Massive aggregator demand response power real-time allocation method and system based on incremental optimization and medium

Abstract

The present application relates to a kind of mass aggregator demand response power real-time distribution method, system and medium based on incremental optimization, the method includes: in preparation stage, based on the operating parameter of aggregated resource and power grid topology parameter, all possible transition point that causes optimal decision variable set to change is identified, mapping relationship from total regulating power demand to variable load aggregator set is established;In real-time distribution stage, the mapping relationship is used to determine the load aggregator that needs to be adjusted currently, and the optimization problem after dimension reduction is solved for load aggregator, so as to obtain globally optimal power distribution instruction.The present application solves the problem that the calculation efficiency is low and the optimal decision is difficult to be considered in the prior art when large-scale load aggregator participates in real-time frequency regulation, and the problem that the existing power distribution method is difficult to meet the requirement of second-level real-time while processing the uncertain response behavior of a large number of participants and obtaining globally optimal economic dispatch result.

Description

Technical Field

[0001] This invention relates to the field of smart grid technology, and in particular to a method, system and medium for real-time allocation of demand response power from massive aggregators based on incremental optimization. Background Technology

[0002] As global demand for carbon emission reduction grows, the penetration rate of renewable energy in power systems continues to increase. However, the inherent volatility and intermittency of renewable energy pose significant challenges to maintaining frequency stability and supply-demand balance in power systems. Demand response, as a key ancillary service, coordinates and manages a large number of distributed load aggregators through virtual power plants, enabling them to participate in electricity markets, including secondary frequency regulation, thus becoming a promising solution to address these challenges.

[0003] While load aggregators theoretically offer a solution, they still face multiple challenges in practical applications. First, most existing power allocation methods fail to adequately consider the stochasticity of demand response. The actual response behavior of residential loads often deviates from expectations, and this uncertainty introduces economic and operational risks to the accurate tracking of dispatch instructions by virtual power plants. Second, with the liberalization of market access policies, hundreds or even thousands of small residential load aggregators will participate in the market in the future. Faced with such a large number of participating entities, existing power allocation algorithms face bottlenecks in computational efficiency. Specifically, while traditional centralized optimization methods can find the optimal solution, their computational time increases dramatically with the number of aggregators, making it difficult to meet the second-level real-time requirements of secondary frequency regulation. Distributed or iterative methods, while offering better scalability, often sacrifice global optimality or fail to guarantee convergence time, leading to unfair cost allocation or inability to ensure real-time performance. Therefore, there is an urgent need for a power allocation method that can simultaneously consider economic optimality, response uncertainty, and large-scale computational efficiency to solve the problem of a large number of load aggregators participating in real-time frequency regulation. Summary of the Invention

[0004] This invention provides a method, system, and medium for real-time power allocation based on incremental optimization for demand response from massive load aggregators. It aims to solve the problems of low computational efficiency and difficulty in achieving optimal decision-making when large-scale load aggregators participate in real-time frequency regulation, as well as the difficulty of existing power allocation methods in handling the uncertain response behavior of a large number of participants and obtaining the globally optimal economic scheduling result while meeting the second-level real-time requirements.

[0005] To achieve the above objectives, this invention provides a method for real-time allocation of demand response power for massive aggregators based on incremental optimization, the method comprising the following steps:

[0006] Step S10: In the preparation phase, before real-time allocation of task execution, based on the operating parameters of aggregated resources and the grid topology parameters, all "turning points" that may cause changes in the optimal decision variable set are identified through analytical calculation, and a mapping relationship from total regulating power demand to variable load aggregate quotient set is established accordingly.

[0007] Step S20: In the real-time allocation phase, upon receiving the real-time total power demand instruction, the mapping relationship is used to determine the few load aggregators that need to be adjusted, and a dimension-reduced optimization problem is solved only for these load aggregators to obtain the globally optimal power allocation instruction.

[0008] As a further improvement of the present invention, the objective function of the power allocation problem is a quadratic function for minimizing the expected total operating cost, and the constraints of the optimization problem consist only of linear constraints, which include at least boundary constraints and power flow constraints for adjusting the upper and lower limits of power for each aggregated resource.

[0009] As a further improvement of the present invention, the calculation of the expected total operating cost takes into account the uncertainty of the aggregated resource response function. Its calculation method depends only on the expected value and covariance matrix of the random variable describing the uncertainty, and is not limited by the specific probability distribution form followed by the random variable.

[0010] As a further improvement of the present invention, the mapping relationship is generated by analytical calculation in the preparation stage, specifically by identifying and determining all inflection points of total regulating power demand that can cause the variable load aggregate quotient to change.

[0011] As a further improvement of the present invention, the identification of the inflection point includes: when the incremental cost rate of an aggregated resource that was originally in a non-variable load aggregator changes to be equal to the system incremental cost rate, the corresponding total regulating power demand value at this moment is determined as an inflection point.

[0012] As a further improvement of the present invention, the identification of the inflection point includes: when the regulating power of an aggregated resource in a variable load aggregater set reaches the upper or lower limit of its own capacity, the corresponding total regulating power demand value at this moment is determined as an inflection point.

[0013] As a further improvement of the present invention, the identification of the inflection point includes: when the incremental cost rate of an aggregated resource that was originally in a non-variable load aggregator changes to be equal to the incremental cost rate of an aggregated resource that is already in a variable load aggregator set, the corresponding total regulation power demand value at this moment is determined as an inflection point.

[0014] As a further improvement of the present invention, the process of identifying and determining all inflection points in the preparation phase is performed by parallel computing, wherein different types of inflection point computing tasks or inflection point computing tasks for different aggregate resource combinations are assigned to different computing threads or processor cores and performed simultaneously.

[0015] To achieve the above objectives, the present invention also proposes a real-time allocation system for demand response power of massive aggregators based on incremental optimization. The system includes a memory, a processor, and a real-time allocation program for demand response power of massive aggregators based on incremental optimization stored on the processor. When the processor runs the real-time allocation program for demand response power of massive aggregators based on incremental optimization, it executes the steps of the method described above.

[0016] To achieve the above objectives, the present invention also proposes a computer-readable storage medium storing a real-time allocation program for demand response power of massive aggregators based on incremental optimization. When the incrementally optimized real-time allocation program for demand response power of massive aggregators is run by a processor, the steps of the method described above are executed.

[0017] The beneficial effects of this invention, which is a real-time power allocation method, system, and medium for demand response from massive aggregators based on incremental optimization, are as follows:

[0018] This invention achieves rapid and optimal power allocation for large-scale demand response resources by decomposing the complex full-scale optimization problem into a pre-computation stage and a real-time dimensionality reduction solution stage. Compared with existing technologies, this invention has the following technical advantages: Through the innovative two-stage architecture of pre-computation in the pre-computation stage and real-time low-dimensional solution, the number of decision variables in the real-time optimization problem is reduced from massive to single digits, ensuring that power allocation commands can be calculated and issued within seconds, meeting the high real-time requirements of secondary frequency regulation. By employing analytical methods to pre-construct a complete "demand-decision" mapping relationship in the pre-computation stage, the real-time solution results are completely consistent with traditional centralized optimization methods, thus technically guaranteeing the global optimality and economic fairness of the allocation results and avoiding the suboptimal solution problem of distributed algorithms. This invention, leveraging the highly parallelizable pre-computation solution process, can efficiently handle ultra-large-scale systems containing hundreds or thousands of aggregators, solving the scalability bottleneck of traditional optimization methods in terms of computational power, and providing a feasible technical path for connecting massive distributed resources to the power grid and providing reliable services. This invention reduces the reliance on accurate prediction of aggregated resource response behavior by modeling the objective function based solely on the expectation and covariance of response uncertainty. This allows the method to maintain good robustness and practicality when faced with complex real-world response behaviors that are difficult to describe with a specific distribution.

[0019] In summary, this invention cleverly balances the contradictions between computational efficiency, decision optimality, and system scale, providing an efficient, reliable, and scalable solution to address the core technical bottleneck of large-scale distributed resources participating in real-time power grid scheduling. Attached Figure Description

[0020] Figure 1 This is a flowchart illustrating a preferred embodiment of the method for real-time allocation of demand response power for massive aggregators based on incremental optimization according to the present invention.

[0021] Figure 2 This is a schematic diagram of a preferred embodiment of the incremental optimization-based real-time power allocation method for demand response of massive aggregators according to the present invention.

[0022] Figure 3 This is a schematic diagram of the system structure of the incremental optimization-based real-time power allocation method for demand response of massive aggregators, a preferred embodiment of the present invention.

[0023] Figure 4 This is a schematic diagram of an IEEE-33 node system with a virtual power plant.

[0024] Figure 5 This is a block diagram of a secondary frequency modulation simulation control.

[0025] Figure 6 This is a schematic diagram of the net load fluctuation curve;

[0026] Figure 7 This is a schematic diagram comparing the system frequency changes of a virtual power plant using different algorithms to coordinate 3,000 load aggregators;

[0027] Figure 8 This is a structural diagram of a real-time power allocation system for demand response based on incremental optimization. Detailed Implementation

[0028] To make the objectives, technical solutions, and advantages of this invention clearer, the invention will be further described in detail below with reference to the accompanying drawings and embodiments.

[0029] This invention proposes a real-time power allocation method for massive aggregator demand response based on incremental optimization, enabling real-time power allocation for large-scale aggregator resources. To address the challenge of balancing computational efficiency and decision optimality in existing technologies when handling large-scale aggregators participating in secondary frequency regulation, this invention employs a two-stage approach. The core idea is to decompose the complex full-scale optimization problem into a preparation stage and a low-dimensional real-time allocation problem, thereby significantly improving real-time computational efficiency while ensuring decision optimality. Preparation stage: Offline pre-calculation establishes a mapping relationship between arbitrary total regulation power demand and the optimal set of variable load aggregators. Real-time allocation stage: Upon receiving a power command online, the mapping relationship is used to directly determine a small number of load aggregators requiring adjustment, and a dimensionality-reduced optimization problem is solved based solely on the corresponding optimal set of variable load aggregators, thus obtaining the globally optimal power allocation command within seconds. This invention, through its innovative "preparation stage dimensionality reduction -- online solution" architecture, significantly improves real-time decision-making speed while ensuring global optimality and robustly handles response uncertainties, providing an efficient, reliable, and scalable solution for massive aggregator resources participating in real-time grid dispatch.

[0030] Specifically, such as Figure 1 As shown, a preferred embodiment of the incremental optimization-based real-time allocation method for demand response power of massive aggregators according to the present invention includes the following steps:

[0031] Step S10: In the preparation phase, before real-time task allocation, based on the operating parameters of aggregated resources and the grid topology parameters, all "turning points" that may cause changes in the optimal decision variable set are identified through analytical calculation, and a mapping relationship from total regulating power demand to variable load aggregate quotient set is established accordingly.

[0032] In this embodiment, before the real-time allocation task is executed, a mapping relationship is pre-calculated and established based on the operating parameters of aggregated resources and the grid topology parameters to map the continuous total regulating power demand range to a pre-determined variable load aggregater set.

[0033] Step S20: In the real-time allocation phase, upon receiving the real-time total power demand instruction, the mapping relationship is used to determine the few load aggregators that need to be adjusted, and a dimension-reduced optimization problem is solved only for these load aggregators to obtain the globally optimal power allocation instruction.

[0034] In this embodiment, the real-time allocation stage involves receiving a real-time total regulation power demand signal, directly indexing and determining the corresponding variable load aggregation quotient set using a mapping relationship, and then constructing and solving a dimension-reduced optimization problem that only uses the regulation power of aggregated resources within the variable load aggregation quotient set as the variable to be solved, so as to quickly solve the real-time power allocation instructions of all aggregated resources.

[0035] Furthermore, in this embodiment, the objective function of the power allocation problem is a quadratic function for minimizing the expected total operating cost, and the constraints of the optimization problem consist only of linear constraints, which at least include boundary constraints for adjusting the upper and lower limits of power for each aggregated resource.

[0036] In this embodiment, the objective function of the power allocation optimization problem is constructed as a quadratic function to minimize the expected total operating cost that takes into account the power deviation penalty; at the same time, the constraints of the problem are simplified to a series of linear constraints, including at least boundary constraints on the upper and lower limits of power adjustment for each aggregated resource, thereby ensuring the convexity of the problem and the solution efficiency.

[0037] Furthermore, in this embodiment, the calculation of the expected total operating cost takes into account the uncertainty of the aggregated resource response function. Its calculation method depends only on the expected value and covariance matrix of the random variable describing the uncertainty, and is not limited by the specific probability distribution form followed by the random variable.

[0038] To address the uncertainty of load response, the calculation model for the expected total operating cost in this embodiment mathematically relies solely on the expected value and covariance matrix of the random variables describing the uncertainty. This approach gives the embodiment good universality, as it eliminates the need to know the specific probability distribution function followed by the random variables, thus enhancing the robustness of the algorithm.

[0039] Furthermore, in this embodiment, the mapping relationship is generated through analytical calculation during the preparation phase, specifically by identifying and determining all inflection points in the total regulating power demand corresponding to changes in the variable load aggregate quotient set.

[0040] The process of identifying inflection points in the preparation phase specifically includes the identification and calculation of the following three situations: First, when the incremental cost rate of a load aggregator that was originally in an invariant state changes to be equal to the system incremental cost rate; Second, when the regulating power of a load aggregator that is already in a variable state reaches the capacity boundary; Third, when the incremental cost rate of an invariant load aggregator catches up with and equals the incremental cost rate of a load aggregator that is already in a variable set.

[0041] Specifically, when the incremental cost rate of an aggregated resource that was originally in the set of invariable load aggregators changes to be equal to the system's incremental cost rate, the corresponding total regulating power demand value at this moment is determined as a turning point.

[0042] When the regulating power of an aggregated resource within a variable load aggregator set reaches the upper or lower limit of its own capacity, the corresponding total regulating power demand value at this moment is determined as a turning point.

[0043] When the incremental cost rate of an aggregated resource that was originally in the set of invariable load aggregators changes to be equal to the incremental cost rate of an aggregated resource that is already in the set of variable load aggregators, the corresponding total regulating power demand value at this moment is determined as a turning point.

[0044] This embodiment supports parallel computing in the inflection point identification process during the preparation phase. Since the calculations for different types of inflection points or for different combinations of aggregated resources are logically independent, they can be allocated to different computing threads or processor cores for simultaneous execution, thereby significantly shortening the time required for the preparation phase and enabling this method to efficiently handle ultra-large-scale power systems containing hundreds or thousands of aggregators.

[0045] The following combination Figures 2 to 3 The present invention provides a more detailed explanation of the real-time allocation method for demand response power of massive aggregators based on incremental optimization.

[0046] The specific interface method for information exchange between the system of this invention and the dispatch center and load aggregator is as follows: Figure 2 As shown. Before the start of a secondary frequency regulation dispatch lasting 15 minutes to 1 hour, the system receives parameters and grid topology information of the load aggregator within the dispatch cycle from the dispatch center via a data interface. This includes capacity, mileage price, expected response power, response correlation coefficient, the node access location of the load aggregator in the grid, power system topology, and power system parameters. Every 6 seconds, the system receives an AGC command from the dispatch center, specifically including the total load value for demand response. After solving the power adjustment scheme of each load aggregator using the real-time power allocation method proposed in this invention, the scheme is sent to each load aggregator via a data interface for execution according to its internal control method.

[0047] In a typical application scenario, a virtual power plant coordinates multiple load aggregators via a communication network to respond to AGC signals issued by the dispatch center. The virtual power plant acquires parameters for each load aggregator, such as rated regulating capacity, mileage pricing, and uncertainty distribution parameters. These parameters remain constant within the secondary frequency regulation interval. The actual power deviation after load regulation... Represented as:

[0048]

[0049] in Let i be the actual response power value of the i-th load aggregator. Let represent the number of load aggregators included in the virtual power plant. A virtual power plant coordinates N load aggregators to participate in the power regulation response to the AGC signal requirements from the dispatch center. Each load aggregator has a random deviation between the power value set for it by the virtual power plant and its actual response power. The actual response value is obtained by adding an uncertainty to the set response value. The demand response uncertainty model for each load aggregator is modeled as follows:

[0050]

[0051]

[0052] in, It is the set response value of the i-th load aggregator. It is the total regulating capacity of the i-th load aggregator. It is the adjustment ratio set by the i-th load aggregator. It is a random variable describing the uncertainty of the i-th load aggregator. The first and second moments are obtained in advance through historical parameters and are random variables with known distributions.

[0053] This invention first constructs a power allocation optimization problem with the objective of minimizing the expected total operating cost of a virtual power plant. The expected total operating cost consists of two parts: the penalty cost incurred due to the deviation between the actual response power and the commanded power, and the mileage cost paid for calling load aggregators to provide regulation services. The penalty cost is modeled as a quadratic function of the power deviation to reflect the severe economic consequences of large frequency deviations. The specific calculation method of the objective function is as follows:

[0054]

[0055]

[0056]

[0057]

[0058]

[0059] in, Indicates to Seeking expectations, This represents the total operating cost. This refers to the mileage price cost that virtual power plants need to pay to load aggregators. This indicates the fine imposed by the dispatch center on the virtual power plant due to the corresponding deviation. This represents the unit power per mileage price (in yuan / MW) for the i-th aggregator. This represents the total available adjustable capacity (in MW) of the i-th aggregator. This represents the proportion of the response power set by the virtual power plant for the i-th load aggregator to the total regulating capacity of the i-th load aggregator (a unitless decimal ranging from 0 to 1). It is the adjustment ratio set by the i-th load aggregator. It is a random variable describing the uncertainty of the i-th load aggregator. , , These represent the coefficients of the penalty function for the deviation of the dispatch center from the total demand response of the virtual power plant. It is the actual power deviation after load adjustment. It is the total demand response regulation power set by the dispatch center for the virtual power plant (multiple load aggregators coordinated by the virtual power plant meet this regulation requirement of the dispatch center).

[0060] The operation mode of the real-time power allocation method for demand response based on incremental optimization described in this invention is as follows: Figure 3 As shown, it comprises two phases: a preparation phase and a real-time allocation phase. The goal of the preparation phase is to identify all inflection points on the optimal solution curve. Each inflection point contains... and Two attribute values. For two adjacent turning points, the previous turning point is denoted as... The next turning point is denoted as Assuming . For the interval [ any It remains effective in the middle. Because the optimal power distribution will change with The inflection point changes continuously, so subsequent inflection points depend on the information of the previous inflection point. Therefore, once the first inflection point is known, all inflection points can be deduced sequentially. In the preparation phase, based on the information of the last inflection point and known parameters, the following optimization problem is solved to determine when the next inflection point occurs. The value of .

[0061]

[0062]

[0063]

[0064]

[0065]

[0066]

[0067]

[0068]

[0069]

[0070] in, This indicates the solution to achieve the total regulation power of the dispatch center corresponding to the next inflection point ( When the minimum is reached, the corresponding optimal set of variable load aggregators ( ), It is the set of variable load aggregators at the next unknown inflection point. and These represent the total demand power of the demand response at two adjacent inflection points. and These are the incremental rates of the i-th load aggregator and the system incremental rate at the next unknown inflection point, respectively. It is the i-th load aggregator. It is the set of all load aggregators. It is the set of immutable load aggregators at the previous known inflection point. , , These correspond to three scenarios for the next turning point. Candidate values. , , Of the three candidate values, the smallest one is selected as the next inflection point for the total demand response power. .calculate From the optimal power distribution and the incremental rate of load aggregator, we can derive... .

[0071] The incremental rate of the i-th load aggregator is calculated as follows:

[0072]

[0073]

[0074]

[0075] in, express Compared to The derivative of . express The variance.

[0076] Relationship between the incremental rate of the i-th load aggregator and the system incremental rate This is a necessary condition for finding the optimal solution to an optimization problem. The same place , , The three have the following relationship:

[0077]

[0078] in, It is the set of all load aggregators. It is the set of optimal variable load aggregators. It is the set of optimal immutable load aggregators.

[0079] During the real-time allocation phase, given and . The usage ratio of medium-load aggregators and The corresponding value at the most recent turning point is the same, because The increment rate is not the system increment rate. Therefore, in real-time power allocation, The usage ratio of medium-load aggregators needs to be based on To determine this. Based on the principle of equal increment rate, The increment rate of all load aggregators is the same, that is , At this point, the objective becomes minimizing the system's incremental rate. Therefore, the real-time allocation phase can be represented as...

[0080]

[0081]

[0082]

[0083]

[0084]

[0085] in, This indicates that solving for the system's incremental rate ( The minimum adjustment ratio set for each load aggregator ( ) ), It is a vector composed of the set adjustment ratios of all load aggregators. It is the lower limit of the node voltage. It is the node voltage at the end of the feeder. , , All variables are to be determined, while other parameters are known. The solution process for the real-time power distribution problem is as follows. Variables can be used The specific formula is as follows:

[0086]

[0087]

[0088] for The first element

[0089]

[0090]

[0091] in, Let be the uncertainty variance of the j-th load aggregator. Through the above simplification, Substitution In the constraints, the optimization problem in the real-time allocation phase is transformed into a problem only concerning... The linear programming problem is then solved. Finally, the solution to this linear programming problem is determined. and will Substitute these values ​​into the original problem to calculate the set usage ratio of all load aggregators.

[0092] The following examples 1 and 2 verify the real-time allocation method for demand response power of massive aggregators based on incremental optimization of the present invention.

[0093] Example 1:

[0094] In a power grid system, its topology and parameters are consistent with the IEEE-33 node system, with a base power of 1MW. See [link to IEEE-33 node numbering and virtual power plant resource location]. Figure 4 As shown.

[0095] Four nodes in the power grid are connected to a load aggregator. Nodes connected to the load aggregator ignore their default load settings in the IEEE-33 node system and only include controllable load resources. The parameters in the objective function... , , Assume that the actual response of each load aggregator follows a Beta distribution. Specific parameters for the load aggregators are shown in Table 1.

[0096] Table 1 Load Aggregator Parameters

[0097]

[0098] In this example, after the control device of the present invention is started, it first receives the load aggregator parameters and grid parameters through the data interface, and performs calculations for the preparation phase. The calculation results for all inflection points are shown in Table 2.

[0099] Table 2 Turning Points

[0100]

[0101] Example 2:

[0102] This example aims to verify the practical performance advantages of the method of the present invention in coordinating the participation of large-scale load aggregators in secondary frequency regulation. The power grid topology, node parameters, and objective function parameters in this example are as follows: , , All parameters are consistent with Example 1. The difference is that this example considers a virtual power plant coordinating 3000 load aggregators to provide secondary frequency regulation ancillary services through demand response. These load aggregators are randomly connected to four designated nodes (#4, #18, #22, #29). The regulating capacity, mileage price, and response uncertainty parameters of each aggregator are randomly generated within a reasonable range to simulate a large-scale, heterogeneous real-world scenario.

[0103] To simulate the dynamic frequency regulation process, this example constructs a closed-loop simulation system that includes primary frequency regulation, secondary frequency regulation, and power system inertial elements. Its system structure is as follows: Figure 5 As shown in the figure. The delay element in the simulation system represents the time consumed by the virtual power plant to calculate the optimal power allocation method after receiving the real-time AGC signal issued by the dispatch center. The simulation system parameters related to the secondary frequency regulation are shown in Table 3. The simulation test duration is 900 seconds, and the net load fluctuation curve that the system needs to cope with is shown in the figure. Figure 6 As shown.

[0104] Table 3 Parameters of the Secondary Frequency Modulation Simulation System

[0105]

[0106] To evaluate the practical effectiveness of the method of this invention, this example compares the system frequency control performance using the method of this invention with that using other centralized optimization methods (implemented by commercial solvers such as Gurobi or CPLEX). The key difference is that the computation time for power allocation decisions is introduced as a control delay into the closed-loop control of secondary frequency modulation. According to experimental tests, when handling 3000 load aggregators, the computation time for a single real-time power allocation using the method of this invention (TEIRP-PDA) is only 0.026 seconds. Using the traditional commercial solver Gurobi, the computation time for a single calculation is 21.677 seconds. Using the traditional commercial solver CPLEX, the computation time for a single calculation is 52.864 seconds.

[0107] Using the above calculation time as the control delay, the resulting system frequency change curve is as follows: Figure 7 As shown.

[0108] from Figure 7 Simulation results show that traditional optimization methods suffer from computational delays of tens of seconds, causing their adjustment commands to lag significantly behind system demands, resulting in a drastic and persistent shift in system frequency. In contrast, the method of this invention, with its sub-second computational speed, can respond to load fluctuations promptly and accurately, strictly controlling the system frequency within a smaller range. When coordinating 3000 load aggregators, the system frequency shift using this method was controlled within the range of [-0.0781 Hz, +0.0066 Hz]. Using the CPLEX solver, the system frequency shift range was expanded to [-0.0956 Hz, +0.0134 Hz]. Compared to the Gurobi solver, this method reduced the peak frequency shift by 18.31%.

[0109] This example fully demonstrates that the two-stage optimization method proposed in this invention can effectively overcome the "curse of dimensionality" caused by large-scale aggregated resource access by greatly reducing the real-time computing burden. The resulting technical effect is not only an improvement in computing speed, but also a direct transformation into a significant improvement in the frequency stability of the power system, demonstrating the great practical value and superiority of this invention in future large-scale demand response applications.

[0110] The beneficial effects of this invention's real-time allocation method for demand response power of massive aggregators based on incremental optimization are:

[0111] This invention achieves rapid and optimal power allocation for large-scale demand response resources by decomposing the complex full-scale optimization problem into a pre-computation stage and a real-time dimensionality reduction solution stage. Compared with existing technologies, this invention has the following technical advantages: Through the innovative two-stage architecture of pre-computation in the pre-computation stage and real-time low-dimensional solution, the number of decision variables in the real-time optimization problem is reduced from massive to single digits, ensuring that power allocation commands can be calculated and issued within seconds, meeting the high real-time requirements of secondary frequency regulation. By employing analytical methods to pre-construct a complete "demand-decision" mapping relationship in the pre-computation stage, the real-time solution results are completely consistent with traditional centralized optimization methods, thus technically guaranteeing the global optimality and economic fairness of the allocation results and avoiding the suboptimal solution problem of distributed algorithms. This invention, leveraging the highly parallelizable pre-computation solution process, can efficiently handle ultra-large-scale systems containing hundreds or thousands of aggregators, solving the scalability bottleneck of traditional optimization methods in terms of computational power, and providing a feasible technical path for connecting massive distributed resources to the power grid and providing reliable services. This invention reduces the reliance on accurate prediction of aggregated resource response behavior by modeling the objective function based solely on the expectation and covariance of response uncertainty. This allows the method to maintain good robustness and practicality when faced with complex real-world response behaviors that are difficult to describe with a specific distribution.

[0112] In summary, this invention cleverly balances the contradictions between computational efficiency, decision optimality, and system scale, providing an efficient, reliable, and scalable solution to address the core technical bottleneck of large-scale distributed resources participating in real-time power grid scheduling.

[0113] To achieve the above objectives, the present invention also proposes a real-time allocation system for demand response power of massive aggregators based on incremental optimization. The system includes a memory, a processor, and a real-time allocation program for demand response power of massive aggregators based on incremental optimization stored on the processor. When the processor runs the real-time allocation program for demand response power of massive aggregators based on incremental optimization, it executes the steps of the method described in the above embodiments.

[0114] Specifically, the incremental optimization-based real-time power allocation system for demand response of massive aggregators includes a memory for storing mapping relationships, a communication interface for receiving real-time total regulation power demand signals, and a processor connected to the communication interface. The mapping relationship maps a continuous range of total regulation power demand to a predetermined set of variable load aggregators. The processor is configured to, upon receiving the demand signal, construct and solve the power allocation problem as described above, and ultimately send power allocation instructions to all aggregators. In this embodiment, the incremental optimization-based real-time power allocation system for demand response of massive aggregators further includes an input interface for receiving data from the power grid dispatching agency and an output interface for issuing power allocation instructions to each aggregator.

[0115] like Figure 8 As shown, Figure 8 This paper explains the connection and deployment diagram of the data interface, power allocation method, and hardware in the incremental optimization-based real-time power allocation system for massive aggregator demand response. The data receiving port receives load aggregator resource parameters and grid parameter information every 15 minutes from the power grid dispatch center, as well as real-time AGC signals every 6 seconds. Upon receiving updated parameters, the high-performance computing center uses a preparation phase algorithm to solve for the intermediate results required for incremental calculation. Upon receiving the real-time AGC signal, the high-performance computing center calculates the allocation results using a real-time allocation phase algorithm and sets the power ratio of each aggregator to its total adjustable power through the decision sending port.

[0116] To achieve the above objectives, the present invention also proposes a computer-readable storage medium storing a real-time allocation program for demand response power of massive aggregators based on incremental optimization. When the incrementally optimized real-time allocation program for demand response power of massive aggregators is run by a processor, the processor is one or more central processing units (CPUs) capable of performing parallel computing.

[0117] The above description, in conjunction with specific preferred embodiments, provides a further detailed explanation of the present invention. It should not be construed that the specific implementation of the present invention is limited to these descriptions. For those skilled in the art, various simple deductions or substitutions can be made without departing from the concept of the present invention, and all such modifications and substitutions should be considered within the scope of protection of the present invention.

Claims

1. A method for real-time allocation of demand response power for massive aggregators based on incremental optimization, characterized in that, The method includes the following steps: Step S10: In the preparation phase, before real-time allocation of task execution, based on the operating parameters of aggregated resources and the grid topology parameters, all "turning points" that cause changes in the optimal decision variable set are identified through analytical calculation, and a mapping relationship from total regulating power demand to variable load aggregate quotient set is established accordingly. The identification of the inflection point includes: when the regulating power of an aggregated resource within a variable load aggregator set reaches the upper or lower limit of its own capacity, the corresponding total regulating power demand value at this moment is determined as an inflection point; or when the incremental cost rate of an aggregated resource that was originally within a non-variable load aggregator set changes to be equal to the incremental cost rate of an aggregated resource that is already within a variable load aggregator set, the corresponding total regulating power demand value at this moment is determined as an inflection point; or when the incremental cost rate of an aggregated resource that was originally within a non-variable load aggregator set changes to be equal to the incremental cost rate of an aggregated resource that is already within a variable load aggregator set, the corresponding total regulating power demand value at this moment is determined as an inflection point. Step S20: In the real-time allocation phase, when the real-time total adjustment power demand instruction is received, the mapping relationship is used to determine the load aggregators that need to be adjusted, and a dimension-reduced optimization problem with the expected total operating cost as the objective is solved only for these load aggregators, so as to obtain the globally optimal power allocation instruction. The objective function of the optimization problem is a quadratic function used to minimize the expected total operating cost. The constraints of the optimization problem consist only of linear constraints, which include at least boundary constraints and power flow constraints for adjusting the upper and lower limits of power for each aggregated resource. The calculation of the expected total operating cost takes into account the uncertainty of the aggregated resource response function. Its calculation method depends only on the expected value and covariance matrix of the random variable describing the uncertainty, and is not limited by the specific probability distribution form followed by the random variable.

2. The method for real-time allocation of demand response power for massive aggregators based on incremental optimization according to claim 1, characterized in that, The mapping relationship is generated through analytical calculation during the preparation phase, specifically by identifying and determining all inflection points in the total regulating power demand that can cause changes in the variable load aggregate quotient set.

3. The method for real-time allocation of demand response power for massive aggregators based on incremental optimization according to claim 2, characterized in that, The process of identifying and determining all inflection points during the preparation phase is performed through parallel computing, where computational tasks for different types of inflection points or for inflection points with different combinations of aggregate resources are assigned to different computing threads or processor cores and performed simultaneously.

4. A real-time power allocation system for demand response from massive aggregators based on incremental optimization, characterized in that, The system includes a memory, a processor, and a real-time power allocation program for demand response of massive aggregators based on incremental optimization stored on the processor. The real-time power allocation program for demand response of massive aggregators based on incremental optimization is executed by the processor to perform the steps of the method as described in any one of claims 1 to 3.

5. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores a real-time power allocation program for demand response of massive aggregators based on incremental optimization, which, when run by a processor, executes the steps of the method as described in any one of claims 1 to 3.

Images