updating the probability parameters in the model according to Bayes' theorem after receiving new measurement data, so that the updated real-time operating characteristic model can reflect the actual behavioral characteristics of the system within the current time window, including any newly emerging coupling relationships not reflected in historical normal data. KL divergence is an index in information theory used to measure the difference between two probability distributions. The KL divergence between the current operating characteristic model and the baseline operating characteristic model is calculated. A significant increase in the divergence value indicates a significant difference between the current system behavior pattern and the historical normal operating pattern, suggesting highly anomalous behavior beyond historical experience, such as the sudden manifestation of strong active-voltage coupling under islanded conditions. The model correction index is identified as a Type II risk indicator signal. The baseline operating characteristic model is implemented using a discrete dynamic Bayesian network. After discretization of the state variables, the state transition probabilities are stored in the form of conditional probability tables. The model parameters are trained based on historical normal operating data using the maximum likelihood estimation method.

[0029] It's important to note that the dispersion metric and the model correction metric assess system risk from two complementary dimensions: the dispersion metric focuses on whether the system is currently "operating smoothly," paying attention to the degree of divergence in the state trajectory; the model correction metric focuses on whether the system is currently "behaving normally," paying attention to the degree of deviation between the behavioral patterns and historical experience. Combining the two allows for a more comprehensive capture of the potential risks faced by the system, including both gradual operational degradation and sudden abnormal behavioral patterns. This dual-dimensional assessment mechanism does not rely on a pre-set fault rule base or a precise physical model, possessing unsupervised and adaptive risk detection capabilities.

[0030] S3. Combine the aforementioned dispersion index and model correction index to generate a comprehensive risk index.

[0031] It should be noted that the fusion refers to integrating the dispersion index and the model correction index, which assess system risk from different dimensions, into a single comprehensive risk index, so as to provide a unified and intuitive decision-making basis for subsequent graded response scheduling. Since the dispersion index and the model correction index have different physical meanings, numerical dimensions, and dynamic ranges, they cannot be directly weighted and combined; therefore, they need to be normalized separately first.

[0032] In an optional embodiment, in S3, the dispersion index and the model correction index are fused to generate a comprehensive risk index, including: The dispersion index and the model correction index are normalized respectively; The normalized dispersion index and the model correction index are weighted and summed according to preset weights to obtain the comprehensive risk index.

[0033] It should be noted that the normalization process refers to mapping the dispersion index and the model correction index to a unified numerical range, such as [0,1], to eliminate the influence of differences in units and dynamic range on the fusion results. Normalization can employ a maximum-minimum normalization method, which uses the minimum and maximum values ​​recorded in the historical operation of each index as a benchmark to linearly map the current value to the [0,1] range. For example, the historical minimum value of the dispersion growth rate is 0, and the historical maximum value is... The normalized dispersion index is the current dispersion growth rate divided by . The historical baseline value of the model correction index is The highest historical value Then the normalized model correction index is Through normalization, the two indicators have the same numerical scale when fused, avoiding the situation where one indicator dominates the calculation of the comprehensive risk index due to its larger absolute value.

[0034] It should be noted that the preset weight and These reflect the relative importance of the dispersion index and the model correction index in the comprehensive risk assessment. The weight values ​​can be set according to the actual operating characteristics and risk preferences of the power system. For example, for systems with variable operating conditions and frequent disturbances, the weight of the dispersion index can be appropriately increased to enhance sensitivity to operational disturbances; for systems with abundant historical operating data and relatively fixed behavior patterns, the weight of the model correction index can be appropriately increased to enhance the ability to detect unknown abnormal behaviors. In a typical configuration, the weight of the model correction index can be increased. and Both are set to 0.5, giving them equal importance.

[0035] It should be noted that the comprehensive risk index fully quantifies the potential risk level currently faced by the system. When the comprehensive risk index is at a low level, it indicates that the system is operating smoothly and behaving normally; when the comprehensive risk index rises, it indicates that the system is experiencing operational disorder or abnormal behavior, requiring the triggering of corresponding levels of response measures based on the degree of increase. By integrating multi-dimensional risk information into a single index, the triggering logic for subsequent tiered responses is simplified, improving the efficiency and consistency of scheduling decisions.

[0036] S4. When the comprehensive risk index meets the first preset condition, emergency stabilization control is triggered, the adjustable load is controlled to provide enhanced frequency damping support and voltage damping support, and flexible load reduction is performed.

[0037] In an optional embodiment, in S4, the first preset condition is that the comprehensive risk index exceeds a preset first threshold.

[0038] It should be noted that S4 corresponds to Level 1 in the hierarchical coordinated response scheduling, namely, emergency stability control response. This level is designed to address emergency situations where the system faces severe risks and requires immediate intervention. The preset first threshold is the risk criterion for triggering this level of response. When the comprehensive risk index exceeds this threshold, it indicates that the degree of system turbulence or abnormal behavior has reached a level that may endanger transient stability, requiring immediate control measures without delay. The specific value of the first threshold can be set based on the stability margin and operating experience of the actual power system.

[0039] It is important to note that the core feature of emergency stability control response lies in local autonomous execution. When the triggering conditions are met, each local control unit does not need to wait for instructions from the central dispatch unit. Instead, it directly adjusts the power according to the preset control law based on locally measured frequency deviation Δf and voltage deviation ΔV. This local autonomous execution mechanism can respond within milliseconds, avoiding missed control windows due to communication delays or time-consuming central decision-making calculations, which is crucial for ensuring the transient stability of the system.

[0040] In an optional embodiment, in S4, controlling the adjustable load to provide enhanced frequency damping support and voltage damping support includes: The adjustable load is controlled to adjust the power. The power adjustment amount is the negative value of the sum of the product of the frequency deviation and the preset frequency damping coefficient and the product of the voltage deviation and the preset voltage damping coefficient, wherein the frequency damping coefficient and the voltage damping coefficient are respectively greater than the corresponding damping coefficients in the normal operation mode. The implementation of flexible load reduction includes controlling lower-priority loads to perform gradual power reduction operations.

[0041] It should be noted that enhanced frequency damping support and voltage damping support refer to controlling the adjustable load to simulate the damping characteristics of a traditional synchronous generator, simultaneously providing active-frequency and reactive-voltage damping effects to the system. Power adjustment amount. The calculation formula is: The negative sign indicates that the frequency decreases (where the negative sign indicates that the frequency decreases). (negative) or voltage drop ( When the frequency is negative, the load reduces power consumption to provide power support to the system; conversely, when the frequency or voltage increases, the load increases power consumption to suppress further increases in frequency and voltage. Frequency damping coefficient. and voltage damping coefficient This is a preset damping coefficient, significantly larger than the corresponding damping coefficient under normal operating conditions, to provide stronger damping support in emergency situations. This enhanced damping control sacrifices some economic efficiency to maximize the suppression of system frequency and voltage fluctuations. Adjustable loads capable of providing this type of damping support include energy storage systems, variable frequency air conditioners, controllable heat pumps, and other equipment with rapid power regulation capabilities.

[0042] It's important to note that flexible load shedding refers to a gradual power reduction operation on lower-priority loads, rather than a traditional instantaneous hard cutoff. Gradual power reduction involves progressively reducing the load's power consumption according to a preset reduction rate or a stepped reduction curve. The advantages of this approach are twofold: firstly, it avoids the secondary power surges caused by instantaneous hard cutoffs, preventing new frequency or voltage fluctuations; secondly, it buys time for the central dispatch unit to perform subsequent refined scheduling. After the system's transient stability is initially contained, the Level 2 adaptive optimization dispatch can more precisely adjust the load. Load priorities can be preset based on factors such as user electricity consumption characteristics, contractual agreements, or system safety importance, with lower-priority loads participating in flexible load shedding first.

[0043] S5. When the comprehensive risk index meets the second preset condition, the abnormal coupling mode is located based on the model correction degree index, and a penalty term is introduced into the preset optimized scheduling model for correction to generate scheduling instructions.

[0044] In an optional embodiment, in S5, the second preset condition is that the comprehensive risk index is between a preset first threshold and a preset second threshold, or falls from above the first threshold to between the first threshold and the second threshold, wherein the first threshold is higher than the second threshold. It should be noted that S5 corresponds to Level 2 in the hierarchical collaborative response scheduling, namely, adaptive optimization scheduling response. This level is activated when the system has been temporarily stabilized by the emergency stabilization control of Level 1 but risks still exist, or when the comprehensive risk index is at a median level, has not yet reached an emergency level but has potential risks. The second preset condition defines the trigger window for Level 2: when the comprehensive risk index is between the first and second thresholds, it indicates that the system has a moderate level of risk and requires fine-tuning to eliminate hidden dangers; when the comprehensive risk index falls from above the first threshold to between the two thresholds, it indicates that the emergency control of Level 1 has initially curbed the deterioration of the risk, and at this time, Level 2 is entered for fundamental governance. The first threshold is higher than the second threshold, forming a transition range between the two, avoiding frequent switching of control modes near a single threshold.

[0045] A penalty term is introduced into the preset optimized scheduling model for correction in order to generate scheduling instructions, including: the preset optimized scheduling model is an optimized scheduling model with the goal of minimizing scheduling cost, and the optimized scheduling model includes a scheduling cost function; Based on the identified abnormal coupling pattern, a risk coupling penalty term specifically designed for that abnormal coupling pattern is added to the scheduling cost function to obtain the corrected optimized scheduling model. The scheduling instructions are generated based on the revised optimized scheduling model.

[0046] It should be noted that the preset optimized scheduling model is a mathematical model aimed at minimizing scheduling costs, with the scheduling cost function at its core. Under normal operating conditions, this model performs optimization calculations based on the traditional active-frequency and reactive-voltage decoupling assumptions to generate the most economically efficient scheduling instructions. However, when abnormal coupling modes exist in the system, using the original decoupling optimization model may generate dangerous scheduling instructions that trigger cascading failures. For example, to cope with frequency drops, the traditional optimization model might suggest cutting off some active loads to restore active balance. However, in the case of islanded operation with strong PV coupling, this operation may lead to abnormal voltage increases, thereby triggering PV grid disconnection, exacerbating the active power deficit, and forming a cascading failure.

[0047] In an optional embodiment, in S5, locating the abnormal coupling mode based on the model correction index includes: For the state transition sequence that causes the model correction index to increase, by comparing the parameter differences of the state transition sequence in the baseline operating characteristic model and the real-time operating characteristic model, the abnormal coupling mode in which the coupling relationship between state variables is abnormally enhanced is located. It is important to note that identifying anomalous coupling patterns is the core analytical step in Level Two. When the model correction index increases, it indicates a significant difference between the current system behavior and historical normal patterns. By analyzing the state transition sequences that lead to the increased model correction index—that is, tracing which state variable changes caused the greatest difference in parameters between the instantaneous model and the baseline model—the anomalous coupling patterns can be located. Specifically, the baseline operating characteristic model encodes the conditional dependencies between state variables under normal operating conditions. For example, under normal operating conditions, the conditional probability between "disconnecting active power load" and "voltage rise" is low. However, if the instantaneous operating characteristic model, updated based on real-time data, shows a significant increase in these conditional probabilities, it indicates an anomalous strong active power-voltage coupling phenomenon in the system. By comparing the parameter differences between the two models on the same state transition sequence, it is possible to identify which state variables have experienced an anomalously enhanced coupling relationship.

[0048] When the identified abnormal coupling mode is a strong coupling between a decrease in active power and an abnormal increase in voltage, and the abnormal coupling mode causes the model correction index to increase, the risk coupling penalty term is constructed as follows: the product of the penalty coefficient and the sum of the absolute values ​​of the active loads to be cut off for each load, multiplied by an indicator function, wherein the indicator function takes a value of 1 when the voltage change predicted based on the real-time operating characteristic model is greater than a preset voltage safety threshold, and otherwise takes a value of 0, and the penalty coefficient is a coefficient positively correlated with the model correction index.

[0049] It should be noted that when the identified abnormal coupling mode is a strong coupling between a significant decrease in active power and an abnormal increase in voltage, the risk coupling penalty term is constructed as follows: ;in, Let i be the active load of the i-th load to be removed. The voltage change is predicted based on the real-time operating characteristic model. For indicator functions, when The value is 1 if the condition is met, and 0 otherwise. The preset voltage safety threshold, This is a penalty coefficient that is positively correlated with the model correction index.

[0050] It should be noted that the introduction of the risk coupling penalty term is precisely to address the aforementioned problem. This penalty term is an additional cost term specifically constructed for the identified abnormal coupling mode. Its function is to impose a high penalty on control behaviors that may trigger abnormal coupling effects during the optimization process of the scheduling model, thereby guiding the optimization algorithm to automatically avoid these dangerous behaviors and select a safer scheduling scheme. Taking the identified abnormal coupling mode as a strong coupling between a decrease in active power and an abnormal increase in voltage as an example, the risk coupling penalty term is constructed as follows: .in, The sum of the absolute values ​​of the active loads to be cut from each load represents the total scale of load shedding. As an indicator function, it represents the voltage change predicted based on the real-time operating characteristic model. Exceeding the preset voltage safety threshold The value is 1 if the condition is met, and 0 otherwise. This is a penalty coefficient positively correlated with the model correction index. The meaning of this penalty term is: when a load shedding scheme is predicted to cause a voltage rise exceeding the safety threshold, a penalty proportional to the load shedding scale and the model correction index is imposed on that scheme; the more load shedding and the higher the model anomaly, the heavier the penalty. Penalty Coefficient The positive correlation with the model correction index ensures that the penalty intensity matches the degree of system anomaly. By incorporating this penalty term into the scheduling cost function, the optimization algorithm, in minimizing the total cost, will proactively avoid load shedding schemes that may induce voltage overruns, and instead choose safer scheduling instructions such as calling the energy storage system to simultaneously issue active power and absorb reactive power, and adopting flexible load smoothing, thus eliminating cascading risks at the decision-making level.

[0051] It should be noted that the form of the risk coupling penalty term is not limited to the examples above. Different forms of penalty terms can be constructed depending on the identified abnormal coupling pattern. For example, if a strong coupling is found between large changes in reactive power and abnormal frequency fluctuations, a corresponding penalty term can be constructed to suppress drastic adjustments in reactive power. The core idea is to quantify the abnormal coupling relationship as a penalty for scheduling costs, enabling the optimized scheduling model to automatically possess risk avoidance capabilities while pursuing economic efficiency.

[0052] Example 2, please refer to Figure 2 This invention provides a multi-user electricity load aggregation response scheduling system, comprising: The trajectory modeling module is used to acquire real-time measurement values ​​of multiple key state variables and construct an instantaneous trajectory probability distribution based on the real-time measurement values. The dual-dimensional evaluation module is used to calculate a dispersion index reflecting the degree of power system disorder based on the instantaneous trajectory probability distribution, and to calculate a model correction index reflecting the degree of deviation of power system behavior from historical normal patterns. The risk fusion module is used to fuse the dispersion index and the model correction index to generate a comprehensive risk index. The emergency control module is used to trigger emergency stabilization control when the comprehensive risk index meets the first preset condition, control the adjustable load to provide enhanced frequency damping support and voltage damping support, and perform flexible load reduction; The adaptive scheduling module is used to locate the abnormal coupling mode based on the model correction index when the comprehensive risk index meets the second preset condition, and accordingly introduce a penalty term into the preset optimized scheduling model for correction to generate scheduling instructions.

[0053] In summary, this invention solves the cascading failure problem caused by the mismatch between traditional decoupled control logic and the strong coupling characteristics of the system under islanded operating conditions, and significantly improves the resilience of the power grid.

[0054] Furthermore, any content not described in detail in this specification is existing technology known to those skilled in the art.

[0055] In the embodiments provided by this invention, it should be understood that the disclosed system or method can be implemented in other ways. For example, the embodiments of the invention described above are merely illustrative; for instance, the division of modules is only a logical functional division, and there may be other division methods in actual implementation.

[0056] The modules described as separate components may or may not be physically separate. The components shown as modules may or may not be physical modules; they may be located in one place or distributed across multiple network modules. Some or all of the modules can be selected to achieve the purpose of this embodiment according to actual needs.

[0057] Furthermore, the functional modules in the various embodiments of the present invention can be integrated into one processing module, or each module can exist physically separately, or two or more modules can be integrated into one module. The integrated module can be implemented in hardware or in the form of hardware plus software functional modules.

[0058] For those skilled in the art, it is obvious that the present invention is not limited to the details of the above exemplary embodiments, and that the present invention can be implemented in other specific forms without departing from the basic characteristics of the present invention.

[0059] The above description is only a preferred embodiment of the present invention, but the scope of protection of the present invention is not limited thereto. Any equivalent substitutions or modifications made by those skilled in the art within the scope of the technology disclosed in the present invention, based on the technical solution and inventive concept of the present invention, should be covered within the scope of protection of the present invention.

Claims

1. A multi-user electricity load aggregation response scheduling method, characterized in that, Includes the following steps: S1. Obtain real-time measurement values ​​of multiple key state variables, and construct an instantaneous trajectory probability distribution based on the real-time measurement values; S2. Based on the instantaneous trajectory probability distribution, calculate the dispersion index reflecting the degree of power system disorder, and calculate the model correction index reflecting the degree of power system behavior deviating from the historical normal pattern. S3. Combine the aforementioned dispersion index and model correction index to generate a comprehensive risk index; S4. When the comprehensive risk index meets the first preset condition, emergency stabilization control is triggered, the adjustable load is controlled to provide enhanced frequency damping support and voltage damping support, and flexible load reduction is performed. S5. When the comprehensive risk index meets the second preset condition, the abnormal coupling mode is located based on the model correction degree index, and a penalty term is introduced into the preset optimized scheduling model for correction to generate scheduling instructions.

2. The multi-user electricity load aggregation response scheduling method as described in claim 1, characterized in that: In S1, the instantaneous trajectory probability distribution is constructed based on the real-time measurement values, including: The multidimensional state space composed of multiple key state variables is discretized into a preset number of state cells. Within a preset sliding time window, the frequency of the power system state trajectory falling into each of the state cells is counted; The instantaneous trajectory probability distribution is constructed based on the frequency; Among them, the multiple key state variables include at least: power system frequency, key node voltage, aggregated load total active power, total reactive power, and distributed generation output.

3. The multi-user electricity load aggregation response scheduling method as described in claim 1, characterized in that: In S2, the dispersion index reflecting the degree of power system disorder is calculated, including: Based on the instantaneous trajectory probability distribution, the state trajectory dispersion index is calculated. The state trajectory dispersion index is the Shannon entropy. When the probability of a certain state cell in the instantaneous trajectory probability distribution is zero, the Shannon entropy term corresponding to the state cell takes the value of zero. Calculate the time change rate of the dispersion index of the state trajectory to obtain the dispersion growth rate, which is used as the dispersion index.

4. The multi-user electricity load aggregation response scheduling method as described in claim 1, characterized in that: In S2, the model correction index, which reflects the degree to which the power system behavior deviates from the historical normal pattern, is calculated, including: Acquire historical normal operation data and train and construct a baseline operation characteristic model based on the historical normal operation data; By utilizing real-time measurements of multiple key state variables, the parameters or state probability distributions of the baseline operating characteristic model are updated based on Bayesian inference, thereby obtaining an instantaneous operating characteristic model that can fit the current power system behavior. Calculate the KL divergence between the real-time operating characteristic model and the baseline operating characteristic model to obtain the model correction index, which is used as the model correction index.

5. The multi-user electricity load aggregation response scheduling method as described in claim 1, characterized in that: In S3, the dispersion index and the model correction index are combined to generate a comprehensive risk index, including: The dispersion index and the model correction index are normalized respectively; The normalized dispersion index and the model correction index are weighted and summed according to preset weights to obtain the comprehensive risk index.

6. The multi-user electricity load aggregation response scheduling method as described in claim 1, characterized in that: In S4, the first preset condition is that the comprehensive risk index exceeds a preset first threshold.

7. The multi-user electricity load aggregation response scheduling method as described in claim 6, characterized in that: In S4, the adjustable load is controlled to provide enhanced frequency damping support and voltage damping support, including: The adjustable load is controlled to adjust the power. The power adjustment amount is the negative value of the sum of the product of the frequency deviation and the preset frequency damping coefficient and the product of the voltage deviation and the preset voltage damping coefficient, wherein the frequency damping coefficient and the voltage damping coefficient are respectively greater than the corresponding damping coefficients in the normal operation mode. The implementation of flexible load reduction includes controlling lower-priority loads to perform gradual power reduction operations.

8. The multi-user electricity load aggregation response scheduling method as described in claim 4, characterized in that: In S5, the second preset condition is that the comprehensive risk index is between a preset first threshold and a preset second threshold, or falls from above the first threshold to between the first threshold and the second threshold, wherein the first threshold is higher than the second threshold. A penalty term is introduced into the preset optimized scheduling model for correction in order to generate scheduling instructions, including: the preset optimized scheduling model is an optimized scheduling model with the goal of minimizing scheduling cost, and the optimized scheduling model includes a scheduling cost function; Based on the identified abnormal coupling pattern, a risk coupling penalty term specifically designed for that abnormal coupling pattern is added to the scheduling cost function to obtain the corrected optimized scheduling model. The scheduling instructions are generated based on the revised optimized scheduling model.

9. The multi-user electricity load aggregation response scheduling method as described in claim 8, characterized in that: In S5, the abnormal coupling mode is located based on the model correction index, including: For the state transition sequence that causes the model correction index to increase, by comparing the parameter differences of the state transition sequence in the baseline operating characteristic model and the real-time operating characteristic model, the abnormal coupling mode in which the coupling relationship between state variables is abnormally enhanced is located. When the identified abnormal coupling mode is a strong coupling between a decrease in active power and an abnormal increase in voltage, and the abnormal coupling mode causes the model correction index to increase, the risk coupling penalty term is constructed as follows: the product of the penalty coefficient and the sum of the absolute values ​​of the active loads to be cut off for each load, multiplied by an indicator function, wherein the indicator function takes a value of 1 when the voltage change predicted based on the real-time operating characteristic model is greater than a preset voltage safety threshold, and otherwise takes a value of 0, and the penalty coefficient is a coefficient positively correlated with the model correction index.

10. A multi-user electricity load aggregation response scheduling system, employing the multi-user electricity load aggregation response scheduling method as described in any one of claims 1-9, characterized in that, include: The trajectory modeling module is used to acquire real-time measurement values ​​of multiple key state variables and construct an instantaneous trajectory probability distribution based on the real-time measurement values. The dual-dimensional evaluation module is used to calculate a dispersion index reflecting the degree of power system disorder based on the instantaneous trajectory probability distribution, and to calculate a model correction index reflecting the degree of deviation of power system behavior from historical normal patterns. The risk fusion module is used to fuse the dispersion index and the model correction index to generate a comprehensive risk index. The emergency control module is used to trigger emergency stabilization control when the comprehensive risk index meets the first preset condition, control the adjustable load to provide enhanced frequency damping support and voltage damping support, and perform flexible load reduction. The adaptive scheduling module is used to locate the abnormal coupling mode based on the model correction index when the comprehensive risk index meets the second preset condition, and accordingly introduce a penalty term into the preset optimized scheduling model for correction to generate scheduling instructions.

Images