Multi-agent flexible resource distributed collaborative optimization privacy protection method

Abstract

The invention relates to the technical field of cooperative computing, and discloses a multi-agent flexible resource distributed cooperative optimization privacy protection method, which comprises the following steps: generating a first computing parameter through parameter initialization; generating a mask value of the current iteration; constructing a first preset function according to the local target function and the first calculation parameter, and calculating a local target parameter corresponding to the minimum value of the first preset function; calculating a first intermediate parameter according to the first calculation parameter, the local target parameter corresponding to the minimum value of the first preset function and the mask value of the current iteration, and sending the first intermediate parameter to the scheduling end; and determining whether the next round of iteration is carried out based on the updated first calculation parameter or not according to whether the scheduling end returns the updated first calculation parameter or not. The method has the advantages of high collaborative optimization precision, low extra calculation and communication overhead and strong privacy protection effect, can protect the security of privacy data on an aggregator side, and also can protect sensitive data on a power dispatching side.

Description

Technical Field

[0001] The present invention relates to the field of collaborative computing technology, and in particular to a privacy protection method for distributed collaborative optimization of multi-agent flexible resources. Background Art

[0002] With the development of technologies such as cloud computing, the Internet of Things, and big data, more and more application scenarios require the processing and analysis of massive amounts of data to improve service quality and efficiency. However, in these application scenarios, there are often multiple data sources or participants, who may compete or cooperate with each other and have high privacy protection requirements for their own data. For example, in a smart grid, power suppliers, power demanders, power dispatchers, and other parties need to jointly optimize power resource allocation and consumption strategies, but they are unwilling to directly expose their sensitive information (such as load curves, price information, etc.) to other parties. Therefore, in these scenarios, the traditional centralized plaintext data processing method cannot meet the privacy protection requirements.

[0003] In a collaborative optimization scenario consisting of N demand-side flexible resource aggregators (such as virtual power plants, smart buildings, electric vehicle operators, etc.) and 1 power dispatcher. Each demand-side flexible resource aggregator has its own electricity demand and cost function, as well as a certain degree of flexibility, and can adjust its electricity consumption according to the price signal or incentive mechanism given by the power dispatcher. The goal of the power dispatcher is to minimize the total system cost, such as minimizing the power generation cost and the demand-side incentive cost. In order to achieve this goal, the power dispatcher needs to know the electricity demand and flexibility of each demand-side flexible resource aggregator, as well as their response function to the price signal or incentive mechanism. However, this information is the privacy data of each demand-side flexible resource aggregator, and they are unwilling to disclose it directly to the power dispatcher or other parties. Based on the relevant distributed optimization and privacy protection methods proposed here, the main ideas and methods can be divided into the following two types: First, using the differential privacy method, before the aggregator sends the original data to the power dispatcher, random noise is added to the original data for perturbation. In this way, when appropriate differential privacy parameters are selected, the accuracy of the collaborative optimization results can be roughly guaranteed; second, using semi-homomorphic encryption methods such as Paillier, the aggregator sends the homomorphically encrypted data to the power dispatcher for homomorphic computing aggregation. The power dispatcher finally decrypts the results of the aggregated calculation to obtain the final collaborative optimization calculation results.

[0004] However, the distributed optimization scenario of flexible resources on the demand side has the characteristics of many aggregators, large amount of collaborative optimization interaction data, high accuracy requirements for collaborative optimization results, and high timeliness requirements for optimization. The aforementioned method 1, which uses differential privacy, has a small amount of additional computation and does not require the introduction of additional communication overhead, but due to the introduction of random noise, the accuracy of the collaborative optimization results has decreased to a certain extent (the more thorough the data privacy protection on the aggregator side, the more the accuracy of collaborative optimization decreases); the aforementioned method 2, which uses a semi-homomorphic encryption algorithm, has high computational accuracy and strong security, but introduces additional homomorphic encryption and decryption computational overhead to both the aggregator side and the power dispatching side, and the ciphertext expansion effect is obvious, which is not completely suitable for collaborative optimization scenarios with high timeliness requirements. Summary of the invention

[0005] In view of this, an embodiment of the present invention provides a privacy protection method for distributed collaborative optimization of multi-agent flexible resources to solve the technical problem that existing distributed optimization solutions cannot meet the requirements of timeliness and accuracy at the same time.

[0006] The technical solution proposed by the present invention is as follows:

[0007] A first aspect of an embodiment of the present invention provides a privacy protection method for distributed collaborative optimization of multi-agent flexible resources, which is applied to the flexible resource end of the aggregator side, including: generating a first calculation parameter through parameter initialization; generating a mask value for this round of iteration, wherein the sum of the mask value generated for this round of iteration and the mask value generated by the flexible resource end of other aggregators is 0; constructing a first preset function according to the local objective function and the first calculation parameter, and calculating the local objective parameter corresponding to the minimum value of the first preset function; calculating a first intermediate parameter according to the first calculation parameter, the local objective parameter corresponding to the minimum value of the first preset function, and the mask value of this round of iteration, and sending the first intermediate parameter to the scheduling end of the power coordination scheduling side; determining whether to perform the next round of iteration based on the updated first calculation parameter according to whether the scheduling end of the power coordination scheduling side returns the updated first calculation parameter, wherein each round of iteration process includes the process from generating the mask value for this round of iteration to sending the first intermediate parameter to the scheduling end of the power coordination scheduling side; outputting the local objective parameter corresponding to the minimum value of the first preset function calculated in the last round of iteration.

[0008] Optionally, generating the mask value for this round of iteration includes: generating a first random vector; receiving a second random vector sent by a pre-matched flexible resource end on the aggregator side, wherein the flexible resource end on the aggregator side pre-matched by each flexible resource end on the aggregator side is different; and generating the mask value for this round of iteration based on the difference between the first random vector and the second random vector.

[0009] Optionally, before generating the first calculation parameter through parameter initialization, it also includes: establishing a secure channel with the dispatching end of the power coordination dispatching side and the flexible resource end of other aggregator sides.

[0010] The second aspect of an embodiment of the present invention provides a distributed collaborative optimization method based on flexible resources, which is applied to the dispatching end on the power coordination dispatching side, including: generating Lagrange multipliers and mask matrices through parameter initialization, and converting the first global constraint parameter and the second global constraint parameter according to the mask matrix to obtain the first conversion parameter and the second conversion parameter; receiving the first intermediate parameter sent by the flexible resource end of all aggregators, and adding all the first intermediate parameters to obtain the second intermediate parameter; updating the Lagrange multiplier according to the Lagrange multiplier, the second intermediate parameter, the second conversion parameter and the preset gradient parameter; judging whether the preset iteration termination condition is met, if the preset iteration termination condition is met, the calculation is terminated, if the preset iteration termination condition is not met, the first calculation parameter is updated according to the first conversion parameter and the updated Lagrange multiplier, the updated first calculation parameter is sent to the flexible resource end of the aggregator side, and the next round of iteration is entered, wherein each round of iteration process includes the process of receiving the first intermediate parameter sent by the flexible resource end of all aggregators to sending the updated first calculation parameter to the flexible resource end of the aggregator side.

[0011] Optionally, the determining whether the preset iteration termination condition is met includes: determining whether the number of iteration rounds reaches a preset number of iterations or whether the variation range of the Lagrange multiplier reaches a preset termination value; if the number of iteration rounds reaches the preset number of iterations or the variation range of the Lagrange multiplier reaches the preset termination value, the preset iteration termination condition is met; if the number of iteration rounds does not reach the preset number of iterations and the variation range of the Lagrange multiplier does not reach the preset termination value, the preset iteration termination condition is not met.

[0012] Optionally, before receiving the first intermediate parameters sent by all flexible resource terminals on the aggregator side, the method further includes: establishing a secure channel with the flexible resource terminal on the aggregator side.

[0013] The third aspect of an embodiment of the present invention provides a joint collaborative computing system based on privacy computing, including: a dispatching end on the power coordination and dispatching side and several flexible resource ends on the aggregator side, each of the flexible resource ends on the aggregator side is communicatively connected to the dispatching end on the power coordination and dispatching side, and any two flexible resource ends on the aggregator side are communicatively connected to each other; the flexible resource end on the aggregator side generates a first calculation parameter through parameter initialization; the dispatching end on the power coordination and dispatching side generates a Lagrange multiplier and a mask matrix through parameter initialization, and converts the first global constraint parameter and the second global constraint parameter according to the mask matrix to obtain a first conversion parameter and a second conversion parameter; the flexible resource end on the aggregator side generates a mask value for this round of iteration, wherein the sum of the mask value generated for this round of iteration and the mask value generated by the flexible resource end on other aggregator sides is 0, a first preset function is constructed according to the local objective function and the first calculation parameter, and the local objective parameter corresponding to the minimum value of the first preset function is calculated, and the first intermediate parameter is calculated according to the first calculation parameter, the local objective parameter corresponding to the minimum value of the first preset function and the mask value of this round of iteration. number, and send the first intermediate parameter to the dispatching end of the power coordination dispatching side; the dispatching end of the power coordination dispatching side receives the first intermediate parameters sent by all flexible resource ends on the aggregator side, adds all the first intermediate parameters to obtain the second intermediate parameter, and updates the Lagrange multiplier according to the Lagrange multiplier, the second intermediate parameter, the second conversion parameter and the preset gradient parameter; the dispatching end of the power coordination dispatching side determines whether the preset iteration termination condition is met. If the preset iteration termination condition is met, the dispatching end of the power coordination dispatching side ends the calculation, and the flexible resource end on the aggregator side outputs the local target parameter corresponding to the minimum value of the first preset function calculated in the last round of iteration process. If the preset iteration termination condition is not met, the dispatching end of the power coordination dispatching side updates the first calculation parameter according to the first conversion parameter and the updated Lagrange multiplier, sends the updated first calculation parameter to the flexible resource end on the aggregator side, and enters the next round of iteration, wherein each round of iteration process includes the process in which the flexible resource end on the aggregator side generates the mask value of this round of iteration to the dispatching end on the power coordination dispatching side and sends the updated first calculation parameter to the flexible resource end on the aggregator side.

[0014] A fourth aspect of an embodiment of the present invention provides a flexible resource end on the aggregator side, characterized in that it includes: a first initialization module, which is used to generate a first calculation parameter through parameter initialization; a mask generation module, which is used to generate a mask value for this round of iteration, wherein the sum of the mask value generated for this round of iteration and the mask value generated by the flexible resource end on the other aggregator side is 0; a minimum value calculation module, which is used to construct a first preset function according to the local objective function and the first calculation parameter, and calculate the local objective parameter corresponding to the minimum value of the first preset function; a first parameter calculation module, which is used to calculate a first intermediate parameter according to the first calculation parameter, the local objective parameter corresponding to the minimum value of the first preset function and the mask value of this round of iteration, and send the first intermediate parameter to the dispatching end on the power coordination dispatching side; a first iteration module, which is used to determine whether to perform the next round of iteration based on the updated first calculation parameter according to whether the dispatching end on the power coordination dispatching side returns the updated first calculation parameter, wherein each round of iteration process includes the process from generating the mask value for this round of iteration to sending the first intermediate parameter to the dispatching end on the power coordination dispatching side; an output module, which is used to output the local objective parameter corresponding to the minimum value of the first preset function calculated in the last round of iteration.

[0015] According to a fifth aspect of an embodiment of the present invention, there is provided a dispatching end on the power coordination dispatching side, comprising: a second initialization module for generating a Lagrange multiplier and a mask matrix through parameter initialization, and converting a first global constraint parameter and a second global constraint parameter according to the mask matrix to obtain a first conversion parameter and a second conversion parameter; a second parameter calculation module for receiving a first intermediate parameter sent by all flexible resource ends on the aggregator side, and adding all the first intermediate parameters to obtain a second intermediate parameter; a multiplier update module for updating the Lagrange multiplier according to the Lagrange multiplier, the second intermediate parameter, the second conversion parameter and a preset gradient parameter; a second iteration module for determining whether a preset iteration termination condition is met, and if the preset iteration termination condition is met, the calculation is terminated; if the preset iteration termination condition is not met, the first calculation parameter is updated according to the first conversion parameter and the updated Lagrange multiplier, and the updated first calculation parameter is sent to the flexible resource end on the aggregator side, and the next round of iteration is entered, wherein each round of iteration process includes a process from receiving the first intermediate parameter sent by all flexible resource ends on the aggregator side to sending the updated first calculation parameter to the flexible resource end on the aggregator side.

[0016] A sixth aspect of an embodiment of the present invention provides a computer-readable storage medium, which stores computer instructions, and the computer instructions are used to enable the computer to execute the distributed collaborative optimization method based on flexible resources as described in any one of the first aspect or the second aspect of the embodiment of the present invention.

[0017] It can be seen from the above technical solutions that the embodiments of the present invention have the following advantages:

[0018] The embodiments of the present invention provide a distributed collaborative optimization method, system and storage medium based on flexible resources. In terms of the privacy data protection effect of the flexible resource end on the aggregator side, the dispatching end on the power coordination and dispatching side cannot obtain any information related to the privacy data of the flexible resource end on the aggregator side from the first intermediate parameter, and the mask value is independently generated by the flexible resource end on each aggregator side in each round, achieving the information theoretic security of "one-time one-key"; in terms of collaborative optimization accuracy, in the process of calculating the mask value by the flexible resource end on the aggregator side, the sum of the mask value generated in this round of iteration and the mask value generated by the flexible resource end on other aggregators is 0, so the mask value has no effect on the collaborative optimization calculation, and the collaborative optimization accuracy is not lost; in terms of additional computing and communication overhead, the calculation process of the mask value additionally introduced to improve the data privacy security of the flexible resource end on the aggregator side does not involve parameter interaction with the dispatching end on the power coordination and dispatching side, so it can be completed offline in advance, and the operation only involves addition and subtraction operations, which can reduce the computing and communication overhead and has high timeliness. Therefore, the embodiments of the present invention have the advantages of high collaborative optimization accuracy, low additional computing and communication overhead, and strong privacy protection effect. In addition to protecting the privacy data security on the aggregator side, it can also protect sensitive data on the power dispatching side, fully meeting the needs of many aggregators, large amounts of collaborative optimization interaction data, high requirements for collaborative optimization result accuracy, and high requirements for optimization timeliness in the demand-side flexible resource distributed collaborative optimization scenario, and can effectively ensure the security of sensitive data on the aggregator side and the power dispatching side. BRIEF DESCRIPTION OF THE DRAWINGS

[0019] In order to more clearly express the technical solutions of the embodiments of the present invention, the drawings required for describing the embodiments are briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without creative work.

[0020] Figure 1 It is a structural diagram of a distributed collaborative optimization system based on flexible resources in an embodiment of the present invention;

[0021] Figure 2 It is a flow chart of a distributed collaborative optimization method based on flexible resources in an embodiment of the present invention;

[0022] Figure 3 is a flow chart of another distributed collaborative optimization method based on flexible resources in an embodiment of the present invention;

[0023] Figure 4 Schematic diagram of the structure of the flexible resource end on the aggregator side in an embodiment of the present invention;

[0024] Figure 5 Schematic diagram of the structure of the dispatching end of the power coordination dispatching side in an embodiment of the present invention;

[0025] Figure 6 Schematic diagram of the structure of a computer-readable storage medium in an embodiment of the present invention. DETAILED DESCRIPTION

[0026] In order to enable those skilled in the art to better understand the scheme of the present invention, the technical scheme in the embodiments of the present invention will be clearly and completely described below in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0027] The distributed collaborative optimization system based on flexible resources in the embodiment of the present invention is applicable to the field of power dispatching, and is particularly applicable to a joint collaborative computing scenario consisting of N demand-side flexible resource aggregators (such as virtual power plants, smart buildings, electric vehicle operators, etc.) and one power coordination dispatcher. The dispatching end on the power coordination dispatching side in the embodiment of the present invention is a terminal device set on the power coordination dispatching side, and the flexible resource end on the aggregator side is a terminal device set on the demand-side flexible resource aggregator side, and the terminal device includes a computer, a mobile phone, etc.

[0028] The embodiment of the present invention provides a distributed collaborative optimization system based on flexible resources, such as Figure 1 As shown, the system includes:

[0029] The dispatching end of the power coordination dispatching side and the flexible resource ends of several aggregators, each flexible resource end of the aggregator side is connected to the dispatching end of the power coordination dispatching side, and any two flexible resource ends of the aggregator side are connected to each other;

[0030] The flexible resource end on the aggregator side generates a first calculation parameter through parameter initialization;

[0031] The dispatching end of the electric power coordination dispatching side generates Lagrange multipliers and mask matrices by initializing parameters, and converts the first global constraint parameter and the second global constraint parameter according to the mask matrix to obtain the first conversion parameter and the second conversion parameter;

[0032] The flexible resource end on the aggregator side generates a mask value for this round of iteration, wherein the sum of the mask value generated for this round of iteration and the mask value generated by the flexible resource end on other aggregator sides is 0, a first preset function is constructed according to the local objective function and the first calculation parameter, and the local objective parameter corresponding to the minimum value of the first preset function is calculated, a first intermediate parameter is calculated according to the first calculation parameter, the local objective parameter corresponding to the minimum value of the first preset function, and the mask value for this round of iteration, and the first intermediate parameter is sent to the dispatching end on the power coordination dispatching side;

[0033] The dispatching end on the power coordination dispatching side receives the first intermediate parameters sent by the flexible resource ends on all aggregator sides, adds all the first intermediate parameters to obtain the second intermediate parameters, and updates the Lagrange multiplier according to the Lagrange multiplier, the second intermediate parameter, the second conversion parameter and the preset gradient parameter;

[0034] The dispatching end on the power coordination and dispatching side determines whether the preset iteration termination condition is met. If the preset iteration termination condition is met, the dispatching end on the power coordination and dispatching side terminates the calculation, and the flexible resource end on the aggregator side outputs the local target parameter corresponding to the value of the first preset function calculated in the last round of iteration process is minimum. If the preset iteration termination condition is not met, the dispatching end on the power coordination and dispatching side updates the first calculation parameter according to the first conversion parameter and the updated Lagrange multiplier, sends the updated first calculation parameter to the flexible resource end on the aggregator side, and enters the next round of iteration, wherein each round of iteration process includes the process in which the flexible resource end on the aggregator side generates a mask value for this round of iteration and the dispatching end on the power coordination and dispatching side sends the updated first calculation parameter to the flexible resource end on the aggregator side.

[0035] The main participants of the distributed collaborative optimization system based on flexible resources in the embodiment of the present invention include N flexible resource terminals on the aggregator side, marked as A1, A2, ..., A N , usually N ≥ 3, and a dispatching end on the power coordination and dispatching side, counted as D.

[0036] Flexible resource end A on each aggregator side i Hold your own private data x i and the local objective function f i (x i ). The dispatching terminal D on the power coordination dispatching side has M global constraints, including the first global constraint parameter B i and the second global constraint parameter c, and it is hoped that under the action of the global constraint and with the cooperation of various demand-side flexible resource aggregators, the global objective function can be minimized. The problem can be described as follows:

[0037]

[0038] in, x=[x1…x N ].

[0039] The Lagrangian function corresponding to the optimization problem of formula (1) is:

[0040]

[0041] Using the dual decomposition method, the solution to equation (2) can be dynamically completed through an iterative algorithm. The general process of the dual decomposition method can be described by the following steps:

[0042]

[0043] Where k is the iteration number mark, α is the preset gradient parameter, and the first step in formula (3) is performed by the flexible resource end A on each aggregator side. i The second step is completed in parallel by the dispatching terminal D on the power coordination dispatching side. To protect the flexible resource terminal A on each aggregator side in formula (3), i Local privacy data x i (such as power consumption, power consumption time, etc.), the first global constraint parameter B held by the dispatching terminal D on the power coordination dispatching side i The second global constraint parameter c is not known to others. The present invention performs privacy protection modification on the method in formula (3). The main process is described as follows:

[0044] Step (1): System initialization phase:

[0045] Step (1-1): Flexible resource end A on the aggregator side i initialization.

[0046] Flexible resource end A on each aggregator side i , initialize the parameters and generate the first calculation parameters Specifically, the flexible resource terminal A on the aggregator side i The initialization formula is:

[0047]

[0048] Step (1-2): The dispatching terminal D on the power coordination and dispatching side performs initialization and mask calculation.

[0049] Step (1-2-1): Initialize the Lagrange multiplier λ 0 :

[0050] λ 0 =0

[0051] Step (1-2-2): Generate a random mask matrix T, where the mask matrix T is a non-singular matrix with M rows and columns, where M is the number of global constraints known by the dispatching terminal D on the power coordination dispatching side. And the first global constraint parameter B is constrained by the mask matrix T i The first conversion parameter is converted into the second global constraint parameter c and the second conversion parameter

[0052]

[0053] Step (2): System iterative optimization calculation phase:

[0054] The iterative optimization calculation is carried out iteratively between step (2-1) and step (2-2) in sequence. The specific steps are as follows:

[0055] Step (2-1): Flexible resource end A on each aggregator side i , perform iterative optimization calculations according to the following steps:

[0056] Step (2-1-1): Flexible resource end A on each aggregator side i , generate a random vector For 1≤i≤N-1, the flexible resource end A on the aggregator side i The random vector Safely sent to A i+1 ; For i = N, the flexible resource terminal A on the aggregator side i The random vector Send to A1 securely;

[0057] Step (2-1-2): Flexible resource end A on each aggregator side i A random vector computed from itself The mask value of this round of iteration is calculated using the random vector sent by the flexible resource end of the other aggregator side, wherein the sum of the mask value of this round of iteration generated and the mask value generated by the flexible resource end of the other aggregator side is 0. Specifically, the mask value is calculated as follows:

[0058] When i=1,

[0059] When 2≤i≤N,

[0060] The mask value is calculated by this method It is convenient to set the flexible resource end on each aggregator side to pre-match the flexible resource end on the aggregator side, and ensure that the sum of all mask values ​​is 0.

[0061] Step (2-1-3): Flexible resource end A on each aggregator side i Obtain the first calculation parameter, wherein, in the non-first round of iteration, the first calculation parameter is obtained from the dispatching terminal D of the power coordination dispatching side In the first iteration, the first calculated parameter The flexible resource end Ai on the aggregator side is initialized by parameters, and each flexible resource end A on the aggregator side is i According to formula (5), the local target parameter corresponding to the minimum value of the first preset function is calculated and recorded as

[0062]

[0063] (5) In the formula, is the first preset function, f i (x i ) is the local objective function. Since the first calculation parameter It is known that the first preset function can be regarded as a function of the local target parameter The local target parameter when the value of the first preset function is minimum is obtained by derivation. The corresponding value

[0064] Step (2-1-4): Flexible resource end A on each aggregator side i According to the first calculation parameter The local target parameter corresponding to the minimum value of the first preset function and the mask value of this iteration Calculate the first intermediate parameter Specifically, the first intermediate parameter is calculated according to the following formula:

[0065]

[0066] And the first intermediate parameter The dispatch terminal D sent to the power coordination dispatch side uses the first calculation parameter and mask value Pair Value Perform the conversion and generate the first intermediate parameter for iteration While protecting privacy data And there is no need for additional encryption and decryption processes, the data processing and transmission volumes remain unchanged, and the timeliness of distributed collaborative optimization is improved.

[0067] Step (2-2) The dispatching terminal D on the power coordination dispatching side performs iterative optimization calculation according to the following steps:

[0068] Step (2-2-1): The dispatching terminal D on the power coordination dispatching side receives the flexible resource terminal A on each aggregator side. i The first intermediate parameter sent All first intermediate parameters Add to get the second intermediate parameter

[0069]

[0070] Step (2-2-2): The dispatching terminal D on the power coordination dispatching side calculates the current Lagrange multiplier λ k , the second intermediate parameter Second conversion parameter And the preset gradient parameter α updates the Lagrange multiplier. Specifically, the updated Lagrange multiplier λ is calculated according to the following formula k+1 :

[0071]

[0072] Step (2-2-3): The dispatching terminal D on the power coordination dispatching side determines the maximum number of iterations or the updated Lagrange multiplier λ according to the preset maximum number of iterations k+1 Lagrange multiplier λ before relative update k The range of changes in determines whether this iteration optimization is terminated; if the preset iteration termination condition is reached, this iteration optimization calculation ends, and the flexible resource end A on the aggregator side i The local target parameter corresponding to the minimum value of the calculated first preset function is the iterative optimization result; if the termination condition is not reached, proceed to step (2-2-4);

[0073] (2-2-4): The dispatching terminal D on the power coordination dispatching side converts the first conversion parameter and the updated Lagrange multiplier λ k+1 Update the first calculation parameter. Specifically, calculate the updated first calculation parameter according to the following formula:

[0074]

[0075] Then, the dispatching terminal D on the power coordination dispatching side updates the first calculation parameter Corresponding to the flexible resource terminal A sent to the aggregator side i , Flexible resource end A on the aggregator side i Receive the updated first calculation parameter Execute steps (2-1) and (2-2) again and repeat the iteration.

[0076] The distributed collaborative optimization system based on flexible resources proposed in the present invention has the advantages of high collaborative optimization accuracy, low additional computing and communication overhead, and strong privacy protection effect. In addition to protecting the privacy data security of the aggregator side, i.e., the flexible resource end of the aggregator side, it can also protect the sensitive data of the dispatching end of the power dispatching side, i.e., the power coordination dispatching side:

[0077] (1) In terms of collaborative optimization accuracy, the mask value is calculated by the flexible resource end on the aggregator side. In the process, the sum of the mask value generated in this iteration and the mask value generated by the flexible resource end of other aggregators is 0, that is, When the dispatching end of the power coordination dispatching side updates the Lagrange multiplier, that is, when calculating formulas (7) and (8), the mask value has no effect on the update of the Lagrange multiplier, so the mask value There is no impact on collaborative optimization calculations, and the collaborative optimization accuracy is not lost;

[0078] (2) In terms of additional computation and communication overhead, the mask value is introduced to improve the data privacy security of the flexible resource end on the aggregator side. The calculation process does not involve parameter interaction with the dispatching end of the power coordination dispatching side, so it can be completed offline in advance, and the operation only involves addition and subtraction operations, which can reduce the calculation and communication overhead;

[0079] (3) In terms of the privacy data protection effect of the flexible resource end on the aggregator side, the dispatch end on the power coordination dispatch side cannot obtain the privacy data from the first intermediate parameter. The privacy data of the flexible resource end of the aggregator side, namely the local target parameters, is obtained Any information about the Each round is independently generated by the flexible resource end of each aggregator, achieving the information-theoretic security of "one-time one-key";

[0080] (4) In terms of sensitive data protection at the dispatching end of the power coordination dispatching side, the dispatching end of the power coordination dispatching side introduces a mask matrix T to act on the first global constraint parameter B. i and the second global constraint parameter c, so that the flexible resource end of the aggregator side cannot be transferred from the dispatch end of the power coordination dispatch side. The exact values ​​of Bi and c are obtained, thereby protecting the first global constraint parameter B i and the safety of the second global constraint parameter c.

[0081] (5) When introducing the mask value There is no need for additional encryption and decryption process based on the value, the data processing and transmission volume remain unchanged, and the timeliness of distributed collaborative optimization is improved.

[0082] In addition, the distributed collaborative optimization system based on flexible resources in the embodiment of the present invention can be easily expanded to multi-level collaborative optimization scenarios, and is still applicable in application scenarios such as multi-level "virtual power plants". In this scenario, it is only necessary to reuse the system in layers, and each layer is independent of each other and does not affect each other.

[0083] In summary, the distributed collaborative optimization system based on flexible resources in the embodiment of the present invention has the advantages of high collaborative optimization accuracy, low additional computing and communication overhead, and strong privacy protection effect. In addition to protecting the privacy data security of the flexible resource end on the aggregator side, it can also protect the sensitive data of the dispatching end on the power coordination and dispatching side. It fully meets the requirements of the distributed collaborative optimization scenario of flexible resources on the demand side, such as the large number of flexible resource ends on the aggregator side, the large amount of collaborative optimization interaction data, the high requirements for the accuracy of collaborative optimization results, and the high requirements for optimization timeliness. It can effectively ensure the security of sensitive data on the flexible resource end on the aggregator side and the dispatching end on the power coordination and dispatching side.

[0084] The embodiment of the present invention also provides a distributed collaborative optimization method based on flexible resources, which is applied to the flexible resource end of the aggregator side, such as Figure 2 As shown, the method includes:

[0085] Step S201, generating first calculation parameters through parameter initialization.

[0086] Flexible resource side A on the aggregator side i , initialize the parameters and generate the first calculation parameters Specifically, the flexible resource terminal A on the aggregator side i The initialization formula is:

[0087]

[0088] Step S202: Generate a mask value for this round of iteration, wherein the sum of the mask value generated for this round of iteration and the mask value generated by the flexible resource end of other aggregator side is 0.

[0089] Specifically, in each iteration, the flexible resource end A of each aggregator side i Generate the mask values ​​for this round of iteration respectively Flexible resource end A on the current aggregator side i The sum of the mask value generated in this iteration and the mask value generated by the flexible resource end of other aggregators is 0, so that the dispatch end D on the power coordination dispatch side does not suffer any loss in the collaborative optimization accuracy when updating the Lagrange multiplier.

[0090] Step S203, constructing a first preset function according to the local objective function and the first calculation parameter, and calculating the local objective parameter corresponding to the minimum value of the first preset function.

[0091] Specifically, the first preset function is Calculate the local target parameter corresponding to the minimum value of the first preset function, that is:

[0092]

[0093] Since the first calculation parameter It is known that the first preset function can be regarded as a function of the local target parameter The local target parameter when the value of the first preset function is minimum is obtained by derivation. The corresponding value

[0094] Step S204, calculating a first intermediate parameter according to the first calculation parameter, the local target parameter corresponding to the minimum value of the first preset function and the mask value of this iteration, and sending the first intermediate parameter to the dispatching end of the power coordination dispatching side.

[0095] Flexible resource side A on the aggregator side i According to the first calculation parameter The local target parameter corresponding to the minimum value of the first preset function and the mask value of this iteration Calculate the first intermediate parameter Specifically, the first intermediate parameter is calculated according to the following formula:

[0096]

[0097] And the first intermediate parameter Sent to the dispatching terminal D on the power coordination and dispatching side.

[0098] Step S205, determines whether to perform the next round of iteration based on the updated first calculation parameter according to whether the scheduling end on the power coordination and scheduling side returns the updated first calculation parameter, wherein each round of iteration process includes the process of generating the mask value of this round of iteration to sending the first intermediate parameter to the scheduling end on the power coordination and scheduling side.

[0099] Specifically, the dispatching terminal D on the power coordination dispatching side receives the flexible resource terminal A on each aggregator side. i The first intermediate parameter sent All first intermediate parameters Add to get the second intermediate parameter

[0100]

[0101] Then the dispatching end D on the power coordination dispatching side is based on the current Lagrange multiplier λ k, the second intermediate parameter Second conversion parameter Update the Lagrange multiplier with the preset gradient parameter α, and calculate the updated Lagrange multiplier λ using the following formula: k+1 :

[0102]

[0103] Then, the dispatching terminal D on the power coordination dispatching side calculates the maximum number of iterations or the updated Lagrange multiplier λ according to the preset maximum number of iterations k+1 Lagrange multiplier λ before relative update k The range of changes in determines whether this iteration optimization is terminated; if the preset iteration termination condition is reached, this iteration optimization calculation ends, and the flexible resource end A on the aggregator side i The local target parameter corresponding to the minimum value of the calculated first preset function is the iterative optimization result; if the termination condition is not reached, the dispatching end D on the power coordination dispatching side is based on the first conversion parameter and the updated Lagrange multiplier λ k+1 Update the first calculation parameter. Specifically, calculate the updated first calculation parameter according to the following formula:

[0104]

[0105] Then, the dispatching terminal D on the power coordination dispatching side updates the first calculation parameter Corresponding to the flexible resource terminal A sent to the aggregator side i , Flexible resource end A on the aggregator side i Receive the updated first calculation parameter Return to step S202 and repeat the iteration

[0106] Step S206, outputting the local target parameter corresponding to the time when the value of the first preset function calculated in the last round of iteration process is minimum.

[0107] Specifically, the local target parameter corresponding to the minimum value of the first preset function calculated in the last round of iteration is the iterative optimization result. The local target parameter is the flexible resource end A of the aggregator side. i The privacy parameters such as electricity consumption, flexible resource terminals A on the aggregator side i Based on the local target parameters, the system adjusts its own power consumption to achieve the global optimum.

[0108] In a distributed collaborative optimization method based on flexible resources according to an embodiment of the present invention, in terms of collaborative optimization accuracy, the flexible resource end of the aggregator side calculates the mask value. In the process, the sum of the mask value generated in this iteration and the mask value generated by the flexible resource end of other aggregators is 0, that is, When updating the Lagrange multiplier at the dispatch end of the power coordination dispatch side, the mask value has no effect on the update of the Lagrange multiplier, so the mask value There is no impact on the collaborative optimization calculation, and the collaborative optimization accuracy is not lost.

[0109] In terms of additional computation and communication overhead, additional mask values ​​are introduced to improve the privacy and security of data on the flexible resource side of the aggregator. The calculation process does not involve parameter interaction with the dispatching end of the power coordination and dispatching side, so it can be completed offline in advance, and the calculation only involves addition and subtraction operations, which can reduce the calculation and communication overhead and has high timeliness.

[0110] In addition, in terms of the privacy data protection effect of the flexible resource end of the aggregator side, the dispatch end of the power coordination dispatch side cannot obtain the privacy data from the first intermediate parameter. The privacy data of the flexible resource end of the aggregator side, namely the local target parameters, is obtained Any information about the Each round is independently generated by the flexible resource end of each aggregator, achieving the information-theoretic security of "one-time, one-key".

[0111] In one embodiment, step S202, generating a mask value for this round of iteration, includes:

[0112] Step S2021, generating a first random vector;

[0113] Step S2022, receiving a second random vector sent by a pre-matched flexible resource terminal on the aggregator side, wherein the pre-matched flexible resource terminal on the aggregator side of each flexible resource terminal on the aggregator side is different;

[0114] Step S2023: Generate a mask value for this round of iteration according to the difference between the first random vector and the second random vector.

[0115] Specifically, for the flexible resource terminal A on the current aggregator side i , the generated random vector is recorded as the first random vector. The random vector generated by the flexible resource end of the pre-matched aggregator side is recorded as the first random vector. The flexible resource end of the pre-matched aggregator side is the flexible resource end A except the current aggregator side. i The flexible resource ends on the aggregator side other than the flexible resource ends on the aggregator side are different from the flexible resource ends on the aggregator side pre-matched, ensuring that the pre-matched flexible resource ends on the aggregator side do not overlap, that is, ensuring that the flexible resource ends on the aggregator side A i The resulting mask value is added to 0.

[0116] For example, all flexible resource terminals A on the aggregator side i Sorting, for 1≤i≤N-1, the flexible resource end A on the aggregator side i The random vector Safely sent to A i+1 , i.e. the flexible resource end A on the aggregator side i+1 The flexible resource end of the pre-matched aggregator side is A i ; For i = N, the flexible resource terminal A on the aggregator side i The random vector It is sent to A1 securely, that is, the flexible resource end of the aggregator side pre-matched by A1 is A N For each aggregator's flexible resource end A i , the mask value is calculated as follows

[0117] When i=1,

[0118] When 2≤i≤N,

[0119] The mask value generated by the above formula Ensures flexible resource end A on each aggregator side i The generated mask value is added to 0. When the dispatching end of the subsequent power coordination dispatching side updates the Lagrange multiplier, the mask value has no effect on the update of the Lagrange multiplier, so the mask value There is no impact on the collaborative optimization calculation, and the collaborative optimization accuracy is not lost.

[0120] In one embodiment, before generating the first calculation parameter by parameter initialization, the method further includes:

[0121] Establish a secure channel with the dispatching end on the power coordination and dispatching side and the flexible resource end on other aggregator sides.

[0122] Specifically, the dispatching terminal D on the power coordination dispatching side and the flexible resource terminal A on the aggregator side i Need to complete identity registration and authentication first, and have flexible resource terminals A on each aggregator side i Flexible resource end A between the aggregator and the i The basic condition for establishing a secure channel with the dispatching terminal D on the power coordination dispatching side. The specific establishment method can be quickly established using the method of the patent "Internet of Things Identity Authentication Method, Device and Power Internet of Things Identity Authentication System (ZL202010845406.5)". After the secure channel is established, the flexible resource terminal A on each aggregator side i Flexible resource end A between the aggregator and thei Parameters are transmitted between the dispatching end D on the power coordination and dispatching side through a secure channel to improve the data security of the system.

[0123] The embodiment of the present invention also provides a distributed collaborative optimization method based on flexible resources, which is applied to the dispatching end of the power coordination dispatching side, such as Figure 3 As shown, the method includes:

[0124] Step S301 : generating Lagrange multipliers and a mask matrix by initializing parameters, and converting the first global constraint parameter and the second global constraint parameter according to the mask matrix to obtain the first conversion parameter and the second conversion parameter.

[0125] Specifically, the dispatching terminal D on the power coordination dispatching side performs parameter initialization including: initializing the Lagrange multiplier λ 0 :

[0126] λ 0 =0

[0127] Generate a random mask matrix T, and the mask matrix T is a non-singular matrix with M rows and columns, where M is the number of global constraints mastered by the dispatching terminal D on the power coordination dispatching side. The first conversion parameter is calculated according to the code matrix T, the first global constraint parameter Bi and the second global constraint parameter c. and the second conversion parameter

[0128]

[0129]

[0130] Step S302: receiving first intermediate parameters sent by all flexible resource terminals on the aggregator side, and adding all first intermediate parameters to obtain second intermediate parameters.

[0131] Specifically, the dispatching terminal D on the power coordination dispatching side receives the flexible resource terminal A on each aggregator side. i The first intermediate parameter sent All first intermediate parameters Add to get the second intermediate parameter

[0132]

[0133] Step S303: updating the Lagrange multiplier according to the Lagrange multiplier, the second intermediate parameter, the second conversion parameter and the preset gradient parameter.

[0134] The dispatching end D on the power coordination dispatching side is based on the current Lagrange multiplier λ k , the second intermediate parameter Second conversion parameter And the preset gradient parameter α updates the Lagrange multiplier. Specifically, the updated Lagrange multiplier λ is calculated according to the following formula k+1 :

[0135]

[0136] Step S304, determine whether the preset iteration termination condition is met. If the preset iteration termination condition is met, terminate the calculation. If the preset iteration termination condition is not met, update the first calculation parameter according to the first conversion parameter and the updated Lagrange multiplier, send the updated first calculation parameter to the flexible resource end on the aggregator side, and enter the next round of iteration, wherein each round of iteration process includes the process of receiving the first intermediate parameter sent by all flexible resource ends on the aggregator side to sending the updated first calculation parameter to the flexible resource end on the aggregator side.

[0137] The dispatching end D on the power coordination dispatching side determines whether this iteration optimization is terminated based on whether the preset iteration termination condition is met; if the preset iteration termination condition is met, this iteration optimization calculation is terminated, and the flexible resource end A on the aggregator side i The local target parameter corresponding to the minimum value of the calculated first preset function is the iterative optimization result; if the termination condition is not reached, the dispatching end D on the power coordination dispatching side converts the local target parameter into the local target parameter according to the first conversion parameter. and the updated Lagrange multiplier λ k+1 Update the first calculation parameter. Specifically, calculate the updated first calculation parameter according to the following formula:

[0138]

[0139] Then, the dispatching terminal D on the power coordination dispatching side updates the first calculation parameter Corresponding to the flexible resource terminal A sent to the aggregator side i , Flexible resource end A on the aggregator side i Receive the updated first calculation parameter Enter the next iteration.

[0140] In the distributed collaborative optimization method based on flexible resources of the embodiment of the present invention, in terms of sensitive data protection of the dispatching end of the power coordination dispatching side, the dispatching end of the power coordination dispatching side introduces a mask matrix T to act on the first global constraint parameter B i and the second global constraint parameter c, so that the flexible resource end of the aggregator side cannot be transferred from the dispatch end of the power coordination dispatch side. Get B i and c, thereby protecting the first global constraint parameter B iand the safety of the second global constraint parameter c.

[0141] In terms of collaborative optimization accuracy, the mask value is calculated by the flexible resource end on the aggregator side. In the process, the sum of the mask value generated in this iteration and the mask value generated by the flexible resource end of other aggregators is 0, that is, When updating the Lagrange multiplier at the dispatch end of the power coordination dispatch side, the mask value has no effect on the update of the Lagrange multiplier, so the mask value There is no impact on the collaborative optimization calculation, and the collaborative optimization accuracy is not lost.

[0142] In one embodiment, in step S304, determining whether a preset iteration termination condition is reached includes:

[0143] It is determined whether the number of iterations reaches the preset number of iterations or whether the range of change of the Lagrange multiplier reaches the preset termination value. If the number of iterations reaches the preset number of iterations or the range of change of the Lagrange multiplier reaches the preset termination value, the preset iteration termination condition is met. If the number of iterations does not reach the preset number of iterations and the range of change of the Lagrange multiplier does not reach the preset termination value, the preset iteration termination condition is not met.

[0144] Specifically, the larger the preset number of iterations, the smaller the variation range of the Lagrange multiplier, and the more accurate the optimization result. Based on this, by setting the preset number of iterations and the variation range of the Lagrange multiplier, an optimization result of corresponding accuracy can be obtained.

[0145] In one embodiment, before receiving the first intermediate parameters sent by the flexible resource terminals of all aggregators, the method further includes:

[0146] Establish a secure channel with the flexible resource end on the aggregator side.

[0147] Specifically, the dispatching terminal D on the power coordination dispatching side and the flexible resource terminal A on the aggregator side i Need to complete identity registration and authentication first, and have flexible resource terminals A on each aggregator side i Flexible resource end A between the aggregator and the i The basic condition for establishing a secure channel with the dispatching terminal D on the power coordination dispatching side. The specific establishment method can be quickly established using the method of the patent "Internet of Things Identity Authentication Method, Device and Power Internet of Things Identity Authentication System (ZL202010845406.5)". After the secure channel is established, the flexible resource terminal A on each aggregator side i Flexible resource end A between the aggregator and the i Parameters are transmitted between the dispatching end D on the power coordination and dispatching side through a secure channel to improve the data security of the system.

[0148] The embodiment of the present invention also provides a flexible resource terminal on the aggregator side, such as Figure 4 As shown, including:

[0149] The first initialization module 401 is used to generate a first calculation parameter by initializing a parameter; for details, please refer to the corresponding part of the above method embodiment, which will not be repeated here.

[0150] The mask generation module 402 is used to generate the mask value of this round of iteration, wherein the sum of the mask value generated for this round of iteration and the mask value generated by the flexible resource end of other aggregators is 0; for specific contents, please refer to the corresponding part of the above method embodiment, which will not be repeated here.

[0151] The minimum value calculation module 403 is used to construct a first preset function based on the local objective function and the first calculation parameter, and calculate the local objective parameter corresponding to the minimum value of the first preset function; the specific content refers to the corresponding part of the above method embodiment, which will not be repeated here.

[0152] The first parameter calculation module 404 is used to calculate the first intermediate parameter based on the first calculation parameter, the local target parameter corresponding to the minimum value of the first preset function, and the mask value of this round of iteration, and send the first intermediate parameter to the dispatching end of the power coordination and dispatching side; for specific contents, please refer to the corresponding part of the above method embodiment, which will not be repeated here.

[0153] The first iteration module 405 is used to determine whether to perform the next round of iteration based on the updated first calculation parameters according to whether the scheduling end on the power coordination and scheduling side returns the updated first calculation parameters, wherein each round of iteration process includes the process of generating the mask value of this round of iteration to sending the first intermediate parameter to the scheduling end on the power coordination and scheduling side; for specific contents, please refer to the corresponding part of the above method embodiment, which will not be repeated here.

[0154] The output module 406 is used to output the local target parameter corresponding to the minimum value of the first preset function calculated in the last round of iteration. For details, please refer to the corresponding part of the above method embodiment, which will not be repeated here.

[0155] The flexible resource end on the aggregator side proposed in the embodiment of the present invention calculates the mask value in terms of collaborative optimization accuracy. In the process, the sum of the mask value generated in this iteration and the mask value generated by the flexible resource end of other aggregators is 0, that is, When updating the Lagrange multiplier at the dispatch end of the power coordination dispatch side, the mask value has no effect on the update of the Lagrange multiplier, so the mask value There is no impact on the collaborative optimization calculation, and the collaborative optimization accuracy is not lost.

[0156] In terms of additional computation and communication overhead, additional mask values ​​are introduced to improve the privacy and security of data on the flexible resource side of the aggregator. The calculation process does not involve parameter interaction with the dispatching end of the power coordination and dispatching side, so it can be completed offline in advance, and the calculation only involves addition and subtraction operations, which can reduce the calculation and communication overhead and has high timeliness.

[0157] In addition, in terms of the privacy data protection effect of the flexible resource end of the aggregator side, the dispatch end of the power coordination dispatch side cannot obtain the privacy data from the first intermediate parameter. The privacy data of the flexible resource end of the aggregator side, namely the local target parameters, is obtained Any information about the Each round is independently generated by the flexible resource end of each aggregator, achieving the information-theoretic security of "one-time, one-key".

[0158] The embodiment of the present invention also provides a dispatching terminal on the power coordination dispatching side, such as Figure 5 As shown, including:

[0159] The second initialization module 501 is used to generate Lagrange multipliers and mask matrices through parameter initialization, and convert the first global constraint parameters and the second global constraint parameters according to the mask matrix to obtain first conversion parameters and second conversion parameters; the specific content refers to the corresponding part of the above method embodiment, which will not be repeated here.

[0160] The second parameter calculation module 502 is used to receive the first intermediate parameters sent by the flexible resource terminals of all aggregators, and add all the first intermediate parameters to obtain the second intermediate parameters; the specific content can be found in the corresponding part of the above method embodiment, which will not be repeated here.

[0161] The multiplier updating module 503 is used to update the Lagrange multiplier according to the Lagrange multiplier, the second intermediate parameter, the second conversion parameter and the preset gradient parameter; the specific content refers to the corresponding part of the above method embodiment, which will not be repeated here.

[0162] The second iteration module 504 is used to determine whether the preset iteration termination condition is reached. If the preset iteration termination condition is reached, the calculation is terminated. If the preset iteration termination condition is not reached, the first calculation parameter is updated according to the first conversion parameter and the updated Lagrange multiplier, and the updated first calculation parameter is sent to the flexible resource end of the aggregator side, and the next round of iteration is entered, wherein each round of iteration process includes receiving the first intermediate parameter sent by the flexible resource end of all the aggregators to sending the updated first calculation parameter to the flexible resource end of the aggregator side. For specific content, please refer to the corresponding part of the above method embodiment, which will not be repeated here.

[0163] The dispatching end of the power coordination dispatching side proposed in the embodiment of the present invention, in terms of sensitive data protection of the dispatching end of the power coordination dispatching side, because the dispatching end of the power coordination dispatching side introduces a mask matrix T acting on the first global constraint parameter B i and the second global constraint parameter c, so that the flexible resource end of the aggregator side cannot be transferred from the dispatch end of the power coordination dispatch side. Get B i and c, thereby protecting the first global constraint parameter B i and the safety of the second global constraint parameter c.

[0164] In terms of collaborative optimization accuracy, the mask value is calculated by the flexible resource end on the aggregator side. In the process, the sum of the mask value generated in this iteration and the mask value generated by the flexible resource end of other aggregators is 0, that is, When updating the Lagrange multiplier at the dispatch end of the power coordination dispatch side, the mask value has no effect on the update of the Lagrange multiplier, so the mask value There is no impact on the collaborative optimization calculation, and the collaborative optimization accuracy is not lost.

[0165] The embodiment of the present invention also provides a computer readable storage medium, such as Figure 6As shown, a computer program 510 is stored thereon, and when the instruction is executed by the processor, the steps of the distributed collaborative optimization method based on flexible resources in the above-mentioned embodiment are implemented. The storage medium also stores audio and video stream data, feature frame data, interactive request signaling, encrypted data, and preset data size, etc. Among them, the storage medium can be a disk, an optical disk, a read-only memory (Read-Only Memory, ROM), a random access memory (Random Access Memory, RAM), a flash memory (Flash Memory), a hard disk (Hard Disk Drive, abbreviated as: HDD) or a solid-state drive (Solid-State Drive, SSD), etc.; the storage medium can also include a combination of the above-mentioned types of memory. Those skilled in the art can understand that the implementation of all or part of the processes in the above-mentioned embodiment method can be completed by instructing the relevant hardware through a computer program. The computer program 510 can be stored in a computer-readable storage medium. When the program is executed, it can include the processes of the embodiments of the above-mentioned methods. Among them, the storage medium can be a magnetic disk, an optical disk, a read-only memory (ROM), a random access memory (RAM), a flash memory (Flash Memory), a hard disk (Hard Disk Drive, abbreviated as: HDD) or a solid-state drive (Solid-State Drive, SSD), etc.; the storage medium can also include a combination of the above types of memory.

[0166] The above embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit the same. Although the present invention has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that the technical solutions described in the aforementioned embodiments may still be modified, or some of the technical features thereof may be replaced by equivalents. However, these modifications or replacements do not deviate the essence of the corresponding technical solutions from the spirit and scope of the technical solutions of the embodiments of the present invention.

Claims

1. A distributed collaborative optimization method based on flexible resources, applied to the flexible resource end of the aggregator side, It is characterized in that include: Generate a first calculation parameter by initializing the parameter; Generate a mask value for this round of iteration, where the sum of the mask value for this round of iteration and the mask value generated by the flexible resource end of other aggregators is 0; Constructing a first preset function according to the local objective function and the first calculation parameter, and calculating the local objective parameter corresponding to the minimum value of the first preset function; Calculating a first intermediate parameter according to the first calculation parameter, the local target parameter corresponding to the minimum value of the first preset function, and the mask value of this round of iteration, and sending the first intermediate parameter to the dispatching end of the power coordination dispatching side; Determining whether to perform the next round of iteration based on the updated first calculation parameter according to whether the dispatching end of the power coordination dispatching side returns the updated first calculation parameter, wherein each round of iteration process includes the process of generating the mask value of the iteration to sending the first intermediate parameter to the dispatching end of the power coordination dispatching side; Output the local target parameter corresponding to the minimum value of the first preset function calculated in the last round of iteration process.

2. The distributed collaborative optimization method based on flexible resources according to claim 1, It is characterized in that The mask value generated in this round of iteration includes: Generate a first random vector; receiving a second random vector sent by a pre-matched flexible resource end on an aggregator side, wherein the pre-matched flexible resource end on an aggregator side of each flexible resource end on an aggregator side is different; A mask value of this round of iteration is generated according to a difference between the first random vector and the second random vector.

3. The distributed collaborative optimization method based on flexible resources according to claim 1, It is characterized in that Before generating the first calculation parameter by parameter initialization, the method further includes: Establish a secure channel with the dispatching end on the power coordination and dispatching side and the flexible resource end on other aggregator sides.

4. A distributed collaborative optimization method based on flexible resources, applied to the dispatching end of power coordination dispatching, It is characterized in that include: Generate Lagrange multipliers and a mask matrix by initializing parameters, and convert the first global constraint parameter and the second global constraint parameter according to the mask matrix to obtain a first conversion parameter and a second conversion parameter; receiving first intermediate parameters sent by all flexible resource terminals on the aggregator side, and adding all the first intermediate parameters to obtain a second intermediate parameter; updating the Lagrange multiplier according to the Lagrange multiplier, the second intermediate parameter, the second conversion parameter and a preset gradient parameter; Determine whether the preset iteration termination condition is met. If the preset iteration termination condition is met, terminate the calculation. If the preset iteration termination condition is not met, update the first calculation parameter according to the first conversion parameter and the updated Lagrange multiplier, send the updated first calculation parameter to the flexible resource end on the aggregator side, and enter the next round of iteration, wherein each round of iteration process includes the process of receiving the first intermediate parameter sent by all flexible resource ends on the aggregator side to sending the updated first calculation parameter to the flexible resource end on the aggregator side.

5. The distributed collaborative optimization method based on flexible resources according to claim 4, It is characterized in that The determining whether a preset iteration termination condition is reached includes: It is determined whether the number of iterations reaches the preset number of iterations or whether the range of change of the Lagrange multiplier reaches the preset termination value. If the number of iterations reaches the preset number of iterations or the range of change of the Lagrange multiplier reaches the preset termination value, the preset iteration termination condition is met. If the number of iterations does not reach the preset number of iterations and the range of change of the Lagrange multiplier does not reach the preset termination value, the preset iteration termination condition is not met.

6. The distributed collaborative optimization method based on flexible resources according to claim 4, It is characterized in that Before receiving the first intermediate parameters sent by all flexible resource terminals on the aggregator side, the method further includes: Establish a secure channel with the flexible resource end on the aggregator side.

7. A joint collaborative computing system based on privacy computing, It is characterized in that include: A dispatching end on the power coordination dispatching side and several flexible resource ends on the aggregator side, each of the flexible resource ends on the aggregator side is communicatively connected to the dispatching end on the power coordination dispatching side, and any two flexible resource ends on the aggregator side are communicatively connected to each other; The flexible resource end on the aggregator side generates a first calculation parameter through parameter initialization; The dispatching end of the electric power coordination dispatching side generates Lagrange multipliers and a mask matrix by initializing parameters, and converts the first global constraint parameter and the second global constraint parameter according to the mask matrix to obtain a first conversion parameter and a second conversion parameter; The flexible resource end on the aggregator side generates a mask value for this round of iteration, wherein the sum of the mask value generated for this round of iteration and the mask value generated by the flexible resource end on other aggregator sides is 0, a first preset function is constructed according to the local objective function and the first calculation parameter, and the local objective parameter corresponding to the minimum value of the first preset function is calculated, a first intermediate parameter is calculated according to the first calculation parameter, the local objective parameter corresponding to the minimum value of the first preset function, and the mask value for this round of iteration, and the first intermediate parameter is sent to the dispatching end on the power coordination dispatching side; The dispatching end on the power coordination dispatching side receives the first intermediate parameters sent by the flexible resource ends on all aggregator sides, adds all the first intermediate parameters to obtain the second intermediate parameters, and updates the Lagrange multiplier according to the Lagrange multiplier, the second intermediate parameter, the second conversion parameter and the preset gradient parameter; The dispatching end on the power coordination and dispatching side determines whether the preset iteration termination condition is met. If the preset iteration termination condition is met, the dispatching end on the power coordination and dispatching side terminates the calculation, and the flexible resource end on the aggregator side outputs the local target parameter corresponding to the value of the first preset function calculated in the last round of iteration process is minimum. If the preset iteration termination condition is not met, the dispatching end on the power coordination and dispatching side updates the first calculation parameter according to the first conversion parameter and the updated Lagrange multiplier, sends the updated first calculation parameter to the flexible resource end on the aggregator side, and enters the next round of iteration, wherein each round of iteration process includes the process in which the flexible resource end on the aggregator side generates a mask value for this round of iteration and the dispatching end on the power coordination and dispatching side sends the updated first calculation parameter to the flexible resource end on the aggregator side.

8. A flexible resource terminal on the aggregator side, It is characterized in that include: A first initialization module, used to generate a first calculation parameter by parameter initialization; A mask generation module, used for generating a mask value for this round of iteration, wherein the sum of the mask value generated for this round of iteration and the mask value generated by the flexible resource end of other aggregator side is 0; A minimum value calculation module, used to construct a first preset function according to the local objective function and the first calculation parameter, and calculate the local objective parameter corresponding to the minimum value of the first preset function; A first parameter calculation module, used to calculate a first intermediate parameter according to a first calculation parameter, a local target parameter corresponding to the minimum value of the first preset function, and a mask value of this round of iteration, and send the first intermediate parameter to a dispatching end on the power coordination dispatching side; A first iteration module, used to determine whether to perform the next round of iteration based on the updated first calculation parameter according to whether the dispatching end of the power coordination and dispatching side returns the updated first calculation parameter, wherein each round of iteration process includes the process of generating the mask value of the iteration to sending the first intermediate parameter to the dispatching end of the power coordination and dispatching side; The output module is used to output the local target parameter corresponding to the minimum value of the first preset function calculated in the last round of iteration process.

9. A dispatching terminal on the power coordination dispatching side, It is characterized in that include: A second initialization module, used for generating Lagrange multipliers and a mask matrix by parameter initialization, and converting the first global constraint parameter and the second global constraint parameter according to the mask matrix to obtain a first conversion parameter and a second conversion parameter; A second parameter calculation module, used for receiving the first intermediate parameters sent by all flexible resource terminals on the aggregator side, and adding all the first intermediate parameters to obtain the second intermediate parameter; A multiplier updating module, used for updating the Lagrange multiplier according to the Lagrange multiplier, the second intermediate parameter, the second conversion parameter and a preset gradient parameter; The second iteration module is used to determine whether the preset iteration termination condition is met. If the preset iteration termination condition is met, the calculation is terminated. If the preset iteration termination condition is not met, the first calculation parameter is updated according to the first conversion parameter and the updated Lagrange multiplier, and the updated first calculation parameter is sent to the flexible resource end on the aggregator side, and the next round of iteration is entered, wherein each round of iteration process includes the process of receiving the first intermediate parameter sent by the flexible resource end on the aggregator side and sending the updated first calculation parameter to the flexible resource end on the aggregator side.

10. A computer-readable storage medium, It is characterized in that The computer-readable storage medium stores computer instructions, and the computer instructions are used to enable the computer to execute the distributed collaborative optimization method based on flexible resources according to any one of claims 1 to 6.

Images