Office building flexible load resource scheduling method, device, system and medium

Abstract

The application relates to the technical field of power load regulation, and provides an office building flexible load resource scheduling method, device, system and medium. The implementation scheme is as follows: a target load group is obtained by performing multi-layer clustering on load groups to which office building equipment belongs; a potential probability prediction value of the target load group at a target time is obtained by performing potential probability prediction on the target load group; an adjustable potential function is obtained by modeling the potential probability prediction value of the target load group at the target time; a relative matching probability constraint of the target load group at the target time is determined based on an integral result of the adjustable potential function; and a target scheduling scheme is obtained by solving a load resource regulation model based on the relative matching probability constraint of the target load group at the target time and a first constraint condition through an improved particle swarm algorithm and / or an improved genetic algorithm. The application can improve resource regulation accuracy and resource configuration efficiency.

Description

Technical Field

[0001] This invention relates to the field of power load regulation technology, and in particular to a method, device, system and medium for scheduling flexible load resources in office buildings. Background Technology

[0002] In energy management of office buildings or commercial parks, flexible load resources, as an important regulatory object connecting the energy consumption side and the power system, directly affect the overall energy regulation accuracy and configuration efficiency of office buildings or commercial parks due to their dynamic characteristics and adjustable potential.

[0003] However, existing flexible load scheduling methods mostly rely on static load models or empirical parameter settings, which fail to fully characterize the differences in load characteristics between different office building equipment groups in the spatiotemporal dimension. As a result, it is difficult to accurately reflect the adjustable potential of various flexible resources under specific operating conditions, and thus it is difficult to achieve high-precision resource matching and dynamic optimization.

[0004] Furthermore, existing flexible load scheduling ignores the uncertainty of the distribution of flexible resource potential and its coupling relationship with external environmental factors, which can easily lead to overly conservative or overly aggressive allocation of system resources, thereby wasting potential control capabilities and reducing resource utilization efficiency. Summary of the Invention

[0005] This invention provides a method, apparatus, system, and medium for scheduling flexible load resources in office buildings, which can solve at least one of the above-mentioned technical problems.

[0006] In a first aspect, embodiments of the present invention provide a method for scheduling flexible load resources in office buildings, comprising:

[0007] Multi-level clustering is performed on the load groups to which the equipment in each office building belongs to obtain the target load group;

[0008] Based on the parameter values ​​of the population topographic density index, local climate correction index, and equipment efficiency correction index at the target time, the potential probability prediction of the target load group is performed to obtain the potential probability prediction value of the target load group at the target time.

[0009] The potential probability prediction value of the target load group at the target time is modeled based on the normal distribution function to obtain the adjustable potential function of the target load group at the target time.

[0010] Based on the integral result of the adjustable potential function of the target load group at the target time, the relative matching probability constraint of the target load group at the target time is determined.

[0011] A load resource regulation model is constructed with the objectives of minimizing unit regulation cost, frequency deviation rate, average absolute error of regulation, comfort failure rate, and unit regulation delay response time.

[0012] Based on the relative matching probability constraint and the first constraint condition of the target load group at the target time, the load resource regulation model is solved by an improved particle swarm optimization algorithm and / or an improved genetic algorithm to obtain a target scheduling scheme, wherein the target scheduling scheme includes the target regulation power, target function value and target constraint result of the target load group at the target time.

[0013] Secondly, embodiments of the present invention provide a scheduling device for flexible load resources in office buildings, comprising:

[0014] The multi-level clustering module is used to perform multi-level clustering of the load groups to which the equipment in each office building belongs, so as to obtain the target load group.

[0015] The potential probability prediction module is used to perform potential probability prediction on the target load group based on the parameter values ​​of each indicator, such as population topographic density index, local climate correction index, and equipment efficiency correction index, at the target time, and obtain the potential probability prediction value of the target load group at the target time.

[0016] The modeling module is used to model the potential probability prediction value of the target load group at the target time based on the normal distribution function, so as to obtain the adjustable potential function of the target load group at the target time.

[0017] The relative matching probability constraint determination module is used to determine the relative matching probability constraint of the target load group at the target time based on the integral result of the adjustable potential function of the target load group at the target time.

[0018] The load resource regulation model construction module is used to construct a load resource regulation model with the objectives of minimizing unit regulation cost, frequency deviation rate, average absolute error of regulation, comfort failure rate, and unit regulation delay response time.

[0019] The solution module is used to solve the load resource regulation model based on the relative matching probability constraint and the first constraint condition of the target load group at the target time, by using an improved particle swarm optimization algorithm and / or an improved genetic algorithm, to obtain a target scheduling scheme. The target scheduling scheme includes the target regulation power, target function value and target constraint result of the target load group at the target time.

[0020] Thirdly, embodiments of the present invention also provide a scheduling system for flexible load resources in office buildings, comprising: at least one processor; and a memory communicatively connected to the at least one processor; wherein the memory stores instructions executable by the at least one processor, the instructions being executed by the at least one processor to enable the at least one processor to perform the method described in any one of the embodiments of the present invention.

[0021] Fourthly, embodiments of the present invention also provide a non-transitory computer-readable storage medium storing computer instructions, wherein the computer instructions are used to cause a computer to perform the method described in any one of the embodiments of the present invention.

[0022] The technical solution of this invention firstly involves multi-level clustering of the load groups belonging to various office building equipment to obtain target load groups that provide a refined data foundation for subsequent potential modeling. Then, potential probability prediction of the target load groups is performed using the parameter values ​​of population topography density, local climate correction, and equipment efficiency correction indicators at the target time. This quantifies the differences in adjustable potential among the flexible loads in the target load group at the target time, improving the accuracy of potential estimation. Next, the potential probability prediction values ​​are modeled using a normal distribution function to obtain the adjustable potential function of the target load group. Relative matching probability constraints are then determined based on the integral result of the adjustable potential function. This allows for a probabilistic expression of the matching degree between the adjustment potential value and the adjustment demand value at the target time, thereby achieving quantitative utilization of the potential control capability of the target load group. Finally, a load resource control model is constructed with the objectives of minimizing unit control cost, frequency deviation rate, average absolute error of control, and unit control delay response time. Based on the relative matching probability constraints and the first constraint condition, the control model is solved using an improved particle swarm optimization algorithm and / or an improved genetic algorithm. In this way, an adaptive balance can be achieved between global search and local convergence, thereby improving the control accuracy and configuration efficiency of flexible load resources in complex operating environments.

[0023] It should be understood that the description in this section is not intended to identify key or essential features of the embodiments of the present invention, nor is it intended to limit the scope of the invention. Other features of the invention will become readily apparent from the following description. Attached Figure Description

[0024] The accompanying drawings are provided for a better understanding of this solution and do not constitute a limitation of the invention. Wherein:

[0025] Figure 1 This is a flowchart of a method for scheduling flexible load resources in office buildings according to an embodiment of the present invention;

[0026] Figure 2This is a structural block diagram of a scheduling device for flexible load resources in office buildings according to an embodiment of the present invention;

[0027] Figure 3 This is a schematic block diagram of an electronic device used to implement the methods of embodiments of the present invention. Detailed Implementation

[0028] The following description, in conjunction with the accompanying drawings, illustrates exemplary embodiments of the present invention, including various details to aid understanding. These details should be considered merely exemplary. Therefore, those skilled in the art will recognize that various changes and modifications can be made to the embodiments described herein without departing from the scope of the invention. Similarly, for clarity and brevity, descriptions of well-known functions and structures are omitted in the following description.

[0029] Figure 1 This is a flowchart of a method for scheduling flexible load resources in office buildings according to an embodiment of the present invention.

[0030] like Figure 1 As shown, the scheduling method for the flexible load resources of this office building may include:

[0031] S110, perform multi-level clustering of the load groups to which the equipment in each office building belongs to obtain the target load group;

[0032] S120, based on the parameter values ​​of each indicator, including population topographic density, local climate correction, and equipment efficiency correction, at the target time, performs potential probability prediction on the target load group and obtains the potential probability prediction value of the target load group at the target time.

[0033] S130, Model the potential probability prediction value of the target load group at the target time based on the normal distribution function to obtain the adjustable potential function of the target load group at the target time;

[0034] S140, Based on the integral result of the adjustable potential function of the target load group at the target time, determine the relative matching probability constraint of the target load group at the target time.

[0035] S150 aims to minimize unit control cost, frequency deviation rate, average absolute error of control, comfort failure rate, and unit control delay response time, and constructs a load resource control model.

[0036] S160, based on the relative matching probability constraint of the target load group at the target time and the first constraint condition, the load resource regulation model is solved by an improved particle swarm algorithm and / or an improved genetic algorithm to obtain the target scheduling scheme, wherein the target scheduling scheme includes the target regulation power of the target load group at the target time, the target function value and the target constraint result.

[0037] For example, office building equipment, such as air conditioners, elevators, lighting fixtures, office computers, etc.

[0038] For example, the target load group refers to the set of load groups with similar regulatory attributes and potential distribution characteristics obtained by multi-level clustering of the load groups to which the equipment of each office building belongs.

[0039] For example, a target load cluster may include multiple load clusters. These could include highly sensitive control clusters, low-sensitive control clusters, and slashable load clusters, where each control cluster is considered a load cluster.

[0040] For example, the load resource regulation model refers to a multi-objective optimization model established with the goal of minimizing unit regulation cost, frequency deviation rate, average absolute error of regulation, and unit regulation delay response time, combined with relative matching probability constraints and multiple other constraints (including constraints on changes in the power supply in the regulated area, constraints on electricity price fluctuations, etc.), to comprehensively achieve the optimal scheduling of flexible load resources.

[0041] In this example, the load resource regulation model can be represented by the following functional expression:

[0042] ;

[0043] In the formula, The decision variable vector (i.e., the first candidate regulation power vector or the second candidate regulation power vector). The evaluation function value is represented by C, which is the unit control cost. Frequency deviation rate; To adjust the mean absolute error; For the rate of comfort not meeting the standard; Adjust the delay response time in units; For the penalty function; This is the penalty coefficient; , , , and These are the weighting coefficients for each normalized sub-objective (cost, frequency deviation rate, MAE, etc.); , , , and All are used as normalization terms, and they represent the theoretical maximum values ​​of each sub-objective function (derived through statistical analysis of historical data). Normalization is used to avoid the influence of dimensions.

[0044] For example, the decision variable vector of the model is the actual regulated active power of the target load group at the target time (a continuous variable), which can be expressed as a function:

[0045] ;

[0046] Where G is the total number of target load groups, and T is the total number of control periods (e.g., 1 hour / 5 minute interval, T=12). The decision variable vector X is arranged in the following order: first, it is divided according to load group 1 (for example, all T time periods of load group 1 first, then all T time periods of load group 2), and then sorted according to the control period t. The decision variable vector must satisfy the relative matching probability constraint.

[0047] In this example, the expression for calculating the unit control cost is as follows:

[0048] ;

[0049] In the formula, C represents the unit control cost, expressed in yuan / kW (kilowatt). Let G be the total cost of regulating target load group i at target time t, specifically covering user compensation costs (used to incentivize users to participate in load regulation), regulation equipment operation and maintenance costs (such as the operation and maintenance costs of load control terminals and communication equipment), and load equipment operation and maintenance costs (such as the loss compensation costs of adjustable loads); G represents the set of all target load groups. The actual controlled load amount of target load group i at target time t; If the values ​​are the same, the absolute value is taken to ensure the non-negativity of the cost calculation; This represents the demand potential for adjustment, and its value is affected by the predicted probability of the target load group at the target time. Electricity price at target time t Joint influence.

[0050] In this example, the formula for adjusting the mean absolute error is as follows:

[0051] ;

[0052] In the formula, T is the length of the control period (e.g., 1 hour, with a communication interval of 5 minutes, then T=12), representing the total number of time nodes that need to be sampled during the control process; G represents the set of all target load groups. The actual control power of target load group i at target time t; To regulate power demand (unit: kW).

[0053] In this example, if the load participates in cross-regional load dispatch, then This refers to the global dispatch issued by the dispatch center to the load in this region; if the load participates in the regional supply and demand balance dispatch, the demand can be controlled based on the droop control strategy. In the operating mechanism of power systems, frequency deviation is a key quantitative indicator characterizing the balance between active power supply and demand. Droop control, as a classic distributed control strategy, can dynamically adjust the real-time active power of flexible loads through a preset ratio of "frequency deviation - power regulation" (i.e., droop characteristics).

[0054] The control principle of droop control is as follows: In the formula, The real-time frequency and real-time active power of the power grid in the regulated area within time t; The frequency reference value is set to 50 Hz. This represents the real-time active power of the load when the regional power grid frequency is the reference value. The active power factor can be calculated using a set of real-time grid frequencies and real-time load active power. At this point... It can be determined by the real-time frequency deviation of the regional power grid. The calculation yields, that is: ; .

[0055] In this example, the expression for calculating the frequency deviation rate is as follows:

[0056] ;

[0057] In the formula, The number of time periods during which the regional power grid frequency exceeds the allowable range (e.g., 50Hz ± 0.2Hz) is adjusted; T is the length of the adjustment period.

[0058] In this example, the formula for calculating the comfort non-compliance rate is as follows:

[0059] ;

[0060] In the formula, This represents the number of periods when comfort levels exceed the standard; T represents the length of the control period. For example, if the office temperature exceeds the standard or the light intensity is below the preset range during a certain period (the specific setting can be adjusted according to actual needs), then that period is counted as a "comfort-exceeding period".

[0061] In this example, the expression for calculating the unit control delay response time is as follows:

[0062] ;

[0063] In the formula, The unit of control delay response time is seconds / kW; Let G be the time interval (in seconds) from receiving the scheduling instruction to actually reaching the instruction target value for target load group i. Take the maximum delay time among all load groups to reflect the worst-case response performance of the control system and ensure that the overall control timeliness meets the standard; G is the total number of target load groups; T is the total number of control periods. The actual controlled load of target load group i at target time t.

[0064] For example, the first constraint condition refers to the condition used to constrain the feasible control range of the target load group and the system operation safety boundary during the solution process of the load resource control model.

[0065] In this example, the first constraints include active power supply and demand balance constraints, reactive power supply and demand balance constraints, power consumption change constraints in the regulated area, and electricity price fluctuation constraints.

[0066] For example, the expression for the active power supply-demand balance constraint is as follows:

[0067] ;

[0068] In the formula, G represents the total number of target load groups; The baseline operating power (unit: kW) of the target load group i at time t before regulation reflects the inherent operating state of the load when it is not subject to regulation. The actual controlled load amount of target load group i at target time t; The surplus of electrical energy transmitted outward during the regulated region at time t (unit: kW); The power shortage introduced from the outside into the regulated region at time t (unit: kW); The total output of all generator units within the regulated area at time t (unit: kW), including distributed power sources and centralized power sources; The active power loss of the regulated area at time t (unit: kW) is mainly due to the active power loss caused by line resistance. The difference between active power production and consumption at time t (unit: kW); This means that the balance constraint must be satisfied at all control moments.

[0069] For example, the expression for the reactive power supply and demand balance constraint is as follows:

[0070]

[0071] In the formula, The baseline reactive power of the target load group i at time t before regulation; The reactive power change at time t after load group i is regulated is affected by the coupling effect of the load group's own characteristics (such as load power factor) and the active load regulation. The reactive power of the regulated region is input from the outside at time t. Let t be the reactive power transmitted outward from the regulated region. The reactive power loss in the regulated area at time t mainly comes from the inductive reactive power loss of the transmission line and the leakage reactance reactive power loss of the power transformer. The total reactive power supply of the reactive power compensation devices in the regulated area at time t includes the inductive reactive power output of the synchronous generator excitation regulation, the switching capacity of static var compensators and parallel capacitors, etc. Let t be the total reactive power output of all generator units in the regulated area; It is the difference between reactive power production and consumption in the regulated area at time t, used to buffer short-term reactive power fluctuations; This means that the balance constraint must be satisfied at all control moments.

[0072] For example, the expression for the constraint on the change in electrical quantity in the regulated area is as follows:

[0073] ;

[0074] In the formula, Let t be the active power loss of the regulated region before regulation. Let t be the change in active power loss after the regulation of the regulated region. It is the maximum active power loss allowed by the power grid; The maximum reactive power output of all reactive power compensation devices within the regulated area is determined by the physical characteristics (such as rated capacity) of the reactive power compensation devices. The maximum allowable active power deviation in the regulated area is derived based on system safety and stability constraints, equipment physical capacity limitations, and the allowable voltage deviation limit of the power system according to power quality standards (GB / T 15945-2008 "Power Quality - Frequency Deviation of Power Systems"). The maximum allowable reactive power deviation of the regulated area is derived from the allowable voltage deviation limit of the power system (such as GB / T 12325-2008 "Power Quality Supply Voltage Deviation").

[0075] For example, for office building areas with rooftop solar panels installed, solar power grid integration constraints must be met. The expression for the solar power grid integration constraint is as follows:

[0076] ;

[0077] In the formula, The actual controlled load amount of target load group i at target time t; The baseline operating power of target load group i at time t before regulation; The real-time output of rooftop photovoltaic power within the target load group i region at time t; The threshold for solar curtailment; This is the photovoltaic absorption coefficient. This constraint mandates that transferable load groups (such as air conditioning pre-cooling and non-core equipment operation) must... > By increasing electricity consumption, it is possible to achieve synergy between "grid regulation" and "local photovoltaic consumption".

[0078] For example, the expression for the electricity price fluctuation constraint is as follows:

[0079] ;

[0080] In the formula, This refers to the change in marginal electricity price of each node in the target load group before and after regulation. It represents the maximum permissible change in the marginal electricity price at the node before and after load regulation (unit: yuan / kWh). Its core function is to filter out solutions that cause grid transmission congestion due to centralized load regulation, which in turn leads to a sudden increase in local electricity prices. , These represent the sum of active power of load group i before and after load regulation, respectively. Let $i$ be the marginal electricity price of node i before and after load adjustment at time $t (unit: yuan / kWh); and $N$ be the set of all nodes within the regulated area. This represents the change in the value of electricity consumption after the load group; The maximum permissible change in marginal electricity value before and after load adjustment (unit: yuan / kWh). The marginal value of electricity consumption after regulation; The marginal value of electricity consumption before regulation; , These are the electricity consumption value functions before and after load group i is regulated, respectively.

[0081] In this example, nodes are mainly divided into generation-side nodes (power supply nodes), transmission / transformation-side nodes (hub nodes), and distribution / consumption-side nodes (load nodes). The marginal electricity price of a node is obtained based on real-time data collection, such as data collected from the electricity meters of the power source.

[0082] For example, the target scheduling scheme is a set of control results obtained by solving the load resource control model through an improved particle swarm optimization algorithm and / or an improved genetic algorithm. It includes the target control power of the target load group at the target time, the optimized objective function value, and the target constraint results.

[0083] In one implementation, based on the parameter values ​​of population topographic density, local climate correction, and equipment efficiency correction indicators at a target time, a potential probability prediction is performed on a target load group to obtain a predicted potential probability value for the target load group at the target time. This includes: determining the parameter value of the population topographic density indicator at the target time based on first load data at the target time, wherein the first load data includes at least the population density and real-time climate data of the area where the target load group is located; determining a first calculation result based on the ratio of the parameter value of the population topographic density indicator at the target time to the parameter value of the local climate correction indicator at the target time; performing a power operation on the first calculation result and the parameter value of the equipment efficiency correction indicator at the target time to obtain a second calculation result; and performing an exponential operation on the negative second calculation result with the natural constant as the base to obtain a predicted potential probability value for the target load group at the target time.

[0084] For example, the target time can be in the form of minutes, hours, or days.

[0085] For example, the first load data refers to a multi-dimensional data set used within the target time period to characterize the operating status of the target load group and external environmental characteristics (such as climate, weather, etc.).

[0086] For example, in an office building area, the population density collected at 14:00 on a weekday is 350 people / 1000m², the real-time temperature is 32℃, and the relative humidity is 65%. Simultaneously, the lighting and air conditioning loads in this area are recorded as 120 kilowatts (kW) and 280 kW, respectively. This data can be used as the primary load data for the target time to determine the parameter values ​​for the population topographic density index.

[0087] For example, the population topographic density index refers to a composite index used to characterize the population activity density and geographical distribution characteristics of the area where the office building is located.

[0088] In this example, the parameter value of the population topographic density index at the target time can be determined by the ratio of the number of people in the office building at the target time to the effective building area of ​​the floor where the people are located.

[0089] For example, the local climate correction index refers to an index that reflects the impact of climate conditions (including temperature, humidity, light intensity, wind speed, etc.) on the energy consumption and regulation potential of office equipment.

[0090] For example, the first calculation result is used to characterize the spatial variability of energy load under the combined effects of population distribution and climate factors. This calculation result can reflect the differences in energy consumption concentration in different regions under the same climatic conditions.

[0091] For example, the equipment efficiency correction index refers to a set of parameters that correct the adjustable potential based on the differences in the operating efficiency, aging degree and energy efficiency level of office building equipment, and is used to characterize the effective response capability of different equipment under the same external conditions.

[0092] For example, the second calculation result is used to comprehensively reflect the degree of influence of external environmental factors and equipment performance degradation on adjustable potential.

[0093] For example, the potential probability prediction value refers to the probability estimate of the potential adjustability of the target load group within the target time, which is used to reflect the distribution characteristics of the adjustment potential of flexible loads within the target time.

[0094] For example, the calculation process of the potential probability prediction value of the target load group at the target time can be expressed by a functional expression as follows: In the formula, denoted by , e represents the predicted potential probability of the region corresponding to the x-coordinate i and y-coordinate j of the centroid in the target load group at the target time d; represents the natural constant; t represents the parameter value of the population topographic density index at the target time. This indicates the parameter value of the local climate correction index at the target time; This indicates the parameter value of the equipment efficiency correction index at the target time.

[0095] In this example, the potential probability prediction value represents the probability that the target load group will achieve optimization within the target time period, and can be used as a basis for judging whether a load event has achieved optimization. In this example, the load potential confidence threshold is set to 95%, that is, when the potential probability prediction value of the target load group is not less than 95%, it is determined that the target load group has achieved optimization.

[0096] According to the above implementation method, the parameter value of the population topographic density index is determined by first load data based on the target time, and then calculated by combining it with local climate correction index and equipment efficiency correction index. This comprehensively reflects the impact of multiple dimensions such as population distribution, environmental climate, and equipment performance on the load adjustment potential. The resulting potential probability prediction value can not only accurately characterize the potential regulation capability of the target load group at the target time, but also effectively reduce the bias caused by single index modeling, improve the accuracy and adaptability of potential prediction, and provide more reliable data support for the optimization and solution of subsequent regulation models.

[0097] In one implementation, a model is created based on the predicted potential value of the target load group at the target time using a normal distribution function to obtain the adjustable potential function of the target load group at the target time. This includes: determining a mean function based on the predicted potential value of the target load group at the target time and a preset first parameter set; taking the actual adjustable potential of the target load group at the target time as a variable, and squaring the difference between the actual adjustable potential and the mean of the mean function at the target time to obtain a squared deviation; determining a third calculation result based on the product of the variance of the actual adjustable potential of the target load group at the target time and 2; determining a negative exponent based on the negative ratio of the squared deviation to the third calculation result; and determining the adjustable potential function of the target load group at the target time based on the product of the normalization coefficient and the negative exponent.

[0098] For example, the first parameter set is used to determine the set of statistical parameters for the distribution characteristics of adjustable potential. This parameter set includes at least the mean parameter, variance parameter, and standardized coefficients, as well as auxiliary parameters such as skewness and kurtosis coefficients to reflect load fluctuation characteristics. The parameter values ​​in the first parameter set can be obtained by relevant personnel through statistical analysis of historical load data.

[0099] For example, in the potential prediction of air conditioning load in office buildings, the mean parameter can represent the average adjustable power of the air conditioning load at the target time (e.g., 250kW), the variance parameter can represent the degree of fluctuation in load adjustment (specifically set according to demand), and the standardization coefficient is used to ensure that the overall integral of the function is 1.

[0100] For example, the adjustable potential function refers to a function used to describe the distribution law of the adjustable potential of the target load group at the target time, which can provide a functional expression for subsequent probabilistic constraints and optimization modeling.

[0101] In this example, the adjustable potential distribution characteristics of the flexible load group are described by a probability density function. Taking the normal distribution as an example (a Beta distribution or other distributions can also be chosen depending on the actual scenario), the adjustable potential function is expressed as: ;

[0102] In the formula, This represents the adjustable potential function of the target load group at the target time. It represents the actual adjustable potential of the target load group at the target time (unit: kWh / kilowatt-hour), which is limited by the physical characteristics of the equipment (load characteristics, regulation efficiency, and transmission losses, etc.); This represents the potential probability prediction value of the target load group in the aforementioned example at the target time, which reflects the impact of the geographical characteristics of the region to which the horizontal axis i and the vertical axis j belong at the target time on the potential. Indicates the first parameter set; This represents the mean function, determined by both geographical characteristics and uncertainty; for example, the more favorable the geographical conditions, the better. The higher, The higher the value, the greater the load forecasting error. The larger, The lower; Represents the normalization coefficient; This represents the standard deviation, which is also known as the actual adjustable potential. The degree of dispersion of the distribution can be determined by expert experience.

[0103] According to the above implementation method, by constructing a mean function based on the predicted potential value and a preset first parameter set, and combining it with the squared deviation and variance of the actual adjustable potential for function modeling, the adjustable potential distribution of the target load group at the target time can be accurately characterized in the form of a normal distribution. This method not only makes the potential function interpretable and continuous, but also quantifies the impact of uncertainty on the load's adjustable range, thereby improving the model's adaptability to different load operating characteristics and prediction accuracy, and providing a highly reliable input basis for subsequent probabilistic constraints and control optimization.

[0104] In one implementation, determining the relative matching probability constraint of the target load group at the target time based on the integral result of the adjustable potential function of the target load group at the target time includes: determining the maximum adjustment potential based on the predicted potential probability value of the target load group at the target time to obtain an integration upper limit; determining the minimum adjustment potential based on the electricity price at the target time and the predicted potential probability value of the target load group at the target time to obtain an integration lower limit; determining an integration interval based on the integration upper limit and the integration lower limit; integrating the adjustable potential function of the target load group at the target time within the integration interval to obtain the relative matching probability of the target load group at the target time; and using the product of the relative matching probability of the target load group at the target time and a preset priority coefficient being greater than or equal to a preset first threshold as the relative matching probability constraint of the target load group at the target time.

[0105] For example, the relative matching probability, i.e., the probability that "actual adjustment potential ≥ adjustment demand potential" (ranging from 0 to 1), is used to measure the feasibility of regulation. The confidence value is set to 95%, meaning that regulation is considered feasible if the relative matching probability is greater than or equal to 95%.

[0106] For example, the probability of matching potential and demand is calculated using the cumulative distribution function, and its core formula is:

[0107] ;

[0108] In the formula, Indicates the relative matching probability; This represents the demand potential for adjustment, and its value is affected by the predicted probability of the target load group at the target time. Electricity price at target time t Joint influence; Indicates maximum regulatory potential; This represents the actual adjustable potential of the target load group at the target time.

[0109] For example, relative matching probability constraints are used to limit the feasible solution space of flexible load resources in the regulation model, ensuring the feasibility and reliability of regulation behavior.

[0110] In this example, the relative matching probability constraint can be expressed as:

[0111] ;

[0112] In the formula, This represents the relative matching probability of target load group i at target time t; This represents the priority coefficient, where the priority coefficient of a highly sensitive regulatory cluster is lower than that of a low-sensitivity regulatory cluster, in order to limit the regulatory amplitude in key regions; the first threshold value is... .

[0113] According to the above implementation method, by determining the maximum and minimum regulation potential based on the predicted potential value and electricity price information, respectively, and then establishing an integral interval and integrating the adjustable potential function, the relative matching probability of the target load group at the target time can be obtained. Combining this probability with a priority coefficient and setting a threshold constraint allows for the dynamic evaluation of the matching degree between load regulation potential and regulation demand, while considering economic efficiency and feasibility. This not only enables a quantitative judgment of the feasibility of load regulation but also improves the accuracy of target response and the rationality of resource allocation when flexible loads participate in scheduling, thereby enhancing the stability and optimization effect of the overall regulation model.

[0114] In one implementation, an improved genetic algorithm is used to solve the load resource regulation model, including: selecting samples from a historical optimal solution database based on the similarity between current and historical regulation needs to obtain historical optimal solution samples; initializing the population of the historical optimal solution samples to obtain an initial population, wherein the initial population includes multiple individuals, each corresponding to a first regulation power vector; calculating the first regulation power vector corresponding to each individual in the initial population based on the objective function corresponding to the load resource regulation model to obtain the evaluation function value of each individual in the initial population; and performing a tournament selection operation based on the evaluation function values ​​of each individual in the initial population to select individuals from the initial population to the parent population. The process begins with the parent population. Based on an arithmetic crossover strategy, parent individuals are randomly paired, and crossover calculations are performed on the first regulatory power vector corresponding to each parent individual pair based on a random crossover factor, resulting in multiple offspring individuals. Gaussian mutation is then performed on each gene locus of these offspring individuals to obtain multiple updated offspring individuals. The parent population and the multiple updated offspring individuals are merged to obtain a merged population. Individuals are selected from the merged population based on evaluation function values ​​to form the next generation population. If the highest evaluation function value among all individuals meets the preset convergence condition, the iteration stops, and the first regulatory power vector corresponding to the individual with the highest evaluation function value is determined as the approximate optimal solution of the load resource regulation model.

[0115] For example, the fields of the historical optimal solution database include: First, regulation demand characteristics: time period (e.g., peak / valley / flat period), load type proportion (proportion of transferable / interruptible loads), meteorological parameters (temperature, sunshine), photovoltaic output forecast, and electricity price. Second, historical optimal solution samples: "regulation power of the load group at the corresponding time" under the corresponding demand, "function value (unit cost)", and "constraint results (e.g., whether the active power supply and demand balance constraint is met; if it is met, the constraint result is that the active power supply and demand balance constraint is met; otherwise, the constraint result is that the active power supply and demand balance constraint is not met)".

[0116] For example, in the historical operation data of an office building park, the database of historical best solutions constructed for the regulation demand during the "14:00-15:00 period on summer weekdays" includes the following fields: Regulation demand characteristics: Time period: peak period; Load type proportion: transferable load proportion 60%, interruptible load proportion 25%, load that can be reduced proportion 15%; Meteorological parameters: outdoor temperature 33℃, solar irradiance 780W / m²; Photovoltaic output forecast value: 180kW; Electricity price: 1.20 yuan / kWh.

[0117] The historical optimal solution sample corresponding to the regulation demand during the "14:00-15:00 period on summer weekdays" includes the regulation power of the load group at the corresponding time: air conditioning load is -60kW (pre-cooling reduction), lighting load is -15kW (reduction), and energy storage system discharge is 40kW (compensation); function value (unit cost): 0.92 yuan / kWh; constraint result: meets the active power supply and demand balance constraint.

[0118] By comparing the current regulation needs (e.g., the period from 11:00 to 12:00 on summer non-working days) with the aforementioned historical samples (e.g., a similarity of 85%), this sample can be selected as one of the historical optimal solution samples. Subsequently, using the regulation power vector [-60, -15, +40] recorded in the sample and its similarity weight as the core, the initial population is guided to be generated, thereby initializing a regulation scheme with high matching with the current operating conditions.

[0119] For example, the historical regulatory needs corresponding to the similarity scores of ≥80% and the top 5 scores after sorting each similarity score in descending order are selected, and the samples corresponding to these historical regulatory needs are taken as the historical optimal solution samples (denoted as X). hist,opt ).

[0120] For example, the similarity calculation formula is as follows:

[0121] ;

[0122] in, The similarity between current regulatory needs and historical regulatory needs; This is the first control power vector for the current control demand; is the control power vector of historical control demand; n is the number of dimensions of the control demand feature vector (e.g., n=5, corresponding to time period, temperature, photovoltaic output, electricity price, and load ratio); k is the k-th dimension in the demand feature vector, with a value range from 1 to n; The feature weights are calculated as follows (e.g., when photovoltaic output fluctuates greatly, w_photovoltaic = 0.3, which is higher than the 0.15 of other features). This is the value of the k-th feature in the current regulatory demand after standardization. This is the value of the k-th feature in the historical regulation demand after standardization. The weighted distance between current demand and historical demand in the multidimensional feature space reflects the degree of difference between the two. The weighted norm of the control power vector for historical control needs is used to normalize the distance values ​​and ensure that the similarity values ​​are within the range of [0, 1].

[0123] For example, the similarity weight of each historical best solution sample is calculated. The calculation formula is as follows:

[0124] ;in, Let be the similarity weight of the j-th historical best solution sample; Let be the similarity value between the historical control demand and the current control demand corresponding to the j-th historical optimal solution sample; The sum of the squared similarities of all historical best solution samples, that is, the sum of the similarities of all historical best solution samples selected (i.e., X). hist,opt , which is usually the sum of the squared similarity values ​​of all samples in the Top-5 similar samples.

[0125] For example, around X hist,opt Targeted generation of the initial population. Actual adjustable potential of the target load population at the target time. Since the values ​​are continuous, this example uses real-number encoding, with each individual corresponding to one first regulation power vector. The encoding length L = G×T. Where G is the number of target load groups; T is the number of time periods within the control cycle; The selection must satisfy the constraint of relative matching probability (i.e. ).

[0126] For example, the algorithm generates an initial population (population size set to N) based on adjusting potential constraints during runtime. GA The initial population consists of a core population and a diversity population.

[0127] First, the core population (accounting for 60%): [Regarding X] hist,opt The regulation power is subject to small perturbations of ±3% to ±8% (the perturbation amplitude is dynamically adjusted based on the current fluctuations in photovoltaic output; a smaller amplitude is used when the fluctuations are large) to ensure that the population remains close to the feasible region. The regulation power P of load group i in the core population... core,i(g,t) The calculation formula is as follows:

[0128]

[0129] In the formula, The inertial weights of the historical best solution samples; Let be the similarity weight of the j-th historical best solution sample; The regulation power of load group i is calculated for the historical best solution sample j; For the disturbance magnitude (e.g., if you want to introduce a 5% disturbance, then...) (A random number between -5% and 5%). If load group i has a large number of rooftop photovoltaic installations and the photovoltaic output fluctuates significantly, then reduce... To avoid excessive disturbances that could cause power to exceed photovoltaic absorption limits; Gaussian random numbers with a mean of 0 and a variance of 1 are used to generate positive and negative symmetrical random perturbations to ensure perturbation diversity. The weight of the random number can be preset by the user, and (To increase the algorithm's computation speed and reduce the number of iterations, increase the value.) To increase population diversity and reduce the population's dependence on historical best solutions, adjust the [adjustment / adjustment] setting. ); These are random numbers generated within the maximum adjustment potential range of load group i.

[0130] Second, diverse populations (40%): in X hist,opt Generate random solutions (such as X) within the “control power boundary range”. hist,opt The central air conditioning load regulation power is 120kW, and the boundary is 100~140kW, so the random solution is generated within this range to avoid population homogenization. Individuals that obviously violate the electricity price fluctuation constraint and regulation potential constraint are eliminated through "pre-verification" (such as solutions that have violated the constraint conditions (such as active power balance and frequency deviation constraints) 3 or more times in history), and invalid populations are directly screened out.

[0131] For example, the tournament selection operation involves randomly selecting K individuals (K is a positive integer greater than or equal to 1) from the initial population and calculating their evaluation function values. ,choose The smallest individual enters the parent pool; repeat the above process N. GA Secondly, ensure that the size of the parent population is consistent with the initial population, retain high-quality individuals, and maintain population diversity.

[0132] For example, individuals in the parent pool are randomly paired, and for each pair of individuals X a X b The calculation process for obtaining offspring individuals through processing can be represented by a function expression as follows:

[0133] ;

[0134] in, For offspring individuals; The crossover factor is randomized to ensure that offspring individuals fall within the feasible region of the decision variables; and All are paternal individuals; For offspring individuals.

[0135] For example, the Gaussian mutation operation, which involves performing Gaussian mutation on each gene locus of each offspring individual, can be represented by a function expression as follows: ;in, This represents the mutated gene locus value, which is the new regulatory power value obtained after Gaussian mutation. The regulatory power value of load group i at time t (i.e., a certain gene locus in offspring individual X); Use Gaussian random numbers; The variable time is positively correlated with the actual adjustment potential to avoid variations exceeding the feasible region. Let be the actual adjustment potential of load group i at time t.

[0136] For example, merging the parent population and the offspring population (i.e., a total of 2N) GA Individuals), select The smallest N GA Individuals form a new generation of population.

[0137] For example, the preset convergence condition is the change in the optimal evaluation function value of the population over 10 consecutive generations. If the value is less than the threshold (the specific value is set according to actual needs), or if the number of iterations reaches the maximum value, stop the iteration and output the first regulation power vector corresponding to the best individual in the population as the approximate optimal solution.

[0138] In one implementation, an improved particle swarm optimization (PSO) algorithm is used to solve a load resource regulation model, comprising: initializing a particle swarm, wherein the particle swarm includes multiple particles, each particle corresponding to a second regulation power vector; evaluating the position of each particle in the current iteration based on the objective function corresponding to the load resource regulation model, obtaining the individual extreme value of each particle and the global extreme value of the particle swarm in the current iteration; updating the inertia weight of the particle swarm based on a memory factor when the global extreme value of the particle swarm in the current iteration does not meet a preset convergence condition, obtaining an updated inertia weight; updating the first velocity of each particle based on the updated inertia weight, individual extreme value, and global extreme value, obtaining a second velocity of each particle; updating the position of each particle based on the second velocity of each particle, obtaining an updated particle swarm; constraining and correcting the position of each particle in the updated particle swarm based on a relative matching probability constraint, obtaining a corrected particle swarm, and using the corrected particle swarm as the particle swarm for the next iteration; stopping the iteration when the global extreme value meets a preset convergence condition, and determining the second regulation power vector corresponding to the global extreme value as the approximate optimal solution of the load resource regulation model.

[0139] For example, each particle corresponds to one second control power vector. (The encoding steps and dimension calculations of the improved genetic algorithm will not be repeated here.) Initial particle position Consistent with the initialization logic of the improved genetic algorithm, and subject to the relative matching probability constraint, the particle swarm size is set to N. PSO Initial velocity Randomly selected from Among them, the maximum speed , ) represents the maximum regulation potential of load group i at time t.

[0140] For example, the calculation process of the individual extrema of each particle and the global extrema of the particle swarm can be represented by the following expression:

[0141] ;

[0142] ;

[0143] In the formula, To evaluate the function value; Let K be the position of particle K at time t; Let be the location of the individual extreme value of particle k; Let be the optimal evaluation function value for particle k; This is an operation to retrieve the minimum value; The location of the global extremum; for The objective function value corresponding to the position is also the optimal evaluation function value of the particle swarm.

[0144] For example, the inertia weight is associated with the "function value of the historical best solution sample", and a memory factor is introduced to achieve dynamic adaptive adjustment. The calculation expression is as follows:

[0145] ;

[0146] In the formula, Let be the inertia weight for the t-th iteration (i.e., the updated inertia weight), with a value range of [0.4, 0.9]. Basic inertia weight (can be set to 0.6 to balance global / local optimization); For memory factor (0.2 < <0.5 reflects the degree of dependence on historical optimal solutions (e.g., when photovoltaic power output fluctuates greatly). Taking 0.5, relying on the stability of historical solutions; when electricity prices are stable. (Take 0.2 to enhance the exploratory nature). The function value of the historical optimal solution (such as the minimum unit control cost). Let be the function value in the t-th iteration.

[0147] For example, the weighting coefficients Substituting into the velocity update formula, and by introducing a decreasing inertia weight strategy, the ability to find local optimizations near historical optimal solutions is enhanced:

[0148] ;

[0149] in, Cognitive / social factor (value 1.5); The result is a random number, and its value is [0, 1]. The updated inertia weights; Let be the first velocity of particle k at the end of the t-th iteration; The location of the individual extreme value; Let be the position of particle k at the end of the t-th iteration / the second control power vector; The location of the global extremum.

[0150] Thus, if we denote the evaluation function value corresponding to the global optimal position gbest in the particle swarm at the t-th iteration as... ; and the optimal function value stored in the historical optimal solution database is denoted as So, when near (e.g., | | / <5%): Reduce (down to a minimum of 0.4) to enhance local optimization accuracy and uncover better solutions near historical optimal solutions (e.g., fine-tuning the control period for low-sensitivity loads); when Deviation (If the difference is >15%) Increase (up to 0.9) to enhance global exploration capabilities and avoid getting stuck in local optima (such as re-exploring the pre-cooling period of transferable loads).

[0151] For example, the expression for position update and constraint correction is: In the formula, For particle k in the first... The position at the end of the next iteration; Particle k in the first The second velocity at the end of the next iteration. If If the relative matching probability exceeds the constraint, it is truncated to the boundary of the feasible region.

[0152] For example, season So take the value at this time. The value is used as the updated position.

[0153] For example, the preset convergence condition is if the global extremum is... Change over 10 consecutive generations If the value is less than the preset threshold, or the number of iterations reaches the maximum value, the iteration stops, and the second control power vector corresponding to the global extremum is output as an approximate optimal solution.

[0154] In one implementation, if the project primarily focuses on day-ahead scheduling (e.g., generating load control schemes 24 hours in advance), the improved genetic algorithm is preferred. The improved genetic algorithm has strong global search capabilities, yielding more accurate optimized solutions and meeting cost and control accuracy requirements. If the project primarily focuses on intraday real-time scheduling (e.g., emergency frequency support, 5-minute load adjustments), the improved particle swarm optimization algorithm is preferred. The improved particle swarm optimization algorithm has a fast convergence speed (only half the time of the improved genetic algorithm), enabling rapid response to dynamic grid demands, and its constraint satisfaction rate meets engineering requirements. For further performance improvements, a hybrid algorithm based on the improved genetic algorithm and the improved particle swarm optimization algorithm can be used. Specifically, firstly, global search optimization is performed based on the improved genetic algorithm. This includes: selecting samples from the historical optimal solution database based on the similarity between current and historical control demands to obtain historical optimal solution samples; initializing the population using these historical optimal solution samples and performing multi-generation iterative operations, including tournament selection, arithmetic crossover, and Gaussian mutation steps; and outputting the individuals with the highest fitness as a global candidate solution set when the evaluation function value converges or the number of iterations reaches its upper limit. Second, local optimization is performed based on an improved particle swarm optimization algorithm. Specifically, this includes: using several individuals from the global candidate solution set as the initial particle swarm's position vector; initializing the particle swarm's velocity vector and calculating the individual and global extrema of each particle according to the objective function; dynamically adjusting particle velocity and position based on adaptive inertia weights, individual and global extrema to achieve local refinement of the global candidate solutions; and correcting the particle swarm position through relative matching probability constraints to ensure the results meet the constraints of the load resource regulation model. Third, after the local optimization stage, the globally optimal particle obtained from the final iteration and its corresponding second regulation power vector are summarized; and this second regulation power vector is determined as the approximate optimal solution of the hybrid algorithm for the load resource regulation model. This approach balances accuracy and efficiency.

[0155] In one implementation, multi-level clustering is performed on the load groups to which the equipment in each office building belongs to obtain multiple target load groups. This includes: performing a first-level clustering on the load groups to which the equipment in each office building belongs based on the type of equipment in each office building to obtain a first-dimensional load group; performing a second-level clustering on each load group in the first-dimensional load group using a clustering algorithm to obtain a second-dimensional load group; performing a third-level clustering on the transferable load group and the interruptible load group in the second-dimensional load group using a comfort sensitivity coefficient to obtain a first sensitive control cluster and a second sensitive control cluster; and determining the target load group based on the first sensitive control cluster, the second sensitive control cluster, and the reduceable load group in the second-dimensional load group.

[0156] For example, the first dimension load group refers to the load group divided based on the type of office building equipment (e.g., HVAC, lighting, office equipment).

[0157] For example, the first dimension of the load group includes three categories: transferable load group, interruptible load group, and load group that can be reduced.

[0158] In this example, the types of loads in the transferable load group are: those that need to run continuously until the task is completed (such as air conditioners that can be pre-cooled / delayed, and elevators that operate during off-peak hours), whose operation cannot be interrupted, but whose overall working hours can be brought forward or delayed. The types of loads in the interruptible load group are: those that are allowed to be interrupted during task execution (such as printers), but whose cumulative running time remains unchanged. The types of loads in the reduceable load group are: those that can reduce power consumption during peak hours and increase power consumption during off-peak hours (such as non-essential lighting).

[0159] For example, the second-dimensional load group refers to the load group obtained by further clustering based on the historical load data (such as peak power, operating cycle, and adjustment rate) of each office device, on the basis of the first-dimensional load group.

[0160] In this example, the second-dimensional load group includes three categories: the updated transferable load group, the interruptible load group, and the load group that can be reduced.

[0161] For example, the clustering algorithm can be a hierarchical clustering algorithm or a K-means (K-Means Clustering Algorithm) algorithm. This example illustrates the use of the K-Means clustering algorithm to cluster the first-dimensional load group. For the transferable load group, interruptible load group, and reduceable load group within the first-dimensional load group, feature matrices are constructed based on the historical load data (including peak power, operating cycle, and adjustment rate) of each office device. Based on these feature matrices, the K-Means clustering algorithm is used to cluster each type of load group. The clusters obtained through clustering show significant differences in operational characteristics, thus forming updated transferable load groups, updated interruptible load groups, and updated reduceable load groups, constituting the second-dimensional load group.

[0162] For example, the first sensitive control cluster refers to the clustering result in the third-level clustering, which is formed by a group of transferable loads with a low comfort sensitivity coefficient. For example, the load group formed by corridor lighting.

[0163] For example, the second sensitive control cluster refers to the clustering result formed by interruptible load groups with high comfort sensitivity coefficients in the third-level clustering. For example, the load group formed by office air conditioners.

[0164] It can be understood that the first sensitive regulatory cluster refers to the same meaning as the low-sensitivity regulatory cluster in the previous example; the second sensitive regulatory cluster refers to the same meaning as the high-sensitivity regulatory cluster in the previous example.

[0165] For example, the calculation process for third-level clustering using the comfort sensitivity coefficient can be expressed by the following formula: In the formula, This represents the weighting factor between load group i and its cluster center j in the first-dimensional load group; This represents the average load peak-to-valley difference within the cluster represented by cluster center j; This represents the peak-to-valley difference of the distributed flexible load i. This represents the average seasonal fluctuation component of the cluster represented by cluster center j. This represents the average seasonal fluctuation component of distributed load i. This represents the amplification factor of the weighting factor.

[0166] In the formula, This represents the weighting coefficient between load group i and its cluster center j in the second-dimensional load group after processing by the comfort sensitivity coefficient; This represents the weighting factor between load group i and its cluster center j in the first-dimensional load group; This represents the comfort sensitivity coefficient (determined by the properties of the area where the load is located, such as offices, corridors, data centers, etc.). This represents the comfort weighting coefficient; This indicates the comfort weighting factor.

[0167] In this example, the specific value of the comfort weight item needs to be determined by combining industry standards and the functional attributes of the area where the load is located, as well as GB50736-2012 "Code for Design of Heating, Ventilation and Air Conditioning of Civil Buildings". "Hard comfort indicators" for different functional areas are extracted as the anchoring basis for the weight item. For example, if the hard comfort indicators for an office / meeting room are "Temperature: 22~26℃ (Celsius) in summer, 18~22℃ (winter); Light intensity: ≥300 lux", then this area is a high-sensitivity area, and the basic "comfort weight item" can be between 0.7 and 0.9; if the hard comfort indicators for a corridor / stairwell are "Temperature: 18~30℃; Light intensity: ≥100 lux", the basic "comfort weight item" can be between 0.3 and 0.5. Furthermore, the specific value of the comfort weight coefficient can also be customized by the user.

[0168] In this example, a feature vector is constructed for each load sample in the transferable load group, and normalization is performed using methods such as minimum-maximum normalization. Each feature vector is weighted, and the weighted Euclidean distance is used as a similarity measure between samples. Based on this distance, a clustering algorithm (such as the weighted K-Means clustering algorithm) is applied to iteratively divide the samples into clusters, thereby obtaining the first sensitive control cluster. The third-level clustering of the interruptible load group is similar to that of the transferable load group, and will not be repeated here.

[0169] For example, a target load group is determined based on the reduceable load groups in the first sensitive control cluster, the second sensitive control cluster, and the second-dimensional load group; that is, all reduceable load groups in the first sensitive control cluster, the second sensitive control cluster, and the second-dimensional load group are considered as target load groups. In the aforementioned example, "target load group i" can be any one of the three target load groups in this example (the reduceable load groups in the first sensitive control cluster, the second sensitive control cluster, and the second-dimensional load group).

[0170] According to the above implementation method, through the above multi-layer clustering process, multi-dimensional features such as equipment type, operating characteristics and user comfort sensitivity can be fully utilized to realize hierarchical identification and differentiated modeling of flexible load resources, thereby improving the representativeness and clustering accuracy of the target load group, and providing higher accuracy and feasibility for subsequent potential prediction and regulation optimization.

[0171] Figure 2 This is a structural block diagram of an office building flexible load resource scheduling device according to an embodiment of the present invention.

[0172] like Figure 2 As shown, the scheduling device for the flexible load resources of the office building may include:

[0173] The multi-level clustering module is used to perform multi-level clustering of the load groups to which the equipment in each office building belongs, so as to obtain the target load group.

[0174] The potential probability prediction module is used to perform potential probability prediction on the target load group based on the parameter values ​​of each indicator, such as population topographic density index, local climate correction index, and equipment efficiency correction index, at the target time, and obtain the potential probability prediction value of the target load group at the target time.

[0175] The modeling module is used to model the potential probability prediction value of the target load group at the target time based on the normal distribution function, so as to obtain the adjustable potential function of the target load group at the target time.

[0176] The relative matching probability constraint determination module is used to determine the relative matching probability constraint of the target load group at the target time based on the integral result of the adjustable potential function of the target load group at the target time.

[0177] The load resource regulation model construction module is used to construct a load resource regulation model with the objectives of minimizing unit regulation cost, frequency deviation rate, average absolute error of regulation, comfort failure rate, and unit regulation delay response time.

[0178] The solution module is used to solve the load resource regulation model based on the relative matching probability constraint and the first constraint condition of the target load group at the target time, by using an improved particle swarm optimization algorithm and / or an improved genetic algorithm, to obtain a target scheduling scheme. The target scheduling scheme includes the target regulation power, target function value and target constraint result of the target load group at the target time.

[0179] In one embodiment, the potential probability prediction module includes:

[0180] The parameter value determination unit is used to determine the parameter value of the population topographic density index at the target time based on the first load data at the target time, wherein the first load data includes at least the population density and real-time climate data of the area where the target load group is located.

[0181] The first calculation result determination unit is used to determine the first calculation result based on the ratio of the parameter value of the population topographic density index at the target time to the parameter value of the local climate correction index at the target time.

[0182] The exponentiation unit is used to perform an exponentiation operation on the first calculation result and the parameter value of the equipment efficiency correction index at the target time to obtain a second calculation result;

[0183] The exponentiation unit is used to perform an exponential operation on the second calculation result (after taking a negative number) and the natural constant, with the natural constant as the base, to obtain the potential probability prediction value of the target load group at the target time.

[0184] In one implementation, the modeling module includes:

[0185] The mean function determination unit is used to determine the mean function based on the potential probability prediction value of the target load group at the target time and a preset first parameter set.

[0186] The squared deviation calculation unit is used to take the actual adjustable potential of the target load group at the target time as a variable, and to square the difference between the actual adjustable potential and the mean of the mean function at the target time to obtain the squared deviation.

[0187] The third calculation result calculation unit is used to determine the third calculation result based on the product of the variance of the actual adjustable potential of the target load group at the target time and 2.

[0188] The negative exponent determination unit is used to determine the negative exponent by taking the negative value of the ratio of the squared deviation to the third calculation result;

[0189] An adjustable potential function determination unit is used to determine the adjustable potential function of the target load group at the target time based on the product of the normalization coefficient and the negative exponent.

[0190] In one implementation, the relative matching probability constraint determination module includes:

[0191] The maximum adjustment potential determination unit is used to determine the maximum adjustment potential based on the predicted potential probability value of the target load group at the target time, so as to obtain the integral upper limit;

[0192] The minimum adjustment potential determination unit is used to determine the minimum adjustment potential based on the electricity price at the target time and the predicted potential probability of the target load group at the target time, so as to obtain the integral lower limit.

[0193] An integration interval determination unit is used to determine an integration interval based on the upper limit of integration and the lower limit of integration;

[0194] An integration unit is used to integrate the adjustable potential function of the target load group at the target time within the integration interval to obtain the relative matching probability of the target load group at the target time.

[0195] The relative matching probability constraint unit is used to ensure that the product of the relative matching probability of the target load group at the target time and a preset priority coefficient is greater than or equal to a preset first threshold, and to use this as the relative matching probability constraint of the target load group at the target time.

[0196] In one embodiment, the solving module includes:

[0197] The sample screening unit is used to screen samples from the historical optimal solution database based on the similarity between the current control demand and the historical control demand, and obtain historical optimal solution samples.

[0198] The population initialization unit is used to initialize the population of the historical optimal solution sample to obtain an initial population, wherein the initial population includes multiple individuals, and each individual corresponds to a first regulation power vector.

[0199] The evaluation function value calculation unit is used to calculate the first regulation power vector corresponding to each individual in the initial population based on the objective function corresponding to the load resource regulation model, so as to obtain the evaluation function value of each individual in the initial population.

[0200] The population iteration unit is used to perform a tournament selection operation based on the evaluation function values ​​of each individual in the initial population to select individuals from the initial population to the parent population, thus obtaining the parent population; based on an arithmetic crossover strategy, the parent individuals in the parent population are randomly paired, and crossover calculation is performed on the first regulatory power vector corresponding to the parent individuals based on a random crossover factor to obtain multiple offspring individuals; Gaussian mutation operation is performed on each gene locus of the multiple offspring individuals to obtain multiple updated offspring individuals; the parent population and the multiple updated offspring individuals are merged to obtain a merged population, and individuals are selected from the merged population based on the evaluation function values ​​to form the next generation population;

[0201] The population iterative convergence unit is used to stop the iteration when the highest evaluation function value among the evaluation function values ​​of each individual meets the preset convergence condition, and to determine the first regulation power vector corresponding to the individual with the highest evaluation function value as the approximate optimal solution of the load resource regulation model.

[0202] In one embodiment, the solving module includes:

[0203] A particle swarm initialization unit is used to initialize a particle swarm, wherein the particle swarm includes multiple particles, and each particle corresponds to a second control power vector.

[0204] The evaluation unit is used to evaluate the position of each particle in the particle swarm in the current iteration based on the objective function corresponding to the load resource regulation model, and to obtain the individual extreme value of each particle and the global extreme value of the particle swarm in the current iteration.

[0205] The particle swarm iteration unit is used to update the inertia weights of the particle swarm based on a memory factor when the global extremum of the particle swarm does not meet the preset convergence condition in the current iteration, thereby obtaining updated inertia weights; based on the updated inertia weights, the individual extremum, and the global extremum, update the first velocity of each particle to obtain the second velocity of each particle; based on the second velocity of each particle, update the position of each particle to obtain an updated particle swarm; and based on the relative matching probability constraint, constrain and correct the position of each particle in the updated particle swarm to obtain a corrected particle swarm, and use the corrected particle swarm as the particle swarm for the next iteration.

[0206] The particle swarm iteration convergence unit is used to stop the iteration when the global extreme value meets the preset convergence condition, and to determine the second regulation power vector corresponding to the global extreme value as the approximate optimal solution of the load resource regulation model.

[0207] In one implementation, the multi-layer clustering module includes:

[0208] The first-level clustering unit is used to perform first-level clustering of the load groups to which the equipment in each office building belongs based on the type of equipment in each office building, thus obtaining the first-dimensional load group.

[0209] The second-layer clustering unit is used to perform second-layer clustering on each load group in the first-dimensional load group using a clustering algorithm to obtain the second-dimensional load group.

[0210] The third-level clustering unit is used to perform third-level clustering on the transferable load group and the interruptible load group in the second-dimensional load group using the comfort sensitivity coefficient, to obtain the first sensitive control cluster and the second sensitive control cluster.

[0211] The target load group determination unit is used to determine the target load group based on the first sensitive control cluster, the second sensitive control cluster, and the slashable load group in the second dimension load group.

[0212] The specific functions and examples of each module and submodule of the system in this embodiment of the invention can be found in the relevant descriptions of the corresponding steps in the above method embodiments, and will not be repeated here.

[0213] The acquisition, storage, and application of user personal information involved in the technical solution of this invention all comply with the provisions of relevant laws and regulations and do not violate public order and good morals.

[0214] This invention also provides a scheduling system for flexible load resources in office buildings, comprising:

[0215] At least one processor; and a memory communicatively connected to said at least one processor;

[0216] The memory stores instructions that can be executed by the at least one processor, which, when executed by the at least one processor, enables the at least one processor to perform the method described in any one of the embodiments of the present invention.

[0217] The beneficial effects of the office building flexible load resource scheduling system of the present invention are equivalent to the beneficial effects of the office building flexible load resource scheduling method described above, and will not be repeated here.

[0218] This invention also provides a non-transitory computer-readable storage medium storing computer instructions, wherein the computer instructions are used to cause a computer to perform the method described in any one of the embodiments of this invention.

[0219] The beneficial effects of the storage medium of the present invention are equivalent to the beneficial effects of the above-described method for scheduling flexible load resources in office buildings, and will not be repeated here.

[0220] Figure 3 A schematic block diagram of an example electronic device 800 that can be used to implement embodiments of the present invention is shown. Electronic device 800 is intended to represent various forms of digital computers, such as laptop computers, desktop computers, workstations, personal digital assistants, servers, blade servers, mainframe computers, and other suitable computers. Electronic device 800 may also represent various forms of mobile devices, such as personal digital assistants, cellular phones, smartphones, wearable devices, and other similar computing devices. The components shown herein, their connections and relationships, and their functions are merely illustrative and are not intended to limit the implementation of the invention described and / or claimed herein.

[0221] like Figure 3 As shown, the electronic device 800 includes a computing unit 801, which can perform various appropriate actions and processes based on a computer program stored in a read-only memory (ROM) 802 or a computer program loaded from a storage unit 808 into a random access memory (RAM) 803. The RAM 803 may also store various programs and data required for the operation of the electronic device 800. The computing unit 801, ROM 802, and RAM 803 are interconnected via a bus 804. An input / output (I / O) interface 805 is also connected to the bus 804.

[0222] Multiple components in electronic device 800 are connected to I / O interface 805, including: input unit 806, such as keyboard, mouse, etc.; output unit 807, such as various types of displays, speakers, etc.; storage unit 808, such as disk, optical disk, etc.; and communication unit 809, such as network card, modem, wireless transceiver, etc. Communication unit 809 allows electronic device 800 to exchange information / data with other devices through computer networks such as the Internet and / or various telecommunications networks.

[0223] The computing unit 801 can be a variety of general-purpose and / or special-purpose processing components with processing and computing capabilities. Some examples of the computing unit 801 include, but are not limited to, a central processing unit (CPU), a graphics processing unit (GPU), various special-purpose artificial intelligence (AI) computing chips, various computing units running machine learning model algorithms, a digital signal processor (DSP), and any suitable processor, controller, microcontroller, etc. The computing unit 801 performs the various methods and processes described above, such as the scheduling method for flexible office building resources. For example, in some embodiments, the scheduling method for flexible office building resources can be implemented as a computer software program tangibly contained in a machine-readable medium, such as storage unit 808. In some embodiments, part or all of the computer program can be loaded and / or installed on the electronic device 800 via ROM 802 and / or communication unit 809. When the computer program is loaded into RAM 803 and executed by the computing unit 801, one or more steps of the scheduling method for flexible office building resources described above can be performed. Alternatively, in other embodiments, computing unit 801 may be configured by any other suitable means (e.g., by means of firmware) to perform a scheduling method for flexible load resources in office buildings.

[0224] Various embodiments of the systems and techniques described above herein can be implemented in digital electronic circuit systems, integrated circuit systems, field-programmable gate arrays (FPGAs), application-specific integrated circuits (ASICs), application-specific standard products (ASSPs), systems-on-a-chip (SoCs), payload-programmable logic devices (CPLDs), computer hardware, firmware, software, and / or combinations thereof. These various embodiments may include implementations in one or more computer programs that can be executed and / or interpreted on a programmable system including at least one programmable processor, which may be a dedicated or general-purpose programmable processor, capable of receiving data and instructions from a storage system, at least one input device, and at least one output device, and transmitting data and instructions to the storage system, the at least one input device, and the at least one output device.

[0225] The program code used to implement the methods of the present invention can be written in any combination of one or more programming languages. This program code can be provided to a processor or controller of a general-purpose computer, special-purpose computer, or other programmable data processing device, such that when executed by the processor or controller, the program code causes the functions / operations specified in the flowcharts and / or block diagrams to be implemented. The program code can be executed entirely on the machine, partially on the machine, as a standalone software package partially on the machine and partially on a remote machine, or entirely on a remote machine or server.

[0226] In the context of this invention, a machine-readable medium can be a tangible medium that may contain or store a program for use by or in conjunction with an instruction execution system, apparatus, or device. A machine-readable medium can be a machine-readable signal medium or a machine-readable storage medium. Machine-readable media can include, but are not limited to, electronic, magnetic, optical, electromagnetic, infrared, or semiconductor systems, apparatus, or devices, or any suitable combination of the foregoing. More specific examples of machine-readable storage media include electrical connections based on one or more wires, portable computer disks, hard disks, random access memory (RAM), read-only memory (ROM), erasable programmable read-only memory (EPROM or flash memory), optical fibers, portable compact disk read-only memory (CD-ROM), optical storage devices, magnetic storage devices, or any suitable combination of the foregoing.

[0227] To provide interaction with a user, the systems and techniques described herein can be implemented on a computer having: a display device for displaying information to the user (e.g., a CRT (cathode ray tube) or LCD (liquid crystal display) monitor); and a keyboard and pointing device (e.g., a mouse or trackball) through which the user provides input to the computer. Other types of devices can also be used to provide interaction with the user; for example, feedback provided to the user can be any form of sensory feedback (e.g., visual feedback, auditory feedback, or tactile feedback); and input from the user can be received in any form (including sound input, voice input, or tactile input).

[0228] The systems and technologies described herein can be implemented in computing systems that include backend components (e.g., as a data server), or computing systems that include middleware components (e.g., an application server), or computing systems that include frontend components (e.g., a user computer with a graphical user interface or web browser through which a user can interact with implementations of the systems and technologies described herein), or any combination of such backend, middleware, or frontend components. The components of the system can be interconnected via digital data communication of any form or medium (e.g., a communication network). Examples of communication networks include local area networks (LANs), wide area networks (WANs), and the Internet.

[0229] Computer systems can include clients and servers. Clients and servers are generally located far apart and typically interact via communication networks. Client-server relationships are created by computer programs running on the respective computers and having a client-server relationship with each other. Servers can be cloud servers, servers in distributed systems, or servers incorporating blockchain technology.

[0230] It should be understood that the various forms of processes shown above can be used to reorder, add, or delete steps. For example, the steps described in this invention can be executed in parallel, sequentially, or in different orders, as long as the desired result of the technical solution disclosed in this invention can be achieved, and this is not limited herein.

[0231] The specific embodiments described above do not constitute a limitation on the scope of protection of this invention. Those skilled in the art should understand that various modifications, combinations, sub-combinations, and substitutions can be made according to design requirements and other factors. Any modifications, equivalent substitutions, and improvements made within the principles of this invention should be included within the scope of protection of this invention.

Claims

1. A method for scheduling flexible load resources in office buildings, characterized in that, include: Multi-level clustering is performed on the load groups to which the equipment in each office building belongs to obtain the target load group; Based on the parameter values ​​of the population topographic density index, local climate correction index, and equipment efficiency correction index at the target time, the potential probability prediction of the target load group is performed to obtain the potential probability prediction value of the target load group at the target time. The potential probability prediction value of the target load group at the target time is modeled based on the normal distribution function to obtain the adjustable potential function of the target load group at the target time. Based on the integral result of the adjustable potential function of the target load group at the target time, the relative matching probability constraint of the target load group at the target time is determined. A load resource regulation model is constructed with the objectives of minimizing unit regulation cost, frequency deviation rate, average absolute error of regulation, comfort failure rate, and unit regulation delay response time. Based on the relative matching probability constraint of the target load group at the target time and the first constraint condition, the load resource regulation model is solved by an improved particle swarm optimization algorithm and / or an improved genetic algorithm to obtain a target scheduling scheme. The target scheduling scheme includes the target regulation power, target function value and target constraint result of the target load group at the target time. The first constraint condition includes active power supply and demand balance constraint, reactive power supply and demand balance constraint, power change constraint in the regulated area, and electricity price fluctuation constraint. The method involves predicting the potential probability of the target load group based on the parameter values ​​of the population topographic density index, local climate correction index, and equipment efficiency correction index at the target time, resulting in the predicted potential probability value of the target load group at the target time, including: Based on the first load data at the target time, the parameter value of the population topographic density index at the target time is determined, wherein the first load data includes at least the population density and real-time climate data of the area where the target load group is located; The first calculation result is determined based on the ratio of the parameter value of the population topographic density index at the target time to the parameter value of the local climate correction index at the target time. The first calculation result and the parameter value of the equipment efficiency correction index at the target time are raised to a power to obtain the second calculation result; Using the natural constant as the base, the second calculation result after taking a negative number is exponentially operated with the natural constant to obtain the potential probability prediction value of the target load group at the target time; The step of solving the load resource regulation model using an improved genetic algorithm includes: Based on the similarity between current regulatory needs and historical regulatory needs, samples are selected from the historical optimal solution database to obtain historical optimal solution samples; The historical optimal solution sample is initialized to obtain an initial population, wherein the initial population includes multiple individuals, and each individual corresponds to a first regulation power vector; Based on the objective function corresponding to the load resource regulation model, the first regulation power vector corresponding to each individual in the initial population is calculated to obtain the evaluation function value of each individual in the initial population. Based on the evaluation function values ​​of each individual in the initial population, a tournament selection operation is performed to select individuals from the initial population to the parent population, thus obtaining the parent population. Using an arithmetic crossover strategy, parent individuals in the parent population are randomly paired, and crossover calculations are performed on the first regulatory power vectors corresponding to each parent individual pair based on a random crossover factor, resulting in multiple offspring individuals. Gaussian mutation is performed on each gene locus of the multiple offspring individuals, resulting in multiple updated offspring individuals. The parent population and the multiple updated offspring individuals are merged to obtain a merged population, and individuals are selected from the merged population based on the evaluation function values ​​to form the next generation population. If the highest evaluation function value among all individuals meets the preset convergence condition, the iteration stops, and the first regulation power vector corresponding to the individual with the highest evaluation function value is determined as the approximate optimal solution of the load resource regulation model. The step of solving the load resource regulation model using an improved particle swarm optimization algorithm includes: Initialize a particle swarm, wherein the particle swarm comprises multiple particles, each particle corresponding to a second control power vector; Based on the objective function corresponding to the load resource regulation model, the position of each particle in the particle swarm in this iteration is evaluated to obtain the individual extreme value of each particle and the global extreme value of the particle swarm in this iteration. If the global extremum of the particle swarm does not meet the preset convergence condition in this iteration, the inertia weight of the particle swarm is updated based on the memory factor to obtain the updated inertia weight; based on the updated inertia weight, the individual extremum, and the global extremum, the first velocity of each particle is updated to obtain the second velocity of each particle; based on the second velocity of each particle, the position of each particle is updated to obtain the updated particle swarm; based on the relative matching probability constraint, the position of each particle in the updated particle swarm is constrained and corrected to obtain the corrected particle swarm, and the corrected particle swarm is used as the particle swarm for the next iteration; The iteration stops when the global extreme value meets the preset convergence condition, and the second regulation power vector corresponding to the global extreme value is determined as the approximate optimal solution of the load resource regulation model.

2. The method according to claim 1, characterized in that, The process of modeling the potential probability prediction value of the target load group at the target time based on the normal distribution function to obtain the adjustable potential function of the target load group at the target time includes: Based on the predicted potential probability of the target load group at the target time and the preset first parameter set, the mean function is determined. The actual adjustable potential of the target load group at the target time is used as a variable. The squared difference between the actual adjustable potential and the mean of the mean function at the target time is taken to obtain the squared deviation. The third calculation result is determined based on the product of the variance of the actual adjustable potential of the target load group at the target time and 2. The negative exponent is determined by taking the negative value of the ratio of the squared deviation to the third calculation result; The adjustable potential function of the target load group at the target time is determined based on the product of the normalization coefficient and the negative exponent.

3. The method according to claim 1, characterized in that, The determination of the relative matching probability constraint of the target load group at the target time based on the integral result of the adjustable potential function of the target load group at the target time includes: Based on the predicted potential probability of the target load group at the target time, the maximum adjustment potential is determined to obtain the integral upper limit. Based on the electricity price at the target time and the predicted potential of the target load group at the target time, the minimum adjustment potential is determined to obtain the integral lower limit. The integration interval is determined based on the upper and lower limits of integration. The adjustable potential function of the target load group at the target time is integrated over the integration interval to obtain the relative matching probability of the target load group at the target time. The product of the relative matching probability of the target load group at the target time and a preset priority coefficient is greater than or equal to a preset first threshold, which is used as the constraint on the relative matching probability of the target load group at the target time.

4. The method according to claim 1, characterized in that, The process of performing multi-level clustering on the load groups to which the equipment in each office building belongs to obtain the target load group includes: Based on the type of equipment in each office building, the first-level clustering of the load groups to which the equipment in each office building belongs is performed to obtain the first-dimensional load groups. The second-dimensional load group is obtained by performing a second-level clustering on each load group in the first-dimensional load group using a clustering algorithm. Using the comfort sensitivity coefficient, the transferable load group and the interruptible load group in the second dimension load group are respectively clustered in the third layer to obtain the first sensitive control cluster and the second sensitive control cluster. The target load group is determined based on the first sensitive control cluster, the second sensitive control cluster, and the slashable load group in the second dimension load group.

5. A scheduling device for flexible load resources in office buildings, characterized in that, include: The multi-level clustering module is used to perform multi-level clustering of the load groups to which the equipment in each office building belongs, so as to obtain the target load group. The potential probability prediction module is used to perform potential probability prediction on the target load group based on the parameter values ​​of each indicator, such as population topographic density index, local climate correction index, and equipment efficiency correction index, at the target time, and obtain the potential probability prediction value of the target load group at the target time. The modeling module is used to model the potential probability prediction value of the target load group at the target time based on the normal distribution function, so as to obtain the adjustable potential function of the target load group at the target time. The relative matching probability constraint determination module is used to determine the relative matching probability constraint of the target load group at the target time based on the integral result of the adjustable potential function of the target load group at the target time. The load resource regulation model construction module is used to construct a load resource regulation model with the objectives of minimizing unit regulation cost, frequency deviation rate, average absolute error of regulation, comfort failure rate, and unit regulation delay response time. The solution module is used to solve the load resource regulation model based on the relative matching probability constraint and the first constraint condition of the target load group at the target time, by using an improved particle swarm optimization algorithm and / or an improved genetic algorithm to obtain the target scheduling scheme. The target scheduling scheme includes the target regulation power, the target function value and the target constraint result of the target load group at the target time. The first constraint condition includes active power supply and demand balance constraint, reactive power supply and demand balance constraint, power change constraint of the regulated area and electricity price fluctuation constraint. The potential probability prediction module includes: The parameter value determination unit is used to determine the parameter value of the population topographic density index at the target time based on the first load data at the target time, wherein the first load data includes at least the population density and real-time climate data of the area where the target load group is located. The first calculation result determination unit is used to determine the first calculation result based on the ratio of the parameter value of the population topographic density index at the target time to the parameter value of the local climate correction index at the target time. The exponentiation unit is used to perform an exponentiation operation on the first calculation result and the parameter value of the equipment efficiency correction index at the target time to obtain a second calculation result; The exponentiation unit is used to perform an exponential operation on the second calculation result (after taking a negative number) and the natural constant, with the natural constant as the base, to obtain the potential probability prediction value of the target load group at the target time. The solution module includes: The sample screening unit is used to screen samples from the historical optimal solution database based on the similarity between the current control demand and the historical control demand, and obtain historical optimal solution samples. The population initialization unit is used to initialize the population of the historical optimal solution sample to obtain an initial population, wherein the initial population includes multiple individuals, and each individual corresponds to a first regulation power vector. The evaluation function value calculation unit is used to calculate the first regulation power vector corresponding to each individual in the initial population based on the objective function corresponding to the load resource regulation model, so as to obtain the evaluation function value of each individual in the initial population. The population iteration unit is used to perform a tournament selection operation based on the evaluation function values ​​of each individual in the initial population to select individuals from the initial population to the parent population, thus obtaining the parent population; based on an arithmetic crossover strategy, the parent individuals in the parent population are randomly paired, and crossover calculation is performed on the first regulatory power vector corresponding to the parent individuals based on a random crossover factor to obtain multiple offspring individuals; Gaussian mutation operation is performed on each gene locus of the multiple offspring individuals to obtain multiple updated offspring individuals; the parent population and the multiple updated offspring individuals are merged to obtain a merged population, and individuals are selected from the merged population based on the evaluation function values ​​to form the next generation population; The population iterative convergence unit is used to stop the iteration when the highest evaluation function value among the evaluation function values ​​of each individual meets the preset convergence condition, and to determine the first regulation power vector corresponding to the individual with the highest evaluation function value as the approximate optimal solution of the load resource regulation model. The solution module includes: A particle swarm initialization unit is used to initialize a particle swarm, wherein the particle swarm includes multiple particles, and each particle corresponds to a second control power vector. The evaluation unit is used to evaluate the position of each particle in the particle swarm in the current iteration based on the objective function corresponding to the load resource regulation model, and to obtain the individual extreme value of each particle and the global extreme value of the particle swarm in the current iteration. The particle swarm iteration unit is used to update the inertia weights of the particle swarm based on a memory factor when the global extremum of the particle swarm does not meet the preset convergence condition in the current iteration, thereby obtaining updated inertia weights; based on the updated inertia weights, the individual extremum, and the global extremum, update the first velocity of each particle to obtain the second velocity of each particle; based on the second velocity of each particle, update the position of each particle to obtain an updated particle swarm; and based on the relative matching probability constraint, constrain and correct the position of each particle in the updated particle swarm to obtain a corrected particle swarm, and use the corrected particle swarm as the particle swarm for the next iteration. The particle swarm iteration convergence unit is used to stop the iteration when the global extreme value meets the preset convergence condition, and to determine the second regulation power vector corresponding to the global extreme value as the approximate optimal solution of the load resource regulation model.

6. A scheduling system for flexible load resources in office buildings, characterized in that, include: At least one processor; and a memory communicatively connected to the at least one processor; The memory stores instructions that can be executed by the at least one processor, which, when executed by the at least one processor, enables the at least one processor to perform the method of any one of claims 1-4.

7. A non-transitory computer-readable storage medium storing computer instructions, characterized in that, The computer instructions are used to cause the computer to perform the method according to any one of claims 1-4.

Images