Resource configuration method with privacy protection function
By transforming the resource allocation problem into an augmented Lagrangian function and introducing an additive homomorphic encryption mechanism, the slow convergence speed of the gradient descent method and the impact of privacy protection on accuracy are solved, thus achieving efficient and secure resource allocation optimization.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- NANJING FORESTRY UNIV
- Filing Date
- 2026-01-28
- Publication Date
- 2026-05-15
AI Technical Summary
In existing technologies, gradient descent methods converge slowly in resource allocation problems, especially under large-scale or highly constrained conditions, and privacy protection methods affect the accuracy of optimization results. There is a lack of effective means to improve convergence speed and privacy protection.
The resource allocation problem is transformed into an augmented Lagrangian function using the primal dual method. An additive homomorphic encryption mechanism is used for privacy protection. A second-order dynamic update mechanism is used to improve the convergence speed. The Paillier method is used to generate keys for encryption and decryption to ensure privacy and security.
It improves the computational efficiency and privacy of resource allocation optimization, reduces the number of iterations, ensures the accuracy and security of optimization results, and has good prospects for engineering applications.
Smart Images
Figure CN122053379A_ABST
Abstract
Description
Technical Field
[0001] This invention belongs to the field of resource allocation technology, and in particular relates to a resource allocation method with privacy protection function. Background Technology
[0002] With the rapid development of smart grids, distributed energy systems, and edge computing networks, resource allocation problems are widespread in these systems. For example, in smart grids, it is necessary to coordinate the output of multiple power generation units, energy storage units, and load units to meet system power balance and operational constraints; in edge computing systems, it is necessary to rationally allocate computing resources among multiple computing nodes to meet quality of service requirements.
[0003] For the aforementioned resource allocation problem, the most commonly used solution method in existing technologies is the iterative optimization algorithm based on gradient descent. Gradient descent updates decision variables stepwise along the negative gradient of the objective function, offering advantages such as simplicity and low computational complexity, thus gaining widespread application in engineering practice. However, gradient descent is essentially a first-order optimization method, and its convergence speed often decreases significantly as it approaches the optimal solution. This is particularly problematic in applications with large problem sizes, numerous constraints, or high real-time requirements, where it can lead to issues such as numerous iterations and slow convergence.
[0004] To improve the performance of solving constrained optimization problems, the primal-dual method addresses constraints by simultaneously updating both primal and dual variables, theoretically guaranteeing feasibility and optimality. However, current research on the primal-dual method primarily focuses on its first-order form, where the update rules still rely on gradient information, resulting in limited improvements in convergence speed. For resource allocation problems, effective techniques for further improving the convergence speed of the primal-dual method while ensuring algorithm stability and implementability remain lacking.
[0005] Furthermore, in actual resource allocation processes, the decision variables and cost parameters of each participating entity often contain sensitive information. If this information is directly exposed during the solution process, it may lead to privacy leaks or security risks. Existing technologies typically employ methods such as noise injection or differential privacy for privacy protection. However, these methods inevitably introduce random perturbations, thereby affecting the accuracy of the optimization results and causing deviations between the solution and the optimal solution to the original problem. Summary of the Invention
[0006] The purpose of this invention is to provide a resource allocation method with privacy protection. It improves the optimization efficiency of smart grid resource allocation by processing system constraints through the primal dual method and achieves privacy protection by introducing an encryption mechanism with additive homomorphic properties.
[0007] To achieve the above objectives, the present invention is implemented using the following technical solution.
[0008] On one hand, the present invention provides a resource allocation method with privacy protection functions, including:
[0009] To obtain an optimization model for smart grid resource allocation with the goal of achieving optimal smart grid resource allocation;
[0010] Based on the aforementioned smart grid resource allocation optimization model, a resource optimization allocation intelligent agent is constructed, and the smart grid resource allocation optimization problem is transformed into an equivalent augmented Lagrangian function through the primal dual method.
[0011] The state variables of the resource optimization agent are encrypted using an encryption mechanism with additive homomorphic properties;
[0012] Based on the augmented Lagrangian function, the gradient of the augmented Lagrangian function is calculated using the state variables of the decrypted resource optimization agent;
[0013] Based on the preset resource allocation optimization objective and constraint domain, the state variables of the resource optimization agent are updated through gradient iteration of the augmented Lagrangian function;
[0014] The resource allocation optimization problem is solved by using an iteratively updated resource optimization agent to obtain the optimal resource allocation parameters for the smart grid.
[0015] Optionally, the encryption of the state variables of the resource optimization agent using an encryption mechanism with additive homomorphic properties includes:
[0016] The Paillier method is used to generate the key for the state variables of the resource optimization agent;
[0017] The state variables of the resource optimization and allocation agent are rounded down to obtain real-valued state variables;
[0018] Based on the real number state variables, the state variables of the resource optimization and allocation agent are obtained by encrypting them with a key.
[0019] The key is also used to decrypt the state variables of the encrypted resource optimization agent to obtain the decrypted state variables of the resource optimization agent.
[0020] By employing the Paillier method to protect the privacy of resource optimization agents, efficient collaborative computation of state variables is achieved while protecting data privacy, thereby enhancing privacy in the process of smart grid resource allocation optimization.
[0021] Optionally, the key for generating the state variables of the resource optimization agent using the Paillier method includes a public key and a private key generated based on two prime numbers respectively;
[0022] The expression for generating a public key based on two prime numbers is:
[0023] (1)
[0024] in, and Let each represent a prime number. The public key representing the secret key;
[0025] The expression for generating a private key based on two prime numbers is:
[0026] (2)
[0027] (3)
[0028] The private key of the secret key is represented as: , and They are two large prime numbers that satisfy specific conditions. To find the least common multiple, the function is... Specifically, it is expressed as follows: ,in For function The independent variable, This involves modulo operations. Generating a key using two randomly selected prime numbers enhances security during the smart grid resource allocation optimization process.
[0029] Optionally, the formula for rounding the state variables of the resource optimization agent is as follows:
[0030] (4)
[0031] in, The first consideration for making decisions by an intelligent agent for optimizing resource allocation The state variables of each resource unit To round up, To determine the number of decimal places to retain, for The corresponding real-valued state variables.
[0032] By rounding the state variables of the resource optimization agent, the computational difficulty in the encryption process is reduced, and the computational efficiency of encrypting state variables is improved.
[0033] Optionally, the calculation formula for encryption using the key based on the real-valued state variable is as follows:
[0034] (5)
[0035] in, for Encrypted state variables, For the first Random numbers generated by the resource optimization agent for each resource unit. and The greatest common divisor is 1. For modulo operation, The public key represents the secret key.
[0036] Optionally, the calculation formula for decrypting the encrypted state variables of the resource optimization and configuration agent using a key is as follows:
[0037] (6)
[0038] (7)
[0039] in, The private key of the secret key is represented as the total resource requirement or overall scheduling target of the distributed system. , and They are two large prime numbers that satisfy specific conditions. The function is specifically represented as: ,in for The independent variable of a function For modulo operation, For intermediate parameters, the calculation formula is as follows: , for Encrypted state variables, for The corresponding real-number state variables, The first consideration for making decisions by an intelligent agent for optimizing resource allocation The state variables of each resource unit The public key represents the private key. To determine the number of decimal places to retain, The total number of resource optimization agents is given, and the state variables of the decrypted resource optimization agents are: .
[0040] By using an asymmetric architecture that encrypts with a public key and decrypts with a private key, resource optimization and allocation are achieved, protecting the privacy of intelligent agents and reducing the cost of key management and the risk of key leakage.
[0041] Optionally, the specific expression of the smart grid resource allocation optimization model is as follows:
[0042] (8)
[0043] in, The first consideration for making decisions by an intelligent agent for optimizing resource allocation The state variables of each resource unit represent the resource output or allocation of that resource unit; For the first The convex cost function of each intelligent agent corresponds to a resource unit, which is used to characterize the operating cost or scheduling cost of the resource unit. For the first The range of values that a resource unit can produce or allocate; This refers to the total resource requirements or overall scheduling target of a distributed system. To optimize the allocation of resources to the total number of intelligent agents; The total operating cost of the system; the resource units include power generation units, energy storage units, communication nodes or computing nodes.
[0044] The smart grid resource allocation optimization model improves the accuracy of resource allocation by optimizing corresponding resource units through intelligent agents.
[0045] Optionally, the expression for the augmented Lagrange function is:
[0046] (9)
[0047] The objective expression for resource allocation optimization is:
[0048] (10)
[0049] in, Optimize the vector representation of the agent for all resources. They are Lagrange multipliers; This is the augmentation factor; This refers to the total resource requirements or overall scheduling target of a distributed system. To optimize the allocation of resources to the total number of intelligent agents; The first consideration for making decisions by an intelligent agent for optimizing resource allocation The state variables of each resource unit For the first The convex cost function of each intelligent agent corresponds to a resource unit, which is used to characterize the operating cost or scheduling cost of the resource unit. For the first The range of values that a resource unit can produce or allocate resources from. To augment the Lagrange function.
[0050] By introducing a second-order dynamic structure into the resource optimization problem of the smart grid using the augmented Lagrangian function, the convergence speed can be improved when optimizing the resource optimization agent.
[0051] Optionally, based on the augmented Lagrangian function, the gradients of the augmented Lagrangian function with respect to the state variables and Lagrange multipliers are calculated using the decrypted state variables of the resource optimization agent. The calculation formula is as follows:
[0052] (11)
[0053] in, To augment the Lagrange function for the th The state variables corresponding to the resource optimization and allocation agent of each resource unit gradient, To augment the Lagrange function for the th The state variables corresponding to the resource optimization and allocation agent of each resource unit The gradient.
[0054] By calculating the gradient of the augmented Lagrange function with respect to the state variables and Lagrange multipliers, the number of iterations of the resource optimization agent can be significantly reduced.
[0055] Optionally, the specific expression of the iteratively updated resource optimization configuration agent is:
[0056] (12)
[0057] in, From the real number field to the set Projection operation, For the number of iterations, and For the first During the nth iteration Auxiliary variables for the resource optimization allocation agent of each resource unit. Step size, For second-order parameters, For the first The decision-making process of the resource optimization agent in the next iteration considers the first... The state variables of each resource unit For the first Lagrange multipliers in the next iteration.
[0058] The resource optimization and allocation agent is a second-order primitive-dual agent, which enables the resource allocation state variables to have stronger dynamic adjustment capabilities during the iteration process. Through the iteratively updated resource optimization and allocation agent with auxiliary variables and second-order parameters, the current state information and historical evolution trends can be used simultaneously when calculating the optimization of smart grid resource allocation, thereby effectively accelerating the convergence process of the system to the optimal solution.
[0059] Compared with the prior art, the beneficial effects achieved by the present invention are as follows:
[0060] This invention transforms the smart grid resource allocation problem into an equivalent augmented Lagrangian function, introduces a second-order dynamic update mechanism, and improves the computational efficiency of the resource optimization agent. Through this resource optimization agent, the dynamic adjustment capability of the resource allocation state variables during the iteration process is enhanced. Updating the resource optimization agent using the gradient of the augmented Lagrangian function improves its convergence speed. Encryption of the resource optimization agent using an additive homomorphic encryption mechanism effectively avoids error accumulation problems. This invention has significant advantages over existing technologies in terms of convergence speed, stability, and privacy protection, and possesses promising engineering application prospects. Attached Figure Description
[0061] Figure 1 This is a schematic diagram of the solver results provided in Embodiment 3 of the present invention;
[0062] Figure 2 A schematic diagram of the solver results provided for Example 3 of the gradient descent method;
[0063] Figure 3 A schematic diagram of the solver results provided in Example 3 of the differential privacy method;
[0064] Figure 4 A schematic diagram of the solver results provided in Embodiment 3 of the present invention when the total resource requirement is set to 100;
[0065] Figure 5 A schematic diagram of the solver results provided in Example 3 of the gradient descent method when the total resource requirement is set to 100;
[0066] Figure 6 A schematic diagram of the solver results provided in Example 3 of the differential privacy method when the total resource requirement is set to 100;
[0067] Figure 7 This is a flowchart illustrating a resource allocation method with privacy protection features. Detailed Implementation
[0068] The technical solution of the present invention will be described in detail below with reference to the accompanying drawings and specific embodiments. It should be understood that the embodiments of the present invention and the specific features in the embodiments are detailed descriptions of the technical solution of the present invention, rather than limitations thereof. In the absence of conflict, the embodiments of the present invention and the technical features in the embodiments can be combined with each other.
[0069] The term "and / or" simply describes the relationship between related objects, indicating that three relationships can exist. For example, A and / or B can represent: A alone, A and B simultaneously, or B alone. Additionally, the character " / " generally indicates that the preceding and following related objects have an "or" relationship.
[0070] Example 1
[0071] This embodiment provides a resource configuration method with privacy protection features, such as... Figure 7 As shown, the specific steps include:
[0072] To obtain an optimization model for smart grid resource allocation with the goal of achieving optimal smart grid resource allocation.
[0073] Based on the aforementioned smart grid resource allocation optimization model, a resource optimization agent is constructed. By solving the resource optimization agent, the optimal resource allocation parameters can be obtained. The resource optimization agent is a second-order primal-dual agent, which enables the resource allocation state variables to have stronger dynamic adjustment capabilities during the iteration process. The smart grid resource allocation problem is transformed into an equivalent augmented Lagrangian function through the primal-dual method, introducing a second-order dynamic update mechanism and improving the computational efficiency of the resource optimization agent.
[0074] By using an encryption mechanism with additive homomorphic properties to encrypt the state variables of the resource optimization agent, the error accumulation problem caused by noise injection-based methods such as differential privacy is avoided in terms of optimization accuracy and convergence performance.
[0075] Based on the augmented Lagrangian function, the gradient of the augmented Lagrangian function is calculated using the state variables of the decrypted resource optimization agent;
[0076] Based on preset resource allocation optimization objectives and constraint domains, which include the optimization requirements of the smart grid and the actual value range of each state variable within the smart grid, the state variables of the resource optimization agent are updated iteratively using the gradient of the augmented Lagrangian function, thus reducing the number of iterations required for the resource optimization agent.
[0077] The resource allocation optimization problem is solved by using an iteratively updated resource optimization agent to obtain the optimal resource allocation parameters for the smart grid.
[0078] In summary, this embodiment has significant advantages over existing technologies in terms of convergence speed, stability, and privacy protection, and has promising prospects for engineering applications.
[0079] Example 2
[0080] Based on the same inventive concept as Embodiment 1, this embodiment provides a resource allocation method with privacy protection functions, the specific steps of which include:
[0081] Based on the aforementioned smart grid resource allocation problem, a resource optimization agent is constructed. The specific expression of the smart grid resource allocation optimization model is as follows:
[0082]
[0083] in, The first consideration for making decisions by an intelligent agent for optimizing resource allocation The state variables of each resource unit represent the resource output or allocation of that resource unit; For the first The convex cost function of each intelligent agent corresponds to a resource unit, which is used to characterize the operating cost or scheduling cost of the resource unit. For the first The range of values that a resource unit can produce or allocate; This refers to the total resource requirements or overall scheduling target of a distributed system. To optimize the allocation of resources to the total number of intelligent agents; This represents the total operating cost of the system.
[0084] The resource units include power generation units, energy storage units, communication nodes, or computing nodes. For the resource optimization and allocation problem of smart grids, intelligent agents optimize the corresponding resource units, thereby improving the accuracy of resource allocation optimization.
[0085] The smart grid resource allocation problem is transformed into an equivalent augmented Lagrangian function using the primal dual method. By introducing a second-order dynamic structure into the smart grid resource optimization problem using the augmented Lagrangian function, the convergence speed in optimizing the resource optimization agent can be improved.
[0086] The expression for the augmented Lagrange function is:
[0087]
[0088] By setting the resource allocation optimization objective, the optimal resource allocation of the smart grid can be accurately reflected. The expression for the resource allocation optimization objective is:
[0089]
[0090] in, Optimize the vector representation of the agent for all resources. They are Lagrange multipliers; This is the augmentation factor; This refers to the total resource requirements or overall scheduling target of a distributed system. To optimize the allocation of resources to the total number of intelligent agents; The first consideration for making decisions by an intelligent agent for optimizing resource allocation The state variables of each resource unit For the first The convex cost function of each intelligent agent corresponds to a resource unit, which is used to characterize the operating cost or scheduling cost of the resource unit. For the first The range of values that a resource unit can produce or allocate resources from. To augment the Lagrange function.
[0091] The state variables of the resource optimization agent are encrypted using an encryption mechanism with additive homomorphic properties; compared to existing methods such as differential privacy based on noise injection, this avoids the error accumulation problem that affects optimization accuracy and convergence performance. Specific steps include:
[0092] The Paillier method is used to generate the key for the state variables of the resource optimization agent;
[0093] Generate a public key and a private key for the secret key based on the two obtained random prime numbers;
[0094] The expression for generating a public key based on two prime numbers is:
[0095]
[0096] in, and Let each represent a prime number. The public key representing the secret key;
[0097] The expression for generating a private key based on two prime numbers is:
[0098]
[0099]
[0100] The private key of the secret key is represented as: , and They are two large prime numbers that satisfy specific conditions. To find the least common multiple, the function is... Specifically, it is expressed as follows: ,in For function The independent variable, This involves modulo operations. Generating a key using two randomly selected prime numbers enhances security during the smart grid resource allocation optimization process.
[0101] The state variables of the resource optimization agent are rounded to obtain real-valued state variables. The specific calculation formula is as follows:
[0102]
[0103] in, The first consideration for making decisions by an intelligent agent for optimizing resource allocation The state variables of each resource unit The rounding up symbol, To determine the number of decimal places to retain, for The corresponding real-valued state variables.
[0104] By rounding the state variables of the resource optimization agent, the computational difficulty in the encryption process is reduced, and the computational efficiency of encrypting state variables is improved.
[0105] Based on the real-valued state variables, the encrypted state variables of the resource optimization agent are obtained by encrypting them with a key. The expression is as follows:
[0106]
[0107] in, for Encrypted state variables, For the first Random numbers generated by the resource optimization agent for each resource unit. and The greatest common divisor is 1. For modulo operation, The public key represents the secret key.
[0108] The encrypted state variables of the resource optimization agent are decrypted using a key to obtain the decrypted state variables of the resource optimization agent.
[0109] The calculation formula for decrypting the state variables of the encrypted resource optimization agent using a key is as follows:
[0110]
[0111]
[0112] in, The private key of the secret key is represented as the total resource requirement or overall scheduling target of the distributed system. , and They are two large prime numbers that satisfy specific conditions. The function is specifically represented as: ,in for The independent variable of a function For modulo operation, For intermediate parameters, the calculation formula is as follows: , for Encrypted state variables, for The corresponding real-number state variables, The first consideration for making decisions by an intelligent agent for optimizing resource allocation The state variables of each resource unit The public key represents the private key. To determine the number of decimal places to retain, The total number of resource optimization agents is given, and the state variables of the decrypted resource optimization agents are: .
[0113] By employing the Paillier method to protect the privacy of resource optimization agents, efficient collaborative computation of state variables is achieved while protecting data privacy, thereby enhancing privacy in the process of smart grid resource allocation optimization.
[0114] Based on the augmented Lagrangian function, the gradient of the augmented Lagrangian function is calculated using the state variables of the decrypted resource optimization agent. The specific expression is as follows:
[0115]
[0116] in, To augment the Lagrange function for the th The state variables of the resource units corresponding to each resource optimization agent gradient, To augment the Lagrange function for the th The state variables of the resource units corresponding to each resource optimization agent The gradient.
[0117] By calculating the gradient of the augmented Lagrange function with respect to the state variables and Lagrange multipliers, the number of iterations of the resource optimization agent can be significantly reduced.
[0118] Based on a preset resource allocation optimization objective and constraint domain, the state variables of the resource optimization agent are updated iteratively through the gradient of the augmented Lagrangian function. The resource optimization agent is a second-order primal-dual agent, enabling stronger dynamic adjustment capabilities of the resource allocation state variables during iteration. Through the iteratively updated resource optimization agent with auxiliary variables and second-order parameters, the current state information and historical evolution trends can be simultaneously utilized when calculating smart grid resource allocation optimization, thereby effectively accelerating the convergence process towards the optimal solution. The specific expression of the resource optimization agent is:
[0119] ;
[0120] in, From the real number field to the set Projection operation, For the number of iterations, and For the first During the nth iteration Auxiliary variables for the resource optimization allocation agent of each resource unit. Step size, For second-order parameters, For the first The decision-making process of the resource optimization agent in the next iteration considers the first... The state variables of each resource unit For the first Lagrange multipliers in the next iteration.
[0121] The optimal resource allocation parameters for the smart grid are obtained by solving the iteratively updated resource optimization agent. This invention can also be applied to resource allocation optimization problems in distributed systems such as distributed energy systems, communication networks, and edge computing systems.
[0122] Example 3
[0123] Based on the same inventive concept as Embodiment 2, this embodiment introduces an application case of a resource allocation method with privacy protection function in a distributed generation power allocation scenario.
[0124] The distributed generation power allocation scenario involves a distributed system comprising five resource units, which can be represented as generation units, energy storage units, or controllable load units in a smart grid. These resource units interact with each other via a communication network. The system needs to collaboratively allocate the output of each resource unit while satisfying total demand constraints, in order to minimize overall operating costs. The corresponding resource allocation optimization problem can be expressed as:
[0125]
[0126] in The coefficients within are shown in the table:
[0127]
[0128] In this scenario, the cost function coefficients and local state variables of each resource unit This information is sensitive and should not be directly disclosed during communication. This invention introduces a privacy-preserving resource optimization and allocation intelligent agent, namely a second-order primal dual solver, to achieve a distributed solution to the aforementioned resource allocation problem while ensuring that the local information of each resource unit is not leaked. The simulation results of using the described resource optimization and allocation intelligent agent to solve the resource allocation problem are as follows: Figure 1 As shown. To compare the convergence performance of different algorithms, the resource optimization allocation agent of this invention, i.e., the second-order primal dual solver, was compared with a solver based on gradient descent. The simulation results of the gradient descent solver are shown below. Figure 2 As shown in the comparison results, the resource optimization allocation agent, i.e., the second-order primal dual solver, can reach a stable convergence state after approximately 10,000 iterations, while the gradient descent solver requires approximately 40,000 iterations to achieve convergence. This indicates that the proposed second-order primal dual solver has a significant advantage in convergence speed. This advantage mainly stems from the auxiliary variables and second-order parameters introduced by the augmented Lagrangian function. These auxiliary variables and second-order parameters constitute the performance improvement of the second-order system in terms of dynamic adjustment and error correction.
[0129] Furthermore, to verify the impact of the proposed privacy protection mechanism on the solution accuracy, the resource optimization agent, i.e., the second-order primal dual solver, was compared with a resource allocation solver based on differential privacy technology. The simulation results are as follows: Figure 3 As shown in the figure. The results show that the solver of this invention can still converge to the vicinity of the optimal solution of the original problem after introducing a privacy protection mechanism. In contrast, the comparison method based on differential privacy introduces random noise during the iteration process, resulting in a significant deviation between its final convergence result and the optimal solution. In comparison, the homomorphic encryption method used in this invention does not introduce additional noise while ensuring privacy security, thus effectively reducing the impact of privacy protection on optimization accuracy. This further demonstrates the technical advantage of this invention in balancing convergence performance and privacy protection.
[0130] Further considering a large-scale distributed resource allocation scenario, taking a distributed system containing 100 resource units as an example, the performance of the second-order primal dual solver with privacy protection described in this invention is verified. The resource unit can be represented as a power generation unit in a smart grid, a node in a communication network, or a computing node in an edge computing system, with each resource unit interacting with information through the communication network.
[0131] In this embodiment, the local cost function parameters corresponding to each resource unit are randomly generated to reflect the differences in operating characteristics or scheduling costs among different resource units. The total system resource requirement is set to a=100, and the resource allocation for each resource unit satisfies the constraints. In the aforementioned distributed network, five representative resource units are selected, and their state variables are iteratively updated under the action of the second-order primal dual solver to illustrate the convergence behavior and stability characteristics of the proposed method in a large-scale distributed scenario. Simulations are performed using the solver proposed in this invention, a solver based on gradient descent, and a solver based on differential privacy, respectively. The results are as follows: Figure 4 , Figure 5 and Figure 6 As shown.
[0132] The simulation results show that even with a significant increase in the number of agents and a larger system scale, the resource optimization agent, i.e., the second-order primal dual solver, still maintains good convergence speed and stability, consistent with the results in Example 1. Furthermore, after introducing a privacy protection mechanism, the solver of this invention significantly outperforms the comparative method based on differential privacy in terms of convergence accuracy. These results demonstrate that the second-order primal dual solver proposed in this invention can balance privacy protection with convergence performance in large-scale distributed resource allocation scenarios, further verifying the effectiveness and scalability of the method. This invention has significant advantages over existing technologies in terms of convergence speed, stability, and privacy protection, and possesses promising engineering application prospects.
[0133] Those skilled in the art will understand that embodiments of the present invention can be provided as methods, systems, or computer program products. Therefore, the present invention can take the form of a completely hardware embodiment, a completely software embodiment, or an embodiment combining software and hardware aspects. Furthermore, the present invention can take the form of a computer program product embodied on one or more computer-usable storage media (including, but not limited to, disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.
[0134] This invention is described with reference to flowchart illustrations and / or block diagrams of methods, apparatus (systems), and computer program products according to embodiments of the invention. It will be understood that each block of the flowchart illustrations and / or block diagrams, and combinations of blocks in the flowchart illustrations and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, special-purpose computer, embedded processor, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, generate instructions for implementing the flowchart illustrations and / or block diagrams. Figure 1 One or more processes and / or boxes Figure 1 A device that provides the functions specified in one or more boxes.
[0135] These computer program instructions may also be stored in a computer-readable storage medium that can direct a computer or other programmable data processing device to function in a particular manner, such that the instructions stored in the computer-readable storage medium produce an article of manufacture including instruction means, which are implemented in a process Figure 1 One or more processes and / or boxes Figure 1 The function specified in one or more boxes.
[0136] These computer program instructions may also be loaded onto a computer or other programmable data processing equipment to cause a series of operational steps to be performed on the computer or other programmable equipment to produce a computer-implemented process, thereby providing instructions that execute on the computer or other programmable equipment for implementing the process. Figure 1 One or more processes and / or boxes Figure 1 The steps of the function specified in one or more boxes.
[0137] The embodiments of the present invention have been described above with reference to the accompanying drawings. However, the present invention is not limited to the specific embodiments described above. The specific embodiments described above are merely illustrative and not restrictive. Those skilled in the art can make many other forms under the guidance of the present invention without departing from the spirit and scope of the claims. All of these forms are within the protection scope of the present invention.
Claims
1. A resource allocation method with privacy protection features, characterized in that, include: To obtain an optimization model for smart grid resource allocation with the goal of achieving optimal smart grid resource allocation; Based on the aforementioned smart grid resource allocation optimization model, a resource optimization allocation intelligent agent is constructed, and the smart grid resource allocation optimization problem is transformed into an equivalent augmented Lagrangian function through the primal dual method. The state variables of the resource optimization agent are encrypted using an encryption mechanism with additive homomorphic properties; Based on the augmented Lagrangian function, the gradient of the augmented Lagrangian function is calculated using the state variables of the decrypted resource optimization agent; Based on the preset resource allocation optimization objective and constraint domain, the state variables of the resource optimization agent are updated through gradient iteration of the augmented Lagrangian function; The resource allocation optimization problem is solved by using an iteratively updated resource optimization agent to obtain the optimal resource allocation parameters for the smart grid.
2. The resource allocation method with privacy protection function according to claim 1, characterized in that, The encryption mechanism with additive homomorphic properties for encrypting the state variables of the resource optimization agent includes: The Paillier method is used to generate the key for the state variables of the resource optimization agent; The state variables of the resource optimization and allocation agent are rounded down to obtain real-valued state variables; Based on the real number state variables, the state variables of the resource optimization and allocation agent are obtained by encrypting them with a key. The key is also used to decrypt the state variables of the encrypted resource optimization agent to obtain the decrypted state variables of the resource optimization agent.
3. The resource allocation method with privacy protection function according to claim 2, characterized in that, The key for generating the state variables of the resource optimization agent using the Paillier method includes a public key and a private key generated based on two prime numbers. The expression for generating a public key based on two prime numbers is: （1） in, and Let each represent a prime number. The public key representing the secret key; The expression for generating a private key based on two prime numbers is: （2） （3） The private key of the secret key is represented as: , and They are two large prime numbers that satisfy specific conditions. To find the least common multiple, the function is... Specifically, it is expressed as follows: ,in For function The independent variable, This is a modulo operation.
4. The resource allocation method with privacy protection function according to claim 2, characterized in that, The formula for rounding the state variables of the resource optimization agent is as follows: （4） in, The first consideration for making decisions by an intelligent agent for optimizing resource allocation The state variables of each resource unit To round up, To determine the number of decimal places to retain, for The corresponding real-valued state variables.
5. The resource allocation method with privacy protection function according to claim 4, characterized in that, Based on the aforementioned real-valued state variables, the calculation formula for encryption using the key is as follows: （5） in, for Encrypted state variables, For the first Random numbers generated by the resource optimization agent for each resource unit. and The greatest common divisor is 1. For modulo operation, The public key represents the secret key.
6. The resource allocation method with privacy protection function according to claim 2, characterized in that, The formula for decrypting the state variables of the encrypted resource optimization agent using a key is as follows: （6） （7） in, The private key of the secret key is represented as the total resource requirement or overall scheduling target of the distributed system. , and They are two large prime numbers that satisfy specific conditions. The function is specifically represented as: ,in for The independent variable of a function For modulo operation, For intermediate parameters, the calculation formula is as follows: , for Encrypted state variables, for The corresponding real-number state variables, The first consideration for making decisions by an intelligent agent for optimizing resource allocation The state variables of each resource unit The public key represents the private key. To determine the number of decimal places to retain, The total number of resource optimization agents is given, and the state variables of the decrypted resource optimization agents are: .
7. The resource allocation method with privacy protection function according to claim 1, characterized in that, The specific expression of the smart grid resource allocation optimization model is as follows: （8） in, The first consideration for making decisions by an intelligent agent for optimizing resource allocation The state variables of each resource unit represent the resource output or allocation of that resource unit; For the first The convex cost function of each intelligent agent corresponds to a resource unit, which is used to characterize the operating cost or scheduling cost of the resource unit. For the first The range of values that a resource unit can produce or allocate; This refers to the total resource requirements or overall scheduling target of a distributed system. To optimize the allocation of resources to the total number of intelligent agents; The total operating cost of the system; the resource units include power generation units, energy storage units, communication nodes or computing nodes.
8. The resource allocation method with privacy protection function according to claim 1, characterized in that, The expression for the augmented Lagrange function is: （9） The objective expression for resource allocation optimization is: （10） in, Optimize the vector representation of the agent for all resources. They are Lagrange multipliers; This is the augmentation factor; This refers to the total resource requirements or overall scheduling target of a distributed system. To optimize the allocation of resources to the total number of intelligent agents; The first consideration for making decisions by an intelligent agent for optimizing resource allocation The state variables of each resource unit For the first The convex cost function of each intelligent agent corresponds to a resource unit, which is used to characterize the operating cost or scheduling cost of the resource unit. For the first The range of values that a resource unit can produce or allocate resources from. To augment the Lagrange function.
9. The resource allocation method with privacy protection function according to claim 8, characterized in that, Based on the augmented Lagrange function, the gradients of the augmented Lagrange function with respect to the state variables and Lagrange multipliers are calculated using the state variables of the decrypted resource optimization agent. The calculation formula is as follows: （11） in, To augment the Lagrange function for the th The state variables corresponding to the resource optimization and allocation agent of each resource unit gradient, To augment the Lagrange function for the th The state variables corresponding to the resource optimization and allocation agent of each resource unit The gradient.
10. The resource allocation method with privacy protection function according to claim 9, characterized in that, The specific expression of the resource optimization and allocation agent after the iterative update is: ； （12） in, From the real number field to the set Projection operation, For the number of iterations, and For the first During the nth iteration Auxiliary variables for the resource optimization allocation agent of each resource unit. Step size, For second-order parameters, For the first The decision-making process of the resource optimization agent in the next iteration considers the first... The state variables of each resource unit For the first Lagrange multipliers in the next iteration.