Demand response intelligent dispatching method and system based on power knowledge graph

Abstract

The present application relates to the technical field of intelligent scheduling, and discloses a demand response intelligent scheduling method and system based on a power knowledge graph, which comprises the following steps: defining feeder segment nodes in the power knowledge graph, injecting time distribution, thermal inertia and power-off duration and other time semantic attributes, and accordingly analyzing the inherent lock time thermal inertia resonance window spectrum of the power restoration block due to the coupling of device lock time concentration release and thermal demand. A peak exposure function is constructed to quantify the system instantaneous current peak value after the inrush current of the to-be-closed block and the peak superposition of the secondary synchronization start wave of the previous block. Based on the function, optimization is performed within the scheduling clock to preferentially select a time point that avoids the resonance window spectrum to avoid peak superposition; if it cannot be avoided, the minimum quota of the desynchronization type demand response is calculated only for the secondary synchronization component. The strategy containing the optimal execution time and the execution quota is output and executed at the present period. The present application solves the window collision misjudgment problem caused by the traditional model and realizes low-cost precise peak shaving in the power restoration process.

Description

Technical Field

[0001] This invention relates to the field of intelligent scheduling technology, and more specifically, to a demand response intelligent scheduling method and system based on power knowledge graphs. Background Technology

[0002] In power distribution network restoration and demand response scenarios, numerous household and commercial refrigeration devices periodically start and stop using temperature control logic to maintain indoor temperatures. When a power outage lasts for a period, these devices simultaneously enter a standby state. However, equipment manufacturers typically incorporate short-cycle timeouts (usually 3 to 10 minutes) in their compressor control circuits, causing the devices to not start immediately upon power restoration, but rather to be put into operation in batches at the moment the timeout expires. Simultaneously, the thermal inertia of the building envelope and refrigeration system causes the ambient temperature to rise continuously during the outage, further increasing the load demand after startup. The combination of these two mechanisms creates a secondary startup peak that is delayed in time and abruptly increases in magnitude, making the power restoration sequence and load scheduling highly time-sensitive. Existing dispatching systems often only monitor instantaneous power changes or execute power restoration commands at uniform intervals, failing to capture and predict these short-term concentrated startup phenomena caused by both equipment control logic and thermal processes.

[0003] Specifically, if power restoration dispatching occurs during such peak startup periods, it will have a cumulative effect with the section being restored, causing the line current to exceed the protection setpoint and triggering overcurrent or timed overcurrent protection actions. Furthermore, the thermal inertia of transformers prevents the immediate release of heat accumulated during short-term peak loads, leading to excessive temperature rise in hotspots and accelerated insulation aging. In demand response scenarios, if the dispatching platform fails to identify these potential synchronization periods, it may issue peak-shaving or start-stop commands during peak load periods, exacerbating local voltage fluctuations and system instability. While existing intelligent dispatching algorithms can utilize historical data for load forecasting, they lack a knowledge carrier that unifies the description of equipment control logic, thermal dynamics, and dispatching events, making it impossible to identify and avoid such sudden peak loads in advance over time. Summary of the Invention

[0004] This invention provides a demand response intelligent scheduling method and system based on power knowledge graph, which solves the technical problems mentioned in the background art.

[0005] The first aspect is a demand response intelligent scheduling method based on power knowledge graphs, including:

[0006] In the power knowledge graph, feeder segment nodes are defined, and time-locking distribution, thermal inertia parameters, and power outage duration are associated with feeder segment nodes as time-history semantic attributes.

[0007] Based on the temporal semantic attributes, the inherent characteristics of the previously closed blocks are analyzed to generate the time-locked thermal inertia resonance window spectrum. The time-locked thermal inertia resonance window spectrum defines the time domain interval of the secondary synchronous start-up peak caused by the concentrated release of equipment lock-up and the thermal inertia coupling of the load.

[0008] A peak exposure function is constructed to quantify the system instantaneous current peak value of the block to be closed under different delay access conditions. The system instantaneous current peak value is composed of the closing inrush current of the block to be closed, the cold load recovery margin of the previously closed block, and the peak value of the secondary synchronous start-up.

[0009] Based on the peak exposure function, the execution time is optimized within the scheduling clock. Priority is given to avoiding the time-locked thermal inertia resonance window spectrum. If it cannot be avoided or exceeds the limit, the minimum quota for desynchronized demand response is calculated for the intensity of the secondary synchronization start peak.

[0010] The output includes the execution time and the scheduling strategy with the minimum quota for desynchronized demand response, and is executed in the current period.

[0011] Secondly, a demand response intelligent scheduling system based on power knowledge graphs, applied to any of the aforementioned demand response intelligent scheduling methods based on power knowledge graphs, includes:

[0012] The power knowledge graph construction module defines feeder segment nodes in the power knowledge graph and associates time-locking distribution, thermal inertia parameters, and power outage duration as time-history semantic attributes to the feeder segment nodes.

[0013] The time-domain interval definition module analyzes the inherent characteristics of the previously closed block based on the time history semantic attributes to generate a time-locked thermal inertia resonance window spectrum. The time-locked thermal inertia resonance window spectrum defines the time-domain interval of the secondary synchronous start-up peak caused by the concentrated release of equipment time-locking and the thermal inertia coupling of the load.

[0014] The peak exposure function construction module constructs a peak exposure function to quantify the system instantaneous current peak value of the block to be closed under different delay access conditions. The system instantaneous current peak value is formed by superimposing the closing inrush current of the block to be closed onto the cold load recovery margin of the previously closed block and the secondary synchronous start peak.

[0015] The demand response quota calculation module optimizes the generation of execution time within the scheduling clock based on the peak exposure function, and prioritizes the selection of time periods that avoid the thermal inertia resonance window spectrum of the lock time. If it cannot be avoided or exceeds the limit, the minimum quota for desynchronized demand response is calculated based on the intensity of the secondary synchronization start peak.

[0016] The quota execution module outputs a scheduling strategy that includes the execution time and the minimum quota for desynchronized demand response, and executes it in the current period.

[0017] The beneficial effects of this invention include: by deconstructing the cross-domain joint mechanism of equipment control logic and load thermodynamics in the power knowledge graph, the time-locked thermal inertia resonance window spectrum in the power restoration process is quantified, thereby enabling accurate identification and avoidance of the risk of secondary synchronization peak superposition that cannot be predicted by the traditional monotonic decay model, or, when it is impossible to avoid, calculating the minimum desynchronization demand response quota only for specific resonance components, thus achieving inherent safety and precise control of the power distribution network restoration process with the lowest control cost. Attached Figure Description

[0018] Figure 1 This is a flowchart of the intelligent demand response scheduling method based on power knowledge graph of the present invention;

[0019] Figure 2 This is a schematic diagram illustrating a specific implementation of the present invention. Detailed Implementation

[0020] The subject matter described herein will now be discussed with reference to exemplary embodiments. It should be understood that these embodiments are discussed only to enable those skilled in the art to better understand and implement the subject matter described herein, and changes may be made to the function and arrangement of the elements discussed without departing from the scope of this specification. Various processes or components may be omitted, substituted, or added as needed in the examples. Furthermore, features described in some examples may be combined in other examples.

[0021] Example 1: As Figure 1 As shown, the intelligent demand response scheduling method based on power knowledge graph includes:

[0022] In the power knowledge graph, feeder segment nodes are defined, and time-locking distribution, thermal inertia parameters, and power outage duration are associated with feeder segment nodes as time-history semantic attributes.

[0023] Based on the temporal semantic attributes, the inherent characteristics of the previously closed blocks are analyzed to generate the time-locked thermal inertia resonance window spectrum. The time-locked thermal inertia resonance window spectrum defines the time domain interval of the secondary synchronous start-up peak caused by the concentrated release of equipment lock-up and the thermal inertia coupling of the load.

[0024] A peak exposure function is constructed to quantify the system instantaneous current peak value of the block to be closed under different delay access conditions. The system instantaneous current peak value is composed of the closing inrush current of the block to be closed, the cold load recovery margin of the previously closed block, and the peak value of the secondary synchronous start-up.

[0025] Based on the peak exposure function, the execution time is optimized within the scheduling clock. Priority is given to avoiding the time-locked thermal inertia resonance window spectrum. If it cannot be avoided or exceeds the limit, the minimum quota for desynchronized demand response is calculated for the intensity of the secondary synchronization start peak.

[0026] The output includes the execution time and the scheduling strategy with the minimum quota for desynchronized demand response, and is executed in the current period.

[0027] In one embodiment of the present invention, feeder segment nodes are defined in a power knowledge graph, and time-locking distribution, thermal inertia parameters, and outage duration are associated with the feeder segment nodes as time-history semantic attributes, including:

[0028] Extract load characteristic data for the corresponding area in the power grid, and write the lockout time distribution, thermal inertia parameters, power outage duration and ambient temperature difference as basic time history semantic attributes into the feeder segment node;

[0029] Based on thermal inertia parameters and power outage duration, a nodal post-power-on temperature function is constructed to characterize the exponential recovery of spatial temperature over time after power restoration.

[0030] By differentiating the lock-time distribution with respect to time, the lock-time arrival density, which characterizes the probability of the device's anti-short cycle timer being reset at a specific moment, is calculated.

[0031] Multiply the lockout density by the probability that the temperature function after the node is powered on is higher than the temperature control threshold, and calculate the secondary synchronous start rate, which represents the probability of instantaneous start-up of the equipment group after power-on due to the overlap of lockout release and heat demand.

[0032] The secondary synchronization start rate is stored as a derived attribute in the feeder segment node.

[0033] Load characteristic data is a collection of electricity consumption characteristics of various electrical equipment in the corresponding area of ​​the power grid. It covers data such as equipment type, operating period, starting current, and power change pattern. The sources include power grid terminal monitoring equipment, electricity enterprise ledgers, and historical operation records.

[0034] Lock time distribution is the statistical distribution of the shutdown lock time set by the anti-short cycle timer of refrigeration equipment in a region. It reflects the proportion of equipment with different lock time. For example, most equipment in the region has a lock time of 5 minutes, while a few have lock times of 3 minutes and 10 minutes, forming a distribution feature with 5 minutes as the core.

[0035] Thermal inertia is a parameter that characterizes the ease with which a building envelope and refrigeration system store and release heat. The larger the value, the slower the temperature change. The value needs to be determined in combination with the building material, insulation performance, and power of the refrigeration equipment.

[0036] The duration of a power outage is the time from the cessation of power supply to the restoration of power supply in the corresponding area of ​​the power grid, which can be accurately obtained through power grid dispatch records or power outage timestamps from distribution terminals.

[0037] The ambient temperature difference is the difference between the actual ambient temperature in the area and the set temperature of the refrigeration equipment. The larger the difference, the stronger the demand for the refrigeration equipment to start. It is calculated by reading the ambient temperature sensor and the set parameters of the equipment.

[0038] The temperature function after a node is restored is a mathematical expression describing the change of regional spatial temperature over time after power restoration. It primarily reflects the influence of thermal inertia and the duration of power outage on temperature recovery, and can accurately predict the spatial temperature at different times. Specifically, the temperature at a certain moment after a node is restored is equal to the set temperature of the refrigeration equipment plus the ambient temperature difference multiplied by (1 minus the exponential term with the natural constant as the base, and the exponential term being (the sum of the power outage duration and the time elapsed after power restoration divided by the thermal inertia parameter)).

[0039] Lock-time arrival density reflects the probability that the anti-short cycle timer of refrigeration equipment in a certain area at a specific time will end its lock-time and meet the start-up conditions. The higher the value, the more equipment that meets the start-up conditions at that time. Specifically, the lock-time arrival density is obtained by performing a time-dimension derivative operation on the lock-time distribution and taking the derivative result of the lock-time distribution function at the corresponding time.

[0040] The temperature control threshold is the critical temperature value for starting the refrigeration equipment. It is the temperature standard used to determine whether the equipment needs to be started in the temperature function after the node is powered on. It is usually set to the set temperature of the refrigeration equipment plus half of the temperature control bandwidth. The temperature control bandwidth is determined according to the equipment's factory parameters.

[0041] The probability that the temperature function after a node is restored is higher than the temperature control threshold is the likelihood that the space temperature will exceed the temperature control threshold at a certain moment after the node is restored. It is used to determine whether the equipment has a cooling requirement. A probability of 1 indicates that all equipment has a cooling requirement, and a probability of 0 indicates that no equipment has a cooling requirement. Specifically, the probability value is obtained by statistically analyzing the number of times the temperature was higher than the temperature control threshold at the corresponding moment after power restoration under the same historical conditions of the same period, the same power outage duration, and thermal inertia parameters in the area, and dividing the result by the total number of statistical counts.

[0042] The secondary synchronous start-up rate reflects the probability that the equipment group will start instantaneously after power restoration due to the end of the lock time and the superimposed heat demand. It is a core parameter for constructing the resonance window spectrum. The higher the value, the greater the risk of concentrated equipment startup. Specifically, the secondary synchronous start-up rate is equal to the lock time arrival density multiplied by the probability that the temperature function of the node after power restoration is higher than the temperature control threshold.

[0043] In one embodiment of the present invention, based on time-history semantic attributes, the inherent characteristics of previously closed blocks are parsed to generate a time-locked thermal-inertial resonance window spectrum, including:

[0044] Obtain the attributes of the previously closed blocks in the power knowledge graph, and calculate the center of the resonance window by finding the time point where the product of the lockout arrival density and the heat demand survival probability reaches its maximum value; the heat demand survival probability is composed of an exponential term minus an exponential term with the natural constant as the base and the sum of the power outage duration and time variable divided by the negative value of the thermal inertia parameter.

[0045] The square root of the sum of the variance of the lock-time distribution and the weighted squared terms of the thermal inertia parameter is used to obtain the resonant window width.

[0046] Using the center of the resonance window as the midpoint and the width of the resonance window as half the width, a time-locked thermal inertia resonance window spectrum is constructed for the peak period of secondary synchronous start-up.

[0047] The attributes of the already closed blocks in the power knowledge graph are a set of basic and derived attributes that have been marked and stored. These include basic attributes such as lockout distribution, thermal inertia parameters, and power outage duration, as well as derived attributes such as secondary synchronization start-up rate. These attributes are the core data support for generating the lockout thermal inertia resonance window spectrum.

[0048] The probability of heat demand persistence is the probability that refrigeration equipment in a certain area will still have a cooling demand at a certain moment after power is restored. The larger the value, the stronger the potential demand for equipment to start. Specifically, the probability of heat demand persistence is equal to a negative exponential term (the sum of power outage duration and time variable divided by the thermal inertia parameter) minus the natural constant.

[0049] The center of the resonance window is the moment when the probability of concentrated startup of equipment in the previously closed block due to the overlap of lock-up release and heat demand is the highest. It is the core reference point of the lock-up thermal inertia resonance window spectrum. Specifically, the effective time range after power restoration is traversed, the product of the lock-up arrival density and the heat demand survival probability corresponding to each time point is calculated, and the time point with the largest product value is taken as the center of the resonance window.

[0050] Lock-in time distribution variance is a parameter that reflects the dispersion of lock-in time of refrigeration equipment in a region. The larger the variance, the more dispersed the lock-in time distribution. It is calculated based on statistical data of lock-in time distribution.

[0051] The width coefficient is a weighted coefficient for the thermal inertia parameter, used to adjust the influence of the thermal inertia parameter in the calculation of the resonant window width. Its value range is fixed from 0.1 to 0.3.

[0052] The weighted square term of the thermal inertia parameter is the square result of the thermal inertia parameter after adjustment by the width coefficient. It is used to reflect the weighted influence of thermal inertia on the width of the resonant window. Specifically, the weighted square term of the thermal inertia parameter is equal to the thermal inertia parameter multiplied by the width coefficient, and then the product is squared.

[0053] The resonance window width is an indicator of the time span of the time-locked thermal inertia resonance window spectrum, which determines the size of the high-start-risk period. Specifically, the resonance window width is equal to the sum of the variance of the time-locked distribution and the weighted square term of the thermal inertia parameter, and then the square root of the sum is taken.

[0054] The time-locked thermal inertia resonance window spectrum defines the time domain interval where the peak of secondary synchronous startup of the equipment is most frequent. Within this interval, closing the circuit breaker is prone to cause the risk of superposition of current peaks, and it is the core reference for peak avoidance in scheduling. Specifically, the start time of the time-locked thermal inertia resonance window spectrum is equal to the center of the resonance window minus the width of the resonance window, and the end time is equal to the center of the resonance window plus the width of the resonance window. The interval from the start time to the end time is the time-locked thermal inertia resonance window spectrum.

[0055] In one embodiment of the present invention, constructing the peak exposure function includes:

[0056] Obtain the cold load recovery amplitude coefficient and decay time constant of the previously closed block, calculate the product of the cold load recovery amplitude coefficient of the previously closed block and the exponential term with the natural constant as the base and the negative delay divided by the decay time constant as the exponent, and obtain the residual current of the previously closed block.

[0057] Obtain the peak inrush current of the block to be closed, the current conversion factor of the previously closed block, and the preset peak sampling half-window width;

[0058] The secondary synchronization start-up rate is integrated within the interval with the delay minus the peak sampling half-window width as the lower limit and the delay plus the peak sampling half-window width as the upper limit. The integral result is multiplied by the current conversion factor to obtain the secondary synchronization additive current.

[0059] The residual current of the previously closed block, the peak inrush current of the block to be closed, and the secondary synchronous boost current are added together to obtain the peak exposure function, which characterizes the extreme value of the total system load under different delays.

[0060] The cold load recovery amplitude coefficient is a parameter that characterizes the initial peak inrush current during the cold load recovery process of a previously closed block. The larger the value, the higher the peak current in the early stage of cold load recovery. It is derived based on the block's historical cold load recovery data and equipment rated current statistics.

[0061] The decay time constant is a parameter that describes how quickly the recovery current of the cold load in the previously closed block decays over time. The larger the value, the slower the current decays. It is determined in combination with the block load type and the impedance characteristics of the power grid line.

[0062] The residual current of the previously closed block is the cold load recovery current of the previously closed block that has not yet completely decayed when the block to be closed is connected with a delay. It is an important component of the total system current. Specifically, the residual current of the previously closed block is equal to the cold load recovery amplitude coefficient multiplied by an exponential term with the natural constant as the base. The exponent of this exponential term is the delay divided by the decay time constant.

[0063] The peak inrush current of the waiting-to-close block is the maximum instantaneous current generated at the moment of closing the waiting-to-close block. It is determined by the type, quantity, and rated power of the equipment in the waiting-to-close block and is obtained through equipment parameter ledgers and closing inrush current test data.

[0064] The current conversion factor is a conversion factor that converts the integral result of the secondary synchronous start-up rate into the actual current value. It is used to unify the dimensions of physical quantities and has a value range of 0.05 to 0.2, which is calibrated according to the block load type.

[0065] The peak sampling half-window width is a set half-value of the integral time interval, used to accurately capture the concentrated area of ​​the secondary synchronization start rate before and after the delay. The preset value range is ten to thirty seconds.

[0066] The secondary synchronization boost current is the superimposed contribution of the secondary synchronous start-up of equipment in the already closed block to the system current. It is the core component in the peak exposure function that reflects the risk of concentrated equipment start-up. Specifically, the secondary synchronization boost current is equal to the integral result of the secondary synchronization start-up rate in the specified interval multiplied by the current conversion factor of the already closed block. The lower limit of the specified interval is the delay minus the peak sampling half-window width, and the upper limit is the delay plus the peak sampling half-window width.

[0067] The peak exposure function is the core function that quantifies the instantaneous peak current of the system when accessing the block to be closed with different delays. Its result is directly used to determine the level of closing risk. Specifically, the peak exposure function is equal to the residual current of the previously closed block plus the inrush peak current of the block to be closed, plus the secondary synchronization boost current.

[0068] In one embodiment of the present invention, the execution time is optimally generated within the scheduling clock based on the peak exposure function, prioritizing the selection of time periods that avoid the time-locked thermal inertia resonance window spectrum. If it cannot be avoided or exceeds the limit, the minimum quota for desynchronized demand response is calculated based on the intensity of the secondary synchronization start-up peak, including:

[0069] Remove time periods that fall into the time-locked thermal inertia resonance window spectrum from the preset set of scheduling clocks to generate a window avoidance candidate delay set composed of safety time slots;

[0070] Obtain the preset system security threshold, and compare the peak exposure function calculated for the selected delay with the system security threshold;

[0071] When the peak exposure function is greater than the system safety threshold, calculate the difference between the peak exposure function and the system safety threshold;

[0072] Using the secondary synchronous addition current as the denominator, the difference is divided by the secondary synchronous addition current to obtain the minimum quota of desynchronized demand response, which characterizes the proportion of phase aggregation that needs to be reduced.

[0073] The preset scheduling clock set is a set of selectable closing times for the blocks to be closed in advance. It includes a series of continuous or discrete time points with uniformly set time intervals to cover the reasonable scheduling period after power restoration.

[0074] The window avoidance candidate delay set is the set of safe closing times after eliminating high-risk periods. Closing at times within the set can avoid the risk of overlapping peaks during secondary synchronization startup. Specifically, the window avoidance candidate delay set is equal to the set of all time points in the preset scheduling clock set that do not fall within the time-locked thermal inertial resonance window spectrum.

[0075] The preset system safety threshold is a critical value that defines whether the system current exceeds the limit. It is used to judge the current risk at the time of closing the circuit and is set comprehensively based on the line rated current, transformer carrying capacity and relay protection settings.

[0076] The selected delay is a specific closing time chosen from the candidate delay set for risk assessment. Each delay can be selected individually for peak exposure function calculation and safety judgment.

[0077] The difference is the portion of the peak exposure function that exceeds the system safety threshold. This value directly determines the amount of current that needs to be reduced and is the core basis for calculating the demand response quota. Specifically, the difference is equal to the result of the peak exposure function calculated for the selected delay minus the preset system safety threshold.

[0078] The minimum desynchronized demand response quota is the minimum load reduction ratio required to bring the system current back to within the safe threshold. It only targets the current component of secondary synchronization startup and will not excessively interfere with system operation. Specifically, the minimum desynchronized demand response quota is equal to the difference between the peak exposure function and the system safe threshold divided by the secondary synchronization boost current.

[0079] In one embodiment of the present invention, the output of a scheduling policy including the execution time and the minimum quota for desynchronized demand response, and its execution in the current period, includes:

[0080] Obtain the preset time weighting coefficient and demand response weighting coefficient;

[0081] For each delay in the candidate delay set for window avoidance, an optimization objective function is constructed. The optimization objective function consists of a time cost term representing the speed of power restoration and a demand response cost term representing the cost of control. The time cost term is the product of the delay and the time weight coefficient, and the demand response cost term is the product of the minimum quota of desynchronized demand response, the demand response triggering flag, and the demand response weight coefficient.

[0082] Minimize the objective function to determine the optimal execution time and the corresponding optimal demand response trigger flag that minimizes the total cost.

[0083] Multiply the optimal demand response trigger flag by the minimum quota of the desynchronized demand response corresponding to the optimal execution time to obtain the final execution quota, and output the scheduling strategy consisting of the optimal execution time and the final execution quota.

[0084] The time weighting coefficient is a parameter used to measure the impact of closing delay on the total dispatch cost. The larger the value, the more priority is given to shortening the power restoration time. The value ranges from 0.2 to 0.5, and is calibrated according to the urgency of power grid restoration.

[0085] The demand response weighting coefficient is a parameter used to measure the impact of demand response quotas on the total dispatch cost. The larger the value, the more it tends to reduce the amount of demand response execution. The value ranges from 0.5 to 0.8, and is set according to the power grid control cost budget.

[0086] Delay is a single closing moment in the candidate delay set for window avoidance. It is one of the core variables in constructing the optimization objective function and is used to calculate the total scheduling cost at different times.

[0087] The optimization objective function is a core indicator that comprehensively measures the timeliness of power restoration and the economy of regulation. The smaller the value, the better the corresponding delay scheduling scheme. Specifically, the optimization objective function is equal to the time cost term plus the demand response cost term. The time cost term is the delay multiplied by the preset time weight coefficient, and the demand response cost term is the minimum quota of desynchronized demand response multiplied by the demand response trigger flag, and then multiplied by the preset demand response weight coefficient.

[0088] The time cost item is an indicator that quantifies the time loss cost caused by the closing delay. The longer the delay and the larger the time weighting coefficient, the higher the value of this item.

[0089] The demand response cost item is an indicator that quantifies the control costs incurred by the execution of demand response. The higher the quota and the larger the response weight coefficient, the higher the value of this item.

[0090] The demand response trigger flag is a binary variable that indicates whether the demand response is executed. It takes the value of one or zero. When the value is one, the demand response is executed, and when the value is zero, it is not executed.

[0091] Minimization optimization is the process of finding the scheduling scheme that minimizes the objective function value by traversing the candidate delay set of the window avoidance through a specific algorithm. Specifically, the gradient descent method is used to traverse the candidate delay set of the window avoidance, calculate the objective function value of each delay under different demand response triggering flags, select the combination with the smallest function value, and the corresponding delay is the optimal execution time, and the corresponding flag is the optimal demand response triggering flag.

[0092] The optimal execution time is the optimal closing time obtained after minimization optimization. This time can balance the power restoration speed and control cost, and is the core time parameter that needs to be executed in the current period.

[0093] The optimal demand response trigger flag is the demand response execution decision obtained after minimization optimization, which determines whether a demand response quota needs to be issued in the current period.

[0094] The final execution quota is the actual amount of demand response that needs to be executed in the current period. It takes the value of zero or the minimum quota for desynchronized demand response, ensuring that control is executed only when necessary. Specifically, the final execution quota is equal to the optimal demand response trigger flag multiplied by the minimum quota for desynchronized demand response corresponding to the optimal execution time.

[0095] The scheduling strategy is a final scheduling instruction that includes a clear execution time and control amount. It can be directly issued to the power grid dispatching system for current execution to ensure that the decision is implemented.

[0096] like Figure 2 As shown, Figure 2 This diagram illustrates the overall architecture of an intelligent demand response dispatching system based on a power knowledge graph. The left side of the diagram shows distribution lines originating from a substation. Through sectionalizing switches, the load is divided into already-closed blocks (currently powered) and pending-close blocks (waiting for power restoration). Both blocks contain temperature-controlled load groups consisting of equipment such as air conditioners, heat pumps, and water heaters. The dispatch center on the right incorporates a power knowledge graph and a demand response dispatcher. On one hand, it acquires real-time line status through instantaneous current monitoring; on the other hand, it sends desynchronized demand response commands to the load groups based on calculation results. This aims to prevent excessive current surges when pending-close blocks are connected (i.e., avoiding the superposition of secondary synchronization start-up peaks and cold loads), ensuring the safe and stable operation of the power grid.

[0097] Example 2: A demand response intelligent scheduling system based on power knowledge graphs, applied to any of the aforementioned demand response intelligent scheduling methods based on power knowledge graphs, includes:

[0098] The power knowledge graph construction module defines feeder segment nodes in the power knowledge graph and associates time-locking distribution, thermal inertia parameters, and power outage duration as time-history semantic attributes to the feeder segment nodes.

[0099] The time-domain interval definition module analyzes the inherent characteristics of the previously closed block based on the time history semantic attributes to generate a time-locked thermal inertia resonance window spectrum. The time-locked thermal inertia resonance window spectrum defines the time-domain interval of the secondary synchronous start-up peak caused by the concentrated release of equipment time-locking and the thermal inertia coupling of the load.

[0100] The peak exposure function construction module constructs a peak exposure function to quantify the system instantaneous current peak value of the block to be closed under different delay access conditions. The system instantaneous current peak value is formed by superimposing the closing inrush current of the block to be closed onto the cold load recovery margin of the previously closed block and the secondary synchronous start peak.

[0101] The demand response quota calculation module optimizes the generation of execution time within the scheduling clock based on the peak exposure function, and prioritizes the selection of time periods that avoid the thermal inertia resonance window spectrum of the lock time. If it cannot be avoided or exceeds the limit, the minimum quota for desynchronized demand response is calculated based on the intensity of the secondary synchronization start peak.

[0102] The quota execution module outputs a scheduling strategy that includes the execution time and the minimum quota for desynchronized demand response, and executes it in the current period.

[0103] The embodiments of this example have been described above. However, this example is not limited to the specific implementation methods described above. The specific implementation methods described above are merely illustrative and not restrictive. Those skilled in the art can make many other forms based on the guidance of this example, and all of them are within the protection scope of this example.

Claims

1. A demand response intelligent scheduling method based on power knowledge graph, characterized in that, include: In the power knowledge graph, feeder segment nodes are defined, and time-locking distribution, thermal inertia parameters, and power outage duration are associated with feeder segment nodes as time-history semantic attributes. Based on the temporal semantic attributes, the inherent characteristics of the previously closed blocks are analyzed to generate the time-locked thermal inertia resonance window spectrum. The time-locked thermal inertia resonance window spectrum defines the time domain interval of the secondary synchronous start-up peak caused by the concentrated release of equipment lock-up and the thermal inertia coupling of the load. A peak exposure function is constructed to quantify the system instantaneous current peak value of the block to be closed under different delay access conditions. The system instantaneous current peak value is formed by superimposing the closing inrush current of the block to be closed onto the cold load recovery margin of the previously closed block and the secondary synchronous start peak. The peak exposure function includes: Obtain the cold load recovery amplitude coefficient and decay time constant of the previously closed block, calculate the product of the cold load recovery amplitude coefficient of the previously closed block and the exponential term with the natural constant as the base and the negative delay divided by the decay time constant as the exponent, and obtain the residual current of the previously closed block. Obtain the peak inrush current of the block to be closed, the current conversion factor of the previously closed block, and the preset peak sampling half-window width; The secondary synchronization start-up rate is integrated within the interval with the delay minus the peak sampling half-window width as the lower limit and the delay plus the peak sampling half-window width as the upper limit. The integral result is multiplied by the current conversion factor to obtain the secondary synchronization additive current. The residual current of the previously closed block, the peak inrush current of the block to be closed, and the secondary synchronous boost current are added together to obtain the peak exposure function, which characterizes the extreme value of the total system load under different delays. Based on the peak exposure function, the execution time is optimized within the scheduling clock. Priority is given to avoiding the time-locked thermal inertia resonance window spectrum. If it cannot be avoided or exceeds the limit, the minimum quota for desynchronized demand response is calculated for the intensity of the secondary synchronization start peak. The output includes the execution time and the scheduling strategy with the minimum quota for desynchronized demand response, and is executed in the current period.

2. The intelligent demand response scheduling method based on power knowledge graph according to claim 1, characterized in that, In the power knowledge graph, feeder segment nodes are defined, and time-locking distribution, thermal inertia parameters, and outage duration are associated with feeder segment nodes as time-history semantic attributes, including: Extract load characteristic data for the corresponding area in the power grid, and write the lockout time distribution, thermal inertia parameters, power outage duration and ambient temperature difference as basic time history semantic attributes into the feeder segment node; Based on thermal inertia parameters and power outage duration, a nodal post-power-on temperature function is constructed to characterize the exponential recovery of spatial temperature over time after power restoration. By differentiating the lock-time distribution with respect to time, the lock-time arrival density, which characterizes the probability of the device's anti-short cycle timer being reset at a specific moment, is calculated. Multiply the lockout density by the probability that the temperature function after the node is powered on is higher than the temperature control threshold, and calculate the secondary synchronous start rate, which represents the probability of instantaneous start-up of the equipment group after power-on due to the overlap of lockout release and heat demand. The secondary synchronization start rate is stored as a derived attribute in the feeder segment node.

3. The intelligent demand response scheduling method based on power knowledge graph according to claim 2, characterized in that, Based on the temporal semantic attributes, the inherent characteristics of previously closed blocks are analyzed to generate a time-locked thermal-inertial resonance window spectrum, including: Obtain the attributes of the previously closed blocks in the power knowledge graph, and calculate the center of the resonance window by finding the time point where the product of the lockout arrival density and the heat demand survival probability reaches its maximum value; the heat demand survival probability is composed of an exponential term minus an exponential term with the natural constant as the base and the sum of the power outage duration and the time variable divided by the negative value of the thermal inertia parameter. The square root of the sum of the variance of the lock-time distribution and the weighted squared terms of the thermal inertia parameter is used to obtain the resonant window width. Using the center of the resonance window as the midpoint and the width of the resonance window as half the width, a time-locked thermal inertia resonance window spectrum is constructed for the peak period of secondary synchronous start-up.

4. The intelligent demand response scheduling method based on power knowledge graph according to claim 3, characterized in that, Based on the peak exposure function, the execution time is optimally generated within the scheduling clock, prioritizing the selection of periods that avoid the time-locked thermal inertia resonance window spectrum. If this cannot be avoided or exceeds the limit, the minimum quota for desynchronized demand response is calculated based on the intensity of the secondary synchronization start-up peak, including: Remove time periods that fall into the time-locked thermal inertia resonance window spectrum from the preset set of scheduling clocks to generate a window avoidance candidate delay set composed of safety time slots; Obtain the preset system security threshold, and compare the peak exposure function calculated for the selected delay with the system security threshold; When the peak exposure function is greater than the system safety threshold, calculate the difference between the peak exposure function and the system safety threshold; Using the secondary synchronous addition current as the denominator, the difference is divided by the secondary synchronous addition current to obtain the minimum quota of desynchronized demand response, which characterizes the proportion of phase aggregation that needs to be reduced.

5. The intelligent demand response scheduling method based on power knowledge graph according to claim 4, characterized in that, The output includes the execution time and the scheduling policy for the minimum quota of desynchronized demand response, and is executed in the current period, including: Obtain the preset time weighting coefficient and demand response weighting coefficient; For each delay in the candidate delay set for window avoidance, an optimization objective function is constructed. The optimization objective function consists of a time cost term representing the speed of power restoration and a demand response cost term representing the cost of control. The time cost term is the product of the delay and the time weight coefficient, and the demand response cost term is the product of the minimum quota of desynchronized demand response, the demand response triggering flag, and the demand response weight coefficient. Minimize the objective function to determine the optimal execution time and the corresponding optimal demand response trigger flag that minimizes the total cost. Multiply the optimal demand response trigger flag by the minimum quota of the desynchronized demand response corresponding to the optimal execution time to obtain the final execution quota, and output the scheduling strategy consisting of the optimal execution time and the final execution quota.

6. A demand response intelligent scheduling system based on power knowledge graph, applied to the demand response intelligent scheduling method based on power knowledge graph as described in any one of claims 1-5, characterized in that, include: The power knowledge graph construction module defines feeder segment nodes in the power knowledge graph and associates time-locking distribution, thermal inertia parameters, and power outage duration as time-history semantic attributes to the feeder segment nodes. The time-domain interval definition module analyzes the inherent characteristics of the previously closed block based on the time history semantic attributes to generate a time-locked thermal inertia resonance window spectrum. The time-locked thermal inertia resonance window spectrum defines the time-domain interval of the secondary synchronous start-up peak caused by the concentrated release of equipment time-locking and the thermal inertia coupling of the load. The peak exposure function construction module constructs a peak exposure function to quantify the system instantaneous current peak value of the block to be closed under different delay access conditions. The system instantaneous current peak value is formed by superimposing the closing inrush current of the block to be closed onto the cold load recovery margin of the previously closed block and the secondary synchronous start peak. The demand response quota calculation module optimizes the generation of execution time within the scheduling clock based on the peak exposure function, and prioritizes the selection of time periods that avoid the thermal inertia resonance window spectrum of the lock time. If it cannot be avoided or exceeds the limit, the minimum quota for desynchronized demand response is calculated based on the intensity of the secondary synchronization start peak. The quota execution module outputs a scheduling strategy that includes the execution time and the minimum quota for desynchronized demand response, and executes it in the current period.

Images