Distributed reactive power resource coordination method based on cooperative game and adaptive learning

Abstract

The invention discloses a cooperative game and adaptive learning-based distributed reactive power resource coordination method, which comprises the following steps of: (1) collecting voltage, current and reactive power contribution data of equipment in real time through a sensor and a monitoring system; (2) constructing a game and income balance model based on voltage, current and reactive power contribution data, and training the game and income balance model; (3) optimizing the model by utilizing meta learning; (4) generating a distribution scheme of reactive power resources; according to the invention, the overall coordination of the system is improved.

Description

Technical Field

[0001] The present invention relates to the technical field of power systems, and particularly to a distributed reactive power resource coordination method based on cooperative game and adaptive learning. Background Art

[0002] In modern power systems, the regulation of reactive power is crucial for maintaining system voltage stability and improving power supply quality. With the increasing access of distributed energy and load fluctuations, traditional centralized reactive power regulation strategies have become difficult to meet dynamic power demands. In the prior art, various reactive power resources (such as inverters, SVG, capacitors) operate independently, lacking unified coordination, which easily leads to problems such as resource waste or uneven interests. Summary of the Invention

[0003] Object of the Invention: The object of the present invention is to provide a distributed reactive power resource coordination method based on cooperative game and adaptive learning, so as to solve the problems that traditional centralized reactive power regulation strategies lack unified coordination and easily lead to resource waste or uneven interests.

[0004] Technical Solution: A distributed reactive power resource coordination method based on cooperative game and adaptive learning according to the present invention includes the following steps:

[0005] (1) Through sensors and monitoring systems, collect voltage, current, and reactive power contribution data of devices in real time;

[0006] (2) Based on the voltage, current, and reactive power contribution data, construct a game and revenue equilibrium model and train it;

[0007] (3) Use meta-learning to optimize the model;

[0008] (4) Generate an allocation scheme for reactive power resources.

[0009] Further, in step (1), the devices include: inverters, SVG, capacitors.

[0010] Further, step (2) is specifically as follows:

[0011] (21) Use a cooperative game model to describe the allocation problem of reactive power resources;

[0012] (22) Use the Shapley value to calculate the revenue allocation;

[0013] (23) Set a deviation threshold, and use deviation feedback to adjust the parameters in the game model to optimize the revenue allocation;

[0014] (24) Calculate the gain allocation based on contributions.

[0015] Further, the formula for step (22) is as follows:

[0016]

[0017] Among them, φi is the payoff distribution of participant i, N is the set of all participants, S is a subset, and v(S) is the value function of the subset.

[0018] Further, step (23) is specifically as follows: Set the deviation threshold ∈, when |D i | > ∈, adjust the allocation strategy of reactive power resources;

[0019] Di = Pi - Fi

[0020] Among them, Di is the payment deviation of participant i, Pi is the actual payoff, and Fi is the expected fair payoff.

[0021] Further, in step (23), the parameter formula in the game model is adjusted using deviation feedback as follows:

[0022]

[0023] Among them, represents the contribution degree after feedback, represents the contribution degree before feedback, and the weight α of the participant is dynamically adjusted with the adjustment step size.

[0024] Further, step (24) is specifically as follows: First, calculate the contribution degree, and the formula is as follows:

[0025]

[0026] Among them, Ci is the contribution degree of participant i, Qi is its reactive power contribution, and ∑j∈NQj is the total contribution of all participants.

[0027] Secondly, calculate the gain allocation, and the formula is as follows:

[0028] Gi = Ci·R

[0029] Among them, Gi is the gain of participant i, and R is the total system payoff.

[0030] Further, the model is optimized using the MAML algorithm.

[0031] An electronic device according to the present invention includes a memory, a processor, and a computer program stored on the memory and executable on the processor. When the computer program is loaded into the processor, it implements a distributed reactive power resource coordination method based on cooperative game and adaptive learning as described in any one of the above.

[0032] A storage medium according to the present invention, wherein the storage medium stores a computer program, and when the computer program is executed by a processor, it implements any one of the distributed reactive power resource coordination methods based on cooperative game and adaptive learning.

[0033] Beneficial effects: Compared with the prior art, the present invention has the following remarkable advantages: introducing the Shapley value and payment deviation detection mechanism in cooperative game to achieve fair distribution of interests of all parties; ensuring the fairness and rationality of reactive power resource allocation by dynamically calculating the marginal contribution of each party; through real-time monitoring and adjustment of payment deviation, the agent can quickly identify and correct the imbalance in interest distribution, thus avoiding resource waste and conflicts among users and improving the overall coordination of the system. BRIEF DESCRIPTION OF THE DRAWINGS

[0034] Figure 1 is the flow chart of the present invention;

[0035] Figure 2 is the flow chart of the game model of the present invention; DETAILED DESCRIPTION OF THE INVENTION

[0036] The technical solution of the present invention will be further described below with reference to the accompanying drawings.

[0037] As Figure 1 shown, an embodiment of the present invention provides a distributed reactive power resource coordination method based on cooperative game and adaptive learning, including the following steps:

[0038] (1) Through sensors and monitoring systems, real-time collect voltage, current, and reactive power contribution data of devices; the devices include: inverters, SVG, and capacitors.

[0039] (2) Build a game and revenue equilibrium model based on voltage, current, and reactive power contribution data and train it; specifically as follows:

[0040] (21) Use a cooperative game model to describe the allocation problem of reactive power resources;

[0041] (22) Use the Shapley value to calculate the revenue distribution; the formula is as follows:

[0042]

[0043] where φi is the revenue distribution of participant i, N is the set of all participants, S is a subset, and v(S) is the value function of the subset.

[0044] (23) Set a deviation threshold, and use deviation feedback to adjust the parameters in the game model to optimize the revenue distribution; specifically as follows: set the deviation threshold ∈, when |D i | > ∈, adjust the allocation strategy of reactive power resources;

[0045] Di = Pi - Fi

[0046] Among them, Di is the payment deviation of participant i, Pi is the actual benefit, and Fi is the expected fair benefit.

[0047] The parameter formula in the game model is adjusted using deviation feedback as follows:

[0048]

[0049] Among them, represents the contribution degree after feedback, represents the contribution degree before feedback, and the weight α of the participant is dynamically adjusted with the adjustment step size.

[0050] (24) Calculate the gain allocation based on contribution. Specifically as follows: First, calculate the contribution degree, and the formula is as follows:

[0051]

[0052] Among them, Ci is the contribution degree of participant i, Qi is its reactive power contribution, and ∑j∈NQj is the total contribution of all participants.

[0053] Secondly, calculate the gain allocation, and the formula is as follows:

[0054] Gi = Ci · R

[0055] Among them, Gi is the gain of participant i, and R is the total benefit of the system.

[0056] (3) Use the MAML algorithm to optimize the model; specifically including the following steps:

[0057] (31) Task definition and data preparation: In reactive power resource coordination, each task can be defined as the reactive power resource allocation problem under different power system states. The data of each task includes: voltage, current, reactive power contribution data: collected in real time through sensors and monitoring systems; equipment operation status: including operation parameters of equipment such as inverters, SVG, and capacitors; system reactive power demand: the reactive power demand status of the current power system. The dataset of each task can be divided into a training set and a validation set, which are used for inner-loop update and outer-loop update respectively.

[0058] (32) Model Initialization: The goal of MAML is to find an initial model parameter θ such that when faced with a new task, this parameter can achieve good performance with only a small number of gradient updates. The initial model parameter θ can be obtained by training on multiple historical tasks. Model Structure: The model can be a revenue equilibrium model based on cooperative games, with input being voltage, current, and reactive power contribution data, and output being the allocation scheme of reactive resources. Initial Parameter θ: Randomly initialized or pre-trained based on historical data.

[0059] (33) Inner Loop Update: The goal of the inner loop update is to perform a small number of gradient updates on the model for each task to obtain task-specific model parameter θ'. The specific steps are as follows:

[0060] (331) Task Sampling: Sample a task T_i from the task distribution. The data of task T_i includes a training set and a validation set.

[0061] (332) Calculate Loss: Use the training data of task T_i to calculate the loss function L_Ti(θ). The loss function can be defined as the deviation of reactive resource allocation or the negative value of system revenue.

[0062] (333) Gradient Update: Perform a gradient update on the model parameter θ to obtain the task-specific parameter θ'_i;

[0063] (34) Outer Loop Update: The goal of the outer loop update is to optimize the initial model parameter θ so that it performs well on all tasks. The specific steps are as follows:

[0064] (341) Calculate Task-Specific Parameters: For each task T_i, use the inner loop update to obtain the task-specific parameter θ'_i.

[0065] (342) Calculate Validation Loss: Use the validation data of task T_i to calculate the loss function L_Ti(θ'_i).

[0066] (343) Update Initial Parameters: Update the initial model parameter θ to minimize the validation loss on all tasks.

[0067] (35) Model Optimization and Adaptation: Through multiple inner loop and outer loop updates, MAML can find an initial model parameter θ such that when faced with a new task, this parameter can achieve good performance with only a small number of gradient updates. In reactive resource coordination, the specific steps of model optimization are as follows:

[0068] (351) Model training: Sample multiple tasks from historical data, where each task corresponds to a different power system state and reactive power demand; Use the MAML algorithm to train the model to find the initial model parameters θ.

[0069] (352) Model adaptation: When the power system state changes, use the data of the new task to perform a small number of gradient updates on the model parameters θ to obtain task-specific parameters θ'; Use θ' to generate the allocation scheme of reactive power resources.

[0070] (353) Model evaluation: Evaluate the performance of the model on the new task to ensure the fairness and reasonableness of the reactive power resource allocation; If the performance does not meet the requirements, the model parameters can be further adjusted or retrained.

[0071] (4) Generate the allocation scheme of reactive power resources.

[0072] An embodiment of the present invention also provides an electronic device, including a memory, a processor, and a computer program stored on the memory and executable on the processor. When the computer program is loaded into the processor, it implements any one of the wireless sensing covert communications based on beacon signals.

[0073] An embodiment of the present invention also provides a storage medium, which stores a computer program. When the computer program is executed by a processor, it implements any one of the wireless sensing covert communications based on beacon signals.

Claims

1. A distributed reactive power resource coordination method based on cooperative game and adaptive learning, characterized in that It includes the following steps: (1) Through sensors and monitoring systems, collect voltage, current, and reactive power contribution data of the device in real time; (2) Build a game and revenue equilibrium model based on the voltage, current, and reactive power contribution data and train it; (3) Use meta-learning to optimize the model; (4) Generate a distribution plan for reactive power resources.

2. The distributed reactive power resource coordination method based on cooperative game and adaptive learning according to claim 1, wherein, In step (1), the device includes: inverters, SVG, and capacitors.

3. A distributed reactive power resource coordination method based on cooperative game and adaptive learning according to claim, characterized in that, Step (2) is specifically as follows: (21) Use a cooperative game model to describe the distribution problem of reactive power resources;; (22) Calculate the revenue distribution using the Shapley value; (23) Set a deviation threshold, and use deviation feedback to adjust the parameters in the game model to optimize the revenue distribution; (24) Calculate the gain distribution based on contributions.

4. A distributed reactive power resource coordination method based on cooperative game and adaptive learning according to claim, characterized in that The formula for step (22) is as follows: Among them, φi is the revenue distribution of participant i, N is the set of all participants, S is a subset, and v(S) is the value function of the subset.

5. A distributed reactive power resource coordination method based on cooperative game and adaptive learning according to claim, characterized in that Step (23) is specifically as follows: Set a deviation threshold ∈. When |D i | > ∈, adjust the allocation strategy of reactive power resources; Di = Pi - Fi Among them, Di is the payment deviation of participant i, Pi is the actual revenue, and Fi is the expected fair revenue.

6. The distributed reactive power resource coordination method based on cooperative game and adaptive learning according to claim, characterized in that In step (23), the formula for using deviation feedback to adjust the parameters in the game model is as follows: Among them, represents the contribution degree after feedback, represents the contribution degree before feedback. The remaining parameters have been mentioned in the previous text. The weight α of the dynamic adjustment participant is the adjustment step size.

7. A distributed reactive power resource coordination method based on cooperative game and adaptive learning according to claim, characterized in that Step (24) is specifically as follows: First, calculate the contribution degree, and the formula is as follows: Among them, Ci is the contribution degree of participant i, Qi is its reactive power contribution, and ∑j∈NQj is the total contribution of all participants. Secondly, calculate the gain distribution, and the formula is as follows: Gi = Ci·R Among them, Gi is the gain of participant i, and R is the total revenue of the system.

8. A distributed reactive power resource coordination method based on cooperative game and adaptive learning according to claim, characterized in that Use the MAML algorithm to optimize the model.

9. An electronic device, comprising a memory, a processor, and a computer program stored on the memory and executable on the processor, characterized in that, When the computer program is loaded into the processor, it implements a distributed reactive power resource coordination method according to any one of claims 1-8.

10. A storage medium storing a computer program, characterized in that, When the computer program is executed by the processor, it implements a distributed reactive power resource coordination method according to any one of claims 1-8.

Images