Switch cabinet electric field intensity optimization strategy determination method and device, equipment and medium
By applying an attention-based dual-depth Q-network adversarial training model in the switchgear, multi-dimensional data is obtained to optimize electric field strength and insulation system margin. This solves the problem of poor electric field strength optimization in existing methods and improves the operational safety and reliability of the switchgear.
Patent Information
- Application Number
- CN202411388512.X
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2024-10-08
- Publication Date
- 2024-11-08
- Estimated Expiration
- 2044-10-08
AI Technical Summary
Existing methods for optimizing the electric field strength of switchgear have the problem of poor optimization effect, which cannot adapt to environmental changes and diverse application requirements, resulting in reduced safety and reliability of switchgear operation.
An electric field strength optimization strategy model is developed using an attention-based dual-deep Q-network for adversarial training. By acquiring multi-dimensional data of target components within the switchgear, such as average dielectric strength, thermal resistance, overall mechanical strength, insulation thickness, and frequency response, the model outputs the optimal optimization strategy to adjust parameters and optimize the electric field strength and insulation system margin.
This improves the decision-making accuracy of electric field strength optimization strategies, enhances the operational safety and reliability of the switchgear, and ensures the stability and efficiency of the system.
Smart Images

Figure CN118917213A_ABST
Abstract
Description
Technical Field
[0001] The present invention relates to the field of power electronics technology, and in particular to a method, device, equipment and medium for determining an optimization strategy for electric field strength of a switch cabinet. Background Art
[0002] With the rapid development of the national power grid and the widespread use of switchgear, switchgear accidents are also common, among which insulation accidents are a very prominent factor in switchgear accidents. Since the switchgear operates under high electric field strength for a long time, the local electric field is severely distorted, which is prone to surface flashover, breakdown and other faults, thereby reducing the operating safety and reliability of the switchgear.
[0003] The existing switchgear electric field strength optimization methods have certain limitations. They often rely on fixed optimization strategies and consider fewer dimensions of electric field strength influencing parameters. They cannot adapt to environmental changes and diverse application requirements, resulting in poor optimization effects on the electric field strength of switchgear. Summary of the invention
[0004] The present invention provides a method, device, equipment and medium for determining an electric field strength optimization strategy for a switch cabinet, so as to solve the problem of poor electric field strength optimization effect existing in the existing methods.
[0005] According to one aspect of the present invention, a method for determining an optimization strategy for electric field strength of a switch cabinet is provided, the method comprising:
[0006] Obtain target data of a target component in a switch cabinet; wherein the target data includes at least one of the following: average dielectric strength, average thermal resistance, average comprehensive mechanical strength, average insulation thickness, and average frequency response; the target component includes at least one of the following switch components: a load switch, a circuit breaker switch, and a grounding switch;
[0007] The target data is input into the electric field strength optimization strategy model to obtain the corresponding optimization strategy; wherein the optimization strategy includes at least one of the following: the parameters to be adjusted of the average dielectric strength, the parameters to be adjusted of the average thermal resistance, the parameters to be adjusted of the average comprehensive mechanical strength, the parameters to be adjusted of the average insulation thickness and the parameters to be adjusted of the average frequency response; the electric field strength optimization strategy model is obtained by adversarial training of a dual deep Q network based on the attention mechanism.
[0008] According to another aspect of the present invention, a device for determining an optimization strategy for electric field strength of a switch cabinet is provided, the device comprising:
[0009] A data acquisition module is used to acquire target data of a target component in a switch cabinet; wherein the target data includes at least one of the following: average dielectric strength, average thermal resistance, average comprehensive mechanical strength, average insulation thickness and average frequency response; the target component includes at least one of the following switch components: a load switch, a circuit breaker switch and a grounding switch;
[0010] An optimization strategy determination module is used to input target data into an electric field strength optimization strategy model to obtain a corresponding optimization strategy; wherein the optimization strategy includes at least one of the following: an adjusted parameter of average dielectric strength, an adjusted parameter of average thermal resistance, an adjusted parameter of average comprehensive mechanical strength, an adjusted parameter of average insulation thickness, and an adjusted parameter of average frequency response; the electric field strength optimization strategy model is obtained by adversarial training of a dual-depth Q network based on an attention mechanism.
[0011] According to another aspect of the present invention, an electronic device is provided, the electronic device comprising:
[0012] at least one processor; and
[0013] a memory communicatively connected to the at least one processor; wherein,
[0014] The memory stores a computer program executable by the at least one processor, and the computer program is executed by the at least one processor so that the at least one processor can execute the method for determining the switch cabinet electric field strength optimization strategy described in any embodiment of the present invention.
[0015] According to another aspect of the present invention, a computer-readable storage medium is provided, wherein the computer-readable storage medium stores computer instructions, and the computer instructions are used to enable a processor to implement the method for determining the switch cabinet electric field strength optimization strategy described in any embodiment of the present invention when executed.
[0016] The technical solution of the embodiment of the present invention obtains the target data of the target component in the switch cabinet; wherein the target data includes at least one of the following: average dielectric strength, average thermal resistance, average comprehensive mechanical strength, average insulation thickness and average frequency response; the target data is input into the electric field strength optimization strategy model to obtain the corresponding optimization strategy; wherein the optimization strategy includes at least one of the following: the parameters to be adjusted for the average dielectric strength, the parameters to be adjusted for the average thermal resistance, the parameters to be adjusted for the average comprehensive mechanical strength, the parameters to be adjusted for the average insulation thickness and the parameters to be adjusted for the average frequency response; the electric field strength optimization strategy model is obtained by adversarial training of a double-depth Q network based on an attention mechanism. This technical solution obtains the target data of the target components in the switch cabinet for different dimensions, and then uses the pre-trained electric field strength optimization strategy model to output the optimization strategy that can achieve the optimal electric field strength and insulation system margin of the switch cabinet in the current state, thereby solving the problem of poor electric field strength optimization effect in the existing method, improving the decision accuracy of the electric field strength optimization strategy, and thus improving the operating safety and reliability of the switch cabinet.
[0017] It should be understood that the contents described in this section are not intended to identify the key or important features of the embodiments of the present invention, nor are they intended to limit the scope of the present invention. Other features of the present invention will become easily understood through the following description. BRIEF DESCRIPTION OF THE DRAWINGS
[0018] In order to more clearly illustrate the technical solutions in the embodiments of the present invention, the following briefly introduces the drawings required for use in the description of the embodiments. 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.
[0019] Figure 1 is a flow chart of a method for determining an electric field strength optimization strategy for a switch cabinet provided in accordance with Embodiment 1 of the present invention;
[0020] Figure 2 is a flow chart of a method for determining a switch cabinet electric field strength optimization strategy provided in accordance with Embodiment 2 of the present invention;
[0021] Figure 3 is a training flow chart of an electric field strength optimization strategy model provided according to Embodiment 3 of the present invention;
[0022] Figure 4 is a dielectric strength distribution diagram provided according to the third embodiment of the present invention;
[0023] Figure 5 is a temperature distribution diagram provided according to the third embodiment of the present invention;
[0024] Figure 6 is a curve diagram showing the variation of mechanical strength with electric field strength according to the third embodiment of the present invention;
[0025] Figure 7 is a 3D surface diagram of the thickness of the insulating layer of the target component provided in accordance with the third embodiment of the present invention;
[0026] Figure 8 is a correlation heat map provided according to the third embodiment of the present invention;
[0027] Fig. 9 is a schematic diagram of the model structure and training provided according to the third embodiment of the present invention;
[0028] Fig.10 2 is a schematic diagram of the structure of a device for determining an electric field strength optimization strategy for a switch cabinet provided according to a fourth embodiment of the present invention;
[0029] Fig.11 It is a structural schematic diagram of an electronic device for implementing the method for determining the electric field strength optimization strategy of a switch cabinet according to an embodiment of the present invention. DETAILED DESCRIPTION
[0030] 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 should fall within the scope of protection of the present invention.
[0031] It should be noted that the terms "first", "second", etc. in the specification and claims of the present invention and the above-mentioned drawings are used to distinguish similar objects, and are not necessarily used to describe a specific order or sequence. It should be understood that the data used in this way can be interchanged where appropriate, so that the embodiments of the present invention described herein can be implemented in an order other than those illustrated or described herein. In addition, the terms "including" and "having" and any variations thereof are intended to cover non-exclusive inclusions, for example, a process, method, system, product or device that includes a series of steps or units is not necessarily limited to those steps or units that are clearly listed, but may include other steps or units that are not clearly listed or inherent to these processes, methods, products or devices.
[0032] Embodiment 1
[0033] Figure 1A flow chart of a method for determining an electric field strength optimization strategy for a switch cabinet is provided for the first embodiment of the present invention. This embodiment can be applied to determine a corresponding electric field strength optimization strategy based on target data of target components in different dimensions in a switch cabinet, so as to assist users (such as operation and maintenance personnel) in optimizing the electric field strength of the switch cabinet. The method can be executed by a switch cabinet electric field strength optimization strategy determination device, which can be implemented in the form of hardware and / or software. The switch cabinet electric field strength optimization strategy determination device can be configured in an electronic device, which can include but is not limited to computers, terminals, servers and other devices with data processing capabilities. Figure 1 As shown, the present embodiment 1 provides a method for determining a switch cabinet electric field strength optimization strategy, which specifically includes the following steps:
[0034] S110. Obtain target data of target components in the switch cabinet; wherein the target data includes at least one of the following: average dielectric strength, average thermal resistance, average comprehensive mechanical strength, average insulation thickness and average frequency response; the target component includes at least one of the following switch components: load switch, circuit breaker switch and grounding switch.
[0035] The target components may include but are not limited to various switches, operating mechanisms, current-carrying devices and other components in the switch cabinet. In one embodiment, the target components include at least one of the following switch components: load switch, circuit breaker switch and grounding switch.
[0036] The target data can be understood as multidimensional parameter data that will affect the electric field distribution and insulation system margin of the switchgear. The target data may include at least one of the following: average dielectric strength, average thermal resistance, average comprehensive mechanical strength, average insulation thickness and average frequency response; among which, the average dielectric strength is used to characterize the overall dielectric strength performance of multiple measurement points on the target component; the average thermal resistance is used to characterize the overall heat dissipation performance of multiple measurement points on the target component; the average comprehensive mechanical strength is used to characterize the overall mechanical strength and electric field distribution of multiple measurement points on the target component; the average insulation thickness is used to characterize the overall insulation layer thickness distribution of multiple measurement points on the target component; and the average frequency response is used to characterize the overall frequency response distribution of multiple measurement points on the target component.
[0037] It is understandable that, in addition to the parameter data of the above dimensions, the target data may also include parameter data of more dimensions, which can be specifically set according to the actual application scenario and business requirements of the switch cabinet, and the embodiment of the present invention does not limit this.
[0038] In an embodiment of the present invention, when the electric field strength of the switch cabinet needs to be optimized, for example, when a preset electric field strength optimization cycle is reached, or when an electric field strength optimization instruction from a user is received, target data of target components in the switch cabinet for different dimensions can be obtained, so that the corresponding electric field strength optimization strategy can be determined according to the target data. The target data includes at least one of the following: average dielectric strength, average thermal resistance, average comprehensive mechanical strength, average insulation thickness, and average frequency response.
[0039] It can be understood that this embodiment does not impose specific restrictions on the number of target components selected and their distribution positions in the switch cabinet, and can be set accordingly according to the actual application scenarios and business needs of the switch cabinet. The more target components are selected, the more accurate the final electric field strength optimization strategy will be, and the better the optimization effect will be. The distribution positions of the target components in the switch cabinet can be selected from but not limited to the following positions: the sharp corners of the switch cabinet (such as the edge of the partition, the busbar bridge connection, etc.), the positions of components directly exposed to the air (such as current-carrying devices), etc.
[0040] S120. Input the target data into the electric field strength optimization strategy model to obtain the corresponding optimization strategy; wherein the optimization strategy includes at least one of the following: the parameters to be adjusted for the average dielectric strength, the parameters to be adjusted for the average thermal resistance, the parameters to be adjusted for the average comprehensive mechanical strength, the parameters to be adjusted for the average insulation thickness, and the parameters to be adjusted for the average frequency response; the electric field strength optimization strategy model is obtained by adversarial training of a dual-depth Q network based on an attention mechanism.
[0041] Among them, the electric field strength optimization strategy model can be understood as a network model used to determine the corresponding electric field strength optimization strategy according to the target data. The output optimization strategy can refer to the best action output by the electric field strength optimization strategy model under the current state, that is, the input target data, that is, the parameters to be adjusted corresponding to the target data of each dimension, so as to achieve the optimal goals of the electric field strength of the switchgear (as small as possible) and the insulation system margin (as large as possible). The electric field strength optimization strategy model can be obtained by pre-adversarial training based on the dual-depth Q network of the attention mechanism.
[0042] As a reinforcement learning algorithm, Double Deep Q Network (DDQN) is an improvement and extension of Deep Q Network (DQN), which aims to solve the problem of overestimation in DQN model when estimating Q value. The DDQN model consists of an evaluation network and a target network with the same network structure but different network parameters. Its basic idea is to separate the selection and evaluation of actions, that is, the evaluation network is used to select the best action, while the target network is used to estimate the Q value of the action, thereby reducing the impact of overestimation.
[0043] The attention mechanism is a mechanism that enables the model to pay attention to important information in the input features. By adaptively adjusting the model's attention to different features, the model can focus on the information most relevant to the current task, improve the model's ability to capture key features, and thus enhance the model's decision-making accuracy.
[0044] Adversarial training is a training method that can improve the robustness of a model. It introduces adversarial samples during model training, forcing the model to learn how to identify and resist the impact of adversarial samples, thereby enhancing the robustness and generalization ability of the model. Adversarial samples refer to samples that make the model make wrong decisions by adding small perturbations to the original sample data. This perturbation is usually imperceptible to humans, but it can have a serious impact on the model.
[0045] In an embodiment of the present invention, after obtaining the target data of target components in the switch cabinet for different dimensions, it can be input into a pre-trained electric field strength optimization strategy model, and the model is used to decide the best action for the switch cabinet to achieve the optimal electric field strength and insulation system margin in the current state, that is, the parameters to be adjusted corresponding to the target data of each dimension, and then each parameter to be adjusted is used as the final optimization strategy. Subsequent users can adjust the relevant parameters of the switch cabinet based on the optimization strategy, such as adjusting the insulating material, optimizing the heat dissipation structure, improving the mechanical design, etc., so as to achieve the goal of optimizing the electric field strength and improving the insulation system margin, thereby improving the operation safety and reliability of the switch cabinet.
[0046] The technical solution of the embodiment of the present invention obtains the target data of the target component in the switch cabinet; wherein the target data includes at least one of the following: average dielectric strength, average thermal resistance, average comprehensive mechanical strength, average insulation thickness and average frequency response; the target component includes at least one of the following switch components: load switch, circuit breaker switch and grounding switch; the target data is input into the electric field strength optimization strategy model to obtain the corresponding optimization strategy; wherein the optimization strategy includes at least one of the following: the parameters to be adjusted for the average dielectric strength, the parameters to be adjusted for the average thermal resistance, the parameters to be adjusted for the average comprehensive mechanical strength, the parameters to be adjusted for the average insulation thickness and the parameters to be adjusted for the average frequency response; the electric field strength optimization strategy model is obtained by adversarial training of a double deep Q network based on an attention mechanism. This technical solution obtains the target data of the target components in the switch cabinet for different dimensions, and then uses the pre-trained electric field strength optimization strategy model to output the optimization strategy that can achieve the optimal electric field strength and insulation system margin of the switch cabinet in the current state, thereby solving the problem of poor electric field strength optimization effect in the existing method, improving the decision accuracy of the electric field strength optimization strategy, and thus improving the operation safety and reliability of the switch cabinet.
[0047] Embodiment 2
[0048] Figure 2 This is a flow chart of a method for determining a switch cabinet electric field strength optimization strategy provided in the second embodiment of the present invention, which is further optimized and expanded based on the above implementation, and can be combined with various optional technical solutions in the above implementation. Figure 2 As shown, the second embodiment provides a method for determining a switch cabinet electric field strength optimization strategy, which specifically includes the following steps:
[0049] S210, acquiring target data of a target component in the switch cabinet.
[0050] In an embodiment of the present invention, the target data corresponding to the target component can be found from a preset data storage location such as a database or Excel file according to identification information such as the target component number, and the target data includes at least one of the following: average dielectric strength, average thermal resistance, average comprehensive mechanical strength, average insulation thickness, and average frequency response.
[0051] Among them, the average dielectric strength is the average value of the dielectric strength corresponding to multiple measurement points on the target component. The dielectric strength is determined by the breakdown voltage collected at the corresponding measurement points and the thickness of the insulation layer of the target component. The breakdown voltage and the thickness of the insulation layer of the target component are collected by the electrode testing device and the thickness measuring device respectively.
[0052] Specifically, measurement points can be set at different positions on the target component, and the positions of the measurement points can be evenly distributed on the target component according to actual needs, or several positions can be randomly selected as measurement points, which is not limited in this embodiment. The breakdown voltage of each measurement point and the thickness of the insulation layer of the target component can be collected using an electrode testing device and a thickness measuring device, respectively, and then the following formula is called to determine the dielectric strength of each measurement point:
[0053]
[0054] In the formula, Represents the dielectric strength of the ith measurement point in kV / mm; represents the breakdown voltage of the ith measurement point, in kV; Indicates the insulation layer thickness of the target component at the i-th measurement point, in mm; represents the electric field uniformity factor.
[0055] Finally, the average value of the dielectric strength corresponding to all measurement points is taken as the average dielectric strength.
[0056] Furthermore, the average thermal resistance is the average value of the thermal resistances corresponding to multiple measurement points on the target component. The thermal resistance is determined by the heating power applied at the corresponding measurement point and the collected temperature difference. The heating power is collected by the thermal testing device, and the initial temperature and final temperature corresponding to the temperature difference are collected by the infrared thermal imager.
[0057] Specifically, the corresponding heat can be applied to the target component by setting the heating power of the thermal test device, and then the temperature difference (the difference between the initial temperature and the final temperature) of each measurement point can be collected by using an infrared thermal imager, and then the thermal resistance of each measurement point can be determined by calling the following formula:
[0058]
[0059] In the formula, represents the thermal resistance of the ith measurement point, in K / W (Kelvin / Watt); represents the temperature difference of the i-th measuring point, in K (Kelvin); P represents the heating power, in W.
[0060] Finally, the average thermal resistance corresponding to all measurement points is taken as the average thermal resistance.
[0061] Furthermore, the average comprehensive mechanical strength is the average value of the comprehensive mechanical strengths corresponding to multiple measurement points on the target component, and the comprehensive mechanical strength is determined by the actual stress and electric field strength collected at the corresponding measurement points, and the actual stress and electric field strength are collected by stress sensors and electric field sensors respectively.
[0062] Specifically, the stress test device and the electric field generator can be used to apply mechanical stress and voltage to the target component, and then the stress sensor and the electric field sensor can be used to collect the actual stress and electric field strength of each measurement point, and then the following formula can be used to determine the comprehensive mechanical strength of each measurement point:
[0063]
[0064] In the formula, Indicates the comprehensive mechanical strength of the i-th measuring point, in N / mm 2 ; Indicates the actual stress at the i-th measuring point, in N / mm 2 ; represents the electric field strength at the ith measurement point, in kV / mm; Indicates the elastic modulus of the target component material in N / mm 2 ; represents the temperature of the ith measurement point in °C; represents the dielectric constant; k1-k5 represent the regression coefficients respectively; Represents a constant.
[0065] Finally, the average value of the comprehensive mechanical strengths corresponding to all measurement points is taken as the average comprehensive mechanical strength.
[0066] Furthermore, the average insulation thickness is an average value of the thickness of the insulation layer of the target component corresponding to multiple measurement points on the target component, and the insulation thickness is acquired by a thickness measuring device.
[0067] Specifically, a thickness measuring device may be used to collect the thickness of the insulation layer of the target component at each measuring point on the target component, and then the average value of the insulation layer thickness of the target component corresponding to all measuring points is taken as the average insulation thickness.
[0068] Furthermore, the average frequency response is an average value of frequency responses corresponding to multiple measurement points on the target component, and the frequency response is acquired by a frequency response analyzer.
[0069] Specifically, a frequency can be applied to the target component using a frequency response analyzer to obtain the frequency response (value) of each measurement point, and then the average value of the frequency responses corresponding to all measurement points is taken as the average frequency response.
[0070] S220, preprocessing the target data to obtain normalized target data, and inputting the normalized target data into a convolutional neural network layer to extract target feature representations corresponding to the normalized target data.
[0071] In an embodiment of the present invention, a pre-trained electric field strength optimization strategy model can be obtained from a storage location such as a local or cloud server, and the model can be deployed in an actual switch cabinet application environment to ensure that it can process and optimize the electric field strength and insulation system margin in real time. At the same time, the model's computational efficiency and hardware compatibility need to be considered during the deployment process to ensure that the model can run efficiently in actual applications. The electric field strength optimization strategy model specifically includes: a convolutional neural network layer, an attention layer, a target network, and an evaluation network, wherein the convolutional neural network layer is used to extract a high-dimensional representation of input features (i.e., target data of different dimensions) to enhance the model's expressive power; the attention layer is used to adaptively adjust the model's attention to different features, so that the model can focus on the information most relevant to the current task, improve the model's ability to capture key features, and thereby enhance the model's decision accuracy; the target network and the evaluation network are used to decide the best action for the switch cabinet to achieve the best electric field strength and insulation system margin in the current state.
[0072] Specifically, the acquired target data can be subjected to preprocessing operations such as noise anomaly removal, missing value filling and standardization in sequence to obtain preprocessed normalized target data, and then the normalized target data is input into the convolutional neural network layer in the electric field strength optimization strategy model for multiple iterations to extract the target feature representation corresponding to the normalized target data. The convolutional neural network layer can be expressed as follows:
[0073]
[0074] In the formula, Represents the output after the convolution operation; x represents the input feature, that is, the multi-dimensional target data; W represents the convolution kernel weight; b represents the bias; ReLU represents the activation function.
[0075] S230. Input the target feature representation into the attention layer, and output the target attention matrix corresponding to the target feature representation.
[0076] In an embodiment of the present invention, the target feature representation output by the convolutional neural network layer can be used as the input of the attention layer in the electric field strength optimization strategy model. By calculating the similarity scores between the three matrices of query (Query), key (Key), and value (Value), and normalizing these scores to obtain the attention weights, the model can perform weighted processing on the input features (the target feature representation output by the convolutional neural network layer) according to these weights, thereby forming attention to specific information and outputting the corresponding target attention matrix. Among them, the target attention matrix output by the attention layer can be expressed as follows:
[0077]
[0078] Where Q represents the query matrix; K represents the key matrix, represents the transpose of the matrix K; V represents the value matrix; represents the dimension of the key matrix; softmax represents the normalization function.
[0079] S240, using the target network and the evaluation network to process the target attention matrix, outputting the parameters to be adjusted corresponding to each target data, and using each parameter to be adjusted as an optimization strategy.
[0080] The parameters to be adjusted may include the amounts to be adjusted or the weights to be adjusted corresponding to the target data of each dimension.
[0081] In an embodiment of the present invention, since after the model training is completed, the target network and the evaluation network have learned how to select the best action according to the current state to maximize the cumulative reward, the attention matrix containing the correlation between each input feature can be input into the target network and the evaluation network respectively for decision processing, so as to output the best action for the switch cabinet to achieve the optimal electric field strength and insulation system margin under the current state, that is, the parameters to be adjusted corresponding to each target data, such as the amount to be adjusted or the weight to be adjusted, and each parameter to be adjusted is used as the final optimization strategy for optimizing the electric field strength of the switch cabinet.
[0082] Among them, the best action output by the model can be expressed as:
[0083]
[0084] In the formula, Indicates the current status The best action to choose; Indicates in status The Q value corresponding to taking action a.
[0085] Furthermore, on the basis of the above-mentioned embodiments of the invention, after obtaining the parameters to be adjusted for each target data output by the model, the processing results between each target data and the corresponding parameters to be adjusted can also be used as the final optimization strategy. For example, if the parameters to be adjusted are the adjusted quantities corresponding to each target data, then the sum of each target data and the corresponding adjusted quantities can be used as the final optimization strategy for optimizing the electric field strength of the switch cabinet; if the parameters to be adjusted are the weights to be adjusted corresponding to each target data, then the product of each target data and the corresponding weights to be adjusted can be used as the final optimization strategy for optimizing the electric field strength of the switch cabinet.
[0086] Furthermore, users can dynamically adjust relevant parameters of the switchgear according to the optimization strategy output by the model, such as adjusting the insulation material and thickness, optimizing the heat dissipation structure, improving the mechanical design, improving the manufacturing process, etc., so as to achieve the goal of optimizing the electric field strength and improving the margin of the insulation system, thereby improving the operating safety and reliability of the switchgear.
[0087] Further, based on the above-mentioned embodiments of the invention, after the electric field strength optimization strategy model is deployed in the actual switch cabinet application environment, the performance of the model can be regularly monitored to ensure its continuous optimization effect in actual applications. If it is found that the model performance has degraded or the environment has changed, the model needs to be retrained or fine-tuned to adapt to the new application scenario. The model performance can be expressed as follows:
[0088]
[0089] In the formula, Performance represents the average performance of the model; Represents the reward for each decision; Indicates the total number of decisions. If the average performance of the model is lower than the preset performance threshold, the model needs to be retrained or fine-tuned.
[0090] The technical solution of the embodiment of the present invention obtains the target data of the target component in the switch cabinet; pre-processes the target data to obtain the normalized target data, and inputs the normalized target data into the convolutional neural network layer to extract the target feature representation corresponding to the normalized target data; inputs the target feature representation into the attention layer, and outputs the target attention matrix corresponding to the target feature representation; uses the target network and the evaluation network to process the target attention matrix, outputs the parameters to be adjusted corresponding to each target data, and uses each parameter to be adjusted as an optimization strategy. This technical solution uses the multi-dimensional parameters that affect the electric field strength of the switch cabinet as data input, and then uses the pre-trained electric field strength optimization strategy model to process the target data of different dimensions, so as to output the optimization strategy that can achieve the optimal electric field strength and insulation system margin of the switch cabinet in the current state, solves the problem of poor electric field strength optimization effect in the existing method, improves the decision accuracy of the electric field strength optimization strategy, and thus improves the operating safety and reliability of the switch cabinet, ensuring the stability and efficiency of the system.
[0091] Embodiment 3
[0092] Figure 3 This is a training flow chart of an electric field strength optimization strategy model provided in the third embodiment of the present invention. Figure 3 As shown in FIG. 1 , the training process of the electric field strength optimization strategy model includes the following steps:
[0093] S310: Obtain historical data of the target component.
[0094] In an embodiment of the present invention, historical data of target components in different dimensions can be obtained from preset data storage locations such as, but not limited to, databases and Excel files; wherein the historical data includes at least one of the following: average dielectric strength historical data, average thermal resistance historical data, average comprehensive mechanical strength historical data, average insulation thickness historical data, and average frequency response historical data. The following describes the collection process of historical data of each dimension:
[0095] ① Collect historical data on average dielectric strength
[0096] First, prepare the target component and experimental equipment. Make sure the surface of the target component is clean and free of impurities, place the target component in the electrode test device, ensure that the target component is in close contact with the electrode plate, and set measurement points at different locations on the target component.
[0097] Then, voltage application and breakdown detection are performed. The voltage applied to the target component is gradually increased, and the changes on the surface of the target component are recorded in real time until the target component breaks down. The breakdown voltage and the thickness of the insulation layer of the target component are recorded at each measurement point, and the dielectric strength of each measurement point is determined based on the breakdown voltage and the thickness of the insulation layer of the target component. .
[0098] Next, data enhancement technology can be used to generate more samples to solve the problem of small samples and increase the diversity of sample data, thereby improving the generalization ability of the model. Specifically, the data enhancement model can be used to simulate different test environments of the target component, output the breakdown voltage and insulation layer thickness of the target component after n data enhancement simulation experiments at each measurement point (n also represents the number of samples in the subsequent model), and then determine the jth ( ) The average value of enhanced dielectric strength obtained in the data augmentation simulation experiment is:
[0099]
[0100] In the formula, represents the average value of the enhanced dielectric strength obtained in the jth data enhancement simulation experiment, in kV / mm; m represents the number of measurement points; represents the breakdown voltage obtained at the i-th measurement point in the j-th data enhancement simulation experiment, in kV; represents the insulation layer thickness of the target component obtained at the i-th measurement point in the j-th data enhancement simulation experiment, in mm; represents the electric field uniformity factor of the jth data enhancement simulation experiment.
[0101] After the above data enhancement processing, the average enhanced dielectric strength of n data enhancement simulation experiments corresponding to the target component can be obtained As the average dielectric strength historical data, n sample data are obtained.
[0102] Furthermore, before using dielectric strength as a parameter affecting the electric field strength of the switchgear, the frequency distribution of dielectric strength can be determined through the following data analysis process to understand the impact of dielectric strength on the electric field strength of the switchgear. The weighted standard deviation can be used in combination with the attention mechanism to assign weights to different samples and calculate the degree of discreteness of the experimental data:
[0103]
[0104] In the formula, Represents the weighted standard deviation in kV / mm; It represents the attention weight of the j-th sample, reflecting the importance of the sample in the overall data.
[0105] Figure 4 This is a dielectric strength distribution diagram after multiple experiments and data enhancement. It shows the dielectric strength distribution of sample data, where the horizontal axis represents dielectric strength (unit: kV / mm) and the vertical axis represents frequency. Through this diagram, we can intuitively observe the overall distribution and concentration trend of dielectric strength data, understand the dielectric strength performance of the target component under different experimental conditions, and thus evaluate its stability and reliability.
[0106] ② Collect historical data of average thermal resistance
[0107] First, make sure the target component surface is clean and free of impurities, and fix the target component in the thermal test device. Use an infrared thermal imager to arrange multiple measurement points on the target component surface to ensure that the temperature distribution on the target component surface can be fully captured.
[0108] Then conduct heating experiments and real-time monitoring. Use high-precision heat source equipment to gradually increase the power applied to the target component, and use an infrared thermal imager to monitor the temperature distribution on the surface of the target component in real time. Record the temperature change (temperature difference) and heating power of each measurement point, and determine the thermal resistance of each measurement point based on the heating power and temperature difference. .
[0109] Next, the collected data is processed and enhanced. An automated data acquisition system can be used to continuously record the temperature data of each measurement point, and preliminary analysis can be performed through data processing software. In order to improve the accuracy and diversity of the data, data enhancement techniques are used to generate more samples, such as expanding the data set by simulating temperature changes under different environmental conditions, and determining the jth ( ) The average value of enhanced thermal resistance obtained in the data enhancement simulation experiment is:
[0110]
[0111] In the formula, It represents the average value of enhanced thermal resistance obtained in the jth data enhancement simulation experiment, in K / W; represents the temperature difference obtained at the i-th measurement point in the j-th data augmentation simulation experiment; represents the heating power obtained in the jth data augmentation simulation experiment.
[0112] After the above data enhancement processing, the average value of the enhanced thermal resistance of n data enhancement simulation experiments corresponding to the target component can be obtained. As the average thermal resistance historical data, n sample data are obtained.
[0113] Furthermore, before using the average thermal resistance as a parameter affecting the electric field strength, the key factors affecting the heat dissipation performance can be determined through the following data analysis process. The collected temperature data can be analyzed in depth using big data analysis and machine learning algorithms. Through cluster analysis and regression modeling, the key factors affecting the heat dissipation performance can be found and a prediction model can be established:
[0114]
[0115] In the formula, Indicates standard deviation in K / W.
[0116] Figure 5 The temperature distribution diagram is generated by recording the temperature data of different temperature measurement points on the surface of the target component through an infrared thermal imager. This diagram shows the temperature distribution of the sample data surface during the heating process, and different colors represent different temperature ranges. Through this diagram, the temperature difference of each area on the surface of the target component can be intuitively observed, and then the heat dissipation performance and heat conduction characteristics of the target component can be analyzed.
[0117] ③ Collect historical data on average comprehensive mechanical strength
[0118] First, the target component is fixed in the stress testing device, and high-precision stress sensors and electric field sensors are arranged at different positions of the target component to ensure that the conditions of the target component under different mechanical stresses and electric fields can be fully captured.
[0119] Then mechanical stress and electric field are applied. The pressure applied to the target component is gradually increased through the automated loading system, and the electric field is applied through the electric field generator. The stress changes and electric field strength of each measuring point are recorded in real time, where the actual stress corresponding to each measuring point can be expressed as follows:
[0120]
[0121] In the formula, Indicates the actual stress at the i-th measuring point, in N / mm 2 ; represents the mechanical stress (pressure) applied at the i-th measuring point; Indicates the cross-sectional area of the i-th measuring point, in mm 2 ; Represents the angle between the stress direction corresponding to the i-th measurement point and the normal direction.
[0122] The electric field strength corresponding to each measurement point can be expressed as follows:
[0123]
[0124] In the formula, represents the electric field intensity at the ith measurement point, in kV / mm; V represents the applied voltage, in kV; represents the electrode spacing at the ith measurement point, in mm; Represents the dielectric constant.
[0125] Afterwards, the data collected by the stress sensor and the electric field sensor can be continuously recorded using an automated data acquisition system, and preliminary analysis can be performed using data processing software. The comprehensive mechanical strength of each measurement point is expressed as follows:
[0126]
[0127] Similarly, data enhancement technology can also be used to generate more samples to obtain the average value of enhanced comprehensive mechanical strength of n data enhancement simulation experiments corresponding to the target component as the average comprehensive mechanical strength historical data.
[0128] Furthermore, before using the comprehensive mechanical strength as an influencing parameter of the electric field strength, the key factors affecting the mechanical strength and the electric field distribution can be determined through the following data analysis process. Through multiple experiments, the average value of the comprehensive mechanical strength can be calculated to find out the key factors affecting the mechanical strength and the electric field distribution, and establish a prediction model:
[0129]
[0130] In the formula, Indicates the average value of comprehensive mechanical strength, unit N / mm 2 .
[0131] Figure 6 This is a graph showing the change of mechanical strength with electric field strength. It shows the distribution of mechanical strength of target component materials under different electric field strengths. Through analysis, we can observe the influence of electric field strength on mechanical strength. Specifically, the horizontal axis represents the electric field strength (unit: kV / mm), and the vertical axis represents the mechanical strength (unit: N / mm 2 ), the curve in the figure shows the change trend of the mechanical strength of the target component material under different electric field strengths. Through this figure, the influence of the electric field on the mechanical properties of the target component material can be intuitively analyzed, which helps to optimize the performance of the target component material in practical applications.
[0132] ④Collect historical data on average insulation thickness
[0133] First, make sure the surface of the target component is clean and free of impurities, and use a high-precision ultrasonic thickness gauge or optical microscope to measure the thickness of the insulation layer of the target component. Arrange measurement points at different locations of the target component to ensure that the distribution of the insulation layer thickness of the target component can be fully captured.
[0134] Then the insulation thickness is measured. The ultrasonic thickness gauge is used to gradually scan each measuring point on the surface of the target component and record the insulation thickness at each measuring point. Similarly, data enhancement technology can be used to generate more samples to obtain the average insulation thickness of the enhanced data, and the average insulation thickness of the enhanced data is used as the average insulation thickness historical data.
[0135] Furthermore, before using the average insulation thickness as a parameter affecting the electric field strength, the key factors affecting the insulation thickness can be determined through the following data analysis process. The thickness uniformity of the target component can be evaluated by comparing the thickness data of each measurement point:
[0136]
[0137] In the formula, Indicates the standard deviation of the insulation thickness in mm; m indicates the total number of measurement points; Indicates the insulation layer thickness of the target component at the i-th measurement point, in mm; Indicates the average insulation thickness of the enhanced data in mm.
[0138] Through cluster analysis and regression modeling, the key factors affecting insulation thickness are found out, and a prediction model is established. Figure 7 It is a 3D surface diagram of the insulation layer thickness of the target component, which shows the distribution of the insulation layer thickness of the target component at different measurement points. Figure 7 The different colored areas in the image can be used to visually observe the difference in insulation thickness at various locations on the surface of the target component. The horizontal and vertical axes represent the X position and Y position (i.e., position coordinates) of the target component, respectively, and the Z axis represents the insulation thickness (unit: mm). Figure 7 It can help analyze the uniformity of the insulation thickness of the target component and evaluate the insulation performance of the material, thereby helping to optimize the manufacturing process and improve product quality.
[0139] ⑤ Collect historical data of average frequency response
[0140] First, the frequency applied to the target component is gradually increased through a frequency response analyzer, and the frequency response values of each measurement point, such as dielectric constant, loss factor, etc., are recorded.
[0141]
[0142] In the formula, represents the frequency response value of the i-th measurement point; represents the frequency applied at the i-th measurement point, in Hz; represents the response amplitude; t represents the time; Indicates phase.
[0143] Similarly, data enhancement technology can also be used to generate more samples, and the average value of the enhanced frequency response of n data enhancement simulation experiments corresponding to the target component can be obtained as the average frequency response historical data.
[0144] Furthermore, before using the frequency response value as the parameter affecting the electric field strength, the key factors affecting the frequency response can be determined through the following data analysis process. The response data of the target component at different frequencies can be analyzed through Fourier transform to evaluate the impact of frequency on the performance of the target component:
[0145]
[0146] In the formula, represents the response function in the frequency domain; Represents the response value in the time domain.
[0147] Finally, the collected frequency response data is analyzed in depth. Through cluster analysis and regression modeling, the key factors affecting the frequency response are found and a prediction model is established:
[0148]
[0149] In the formula, represents the predicted frequency response value; represents the temperature of the i-th measurement point; Represents the pressure at the i-th measuring point, in Pa; represents the dielectric constant of the i-th measurement point; NN represents the neural network model.
[0150] Figure 8 This is a correlation heat map of the impact of related factors on frequency response. The figure shows the correlation between various factors, including frequency, temperature, pressure, dielectric constant and (frequency) response value. Through the colors and values in the figure, you can intuitively observe the degree of correlation between these factors. The horizontal axis and the vertical axis represent different factors respectively. The redder the color (the greater the correlation), the stronger the positive correlation, and the bluer the color (the smaller the correlation), the stronger the negative correlation. From this figure, it can be found that there is a strong positive correlation between frequency and response value, while the correlation between other factors is relatively weak.
[0151] S320: Preprocess the historical data to obtain normalized historical data.
[0152] In the embodiment of the present invention, the acquired historical data may be subjected to noise anomaly elimination, missing value filling and standardization processing in sequence to obtain preprocessed normalized historical data, wherein the data standardization formula is as follows:
[0153]
[0154] In the formula, X represents the standardized data; x represents the original data; represents the average value of the original data; represents the standard deviation of the original data; max(X) and min(X) represent the maximum and minimum values of the standardized data respectively; n represents the number of samples.
[0155] S330, performing feature mapping encoding on the normalized historical data to obtain a normalized data set.
[0156] In an embodiment of the present invention, the normalized historical data can be feature mapped and encoded to obtain a corresponding normalized data set, that is, the historical data of each dimension is mapped to a specific value to ensure that all states and factors can be accurately encoded and analyzed. The specific mapping method is as follows: {average dielectric strength: 1, average thermal resistance: 2, average comprehensive mechanical strength: 3, average insulation thickness: 4, average frequency response: 5}. This mapping method can ensure that each feature has a unique encoding, which is convenient for subsequent data analysis and model training.
[0157] S340, dividing the normalized data set into a training set, a validation set, and a test set according to a preset data set segmentation ratio, and initializing and constructing an initial model of the electric field strength optimization strategy.
[0158] In an embodiment of the present invention, the above-mentioned normalized data set can be divided into a training set, a validation set, and a test set according to a preset data set segmentation ratio of 70%:15%:15%, and the initial model of the electric field strength optimization strategy to be trained is initialized to initialize the model parameters; wherein the electric field strength optimization strategy model includes: a convolutional neural network layer, an attention layer, a target network, and an evaluation network. It can be understood that the above-mentioned preset data set segmentation ratio is only used as an example, and other segmentation ratios, such as 6:3:1, etc., can also be used in actual applications, and this embodiment does not limit this.
[0159] S350, adding adversarial disturbance to the sample data of the training set and the verification set, and training the initial model of the electric field strength optimization strategy according to the training set and the verification set to obtain the basic model of the electric field strength optimization strategy.
[0160] In the embodiment of the present invention, in order to enhance the robustness and generalization ability of the model, this embodiment introduces adversarial training technology, that is, adding small disturbances to the sample data of the training set and the verification set, and then using the training set and the verification set with disturbances to train the initial model of the electric field strength optimization strategy, so as to obtain the corresponding basic model of the electric field strength optimization strategy (without model testing). Among them, adversarial training can be expressed as follows:
[0161]
[0162] In the formula, It means fighting against loss; Represents the expectation under the joint distribution of state s and action a; Representation strategy; Indicates the Q value corresponding to taking action a in state s; represents adversarial disturbance; represents a perturbation set.
[0163] By adding adversarial perturbations during model training to simulate possible attacks or environmental changes, the model's performance can be enhanced when facing unknown and changing environments.
[0164] Further, based on the above-mentioned embodiment of the invention, S350 specifically includes the following steps:
[0165] S3501, inputting the sample data in the training set into the convolutional neural network layer of the initial model of the electric field strength optimization strategy respectively, and extracting the sample feature representation corresponding to each sample data;
[0166] S3502, inputting each sample feature representation into the attention layer of the initial model of the electric field strength optimization strategy respectively, and obtaining a sample attention matrix corresponding to each sample feature representation;
[0167] S3503, using each sample attention matrix as the state in the state space corresponding to the target network and the evaluation network of the initial model of the electric field strength optimization strategy, and obtaining the action space and reward function of the configuration of the initial model of the electric field strength optimization strategy;
[0168] S3504, selecting a state from the state space as the current state, and selecting a current action in the action space based on the current state using a preset dynamic exploration strategy, so as to obtain an updated state according to the interaction between the current action and the environment;
[0169] S3505, determining an update action according to the update state by evaluating the network; wherein the update action is used to interact with the environment to obtain a new update state;
[0170] S3506, determining the actual Q value corresponding to the current state and the current action through the evaluation network and the reward function;
[0171] S3507, determining the target Q value corresponding to the update state and the update action through the target network;
[0172] S3508, updating the network parameters of the evaluation network according to the loss function values corresponding to the actual Q value and the target Q value;
[0173] S3509, copying the network parameters of the evaluation network to the target network at every preset time period;
[0174] S35010, when the number of model iterations of the initial model of the electric field strength optimization strategy reaches a preset number of iterations, or the loss function value reaches a preset threshold, the current initial model of the electric field strength optimization strategy is used as the target model of the electric field strength optimization strategy;
[0175] S35011. Use the verification set to verify the electric field strength optimization strategy target model, and adjust the model parameters until the electric field strength optimization strategy target model meets the preset model verification indicators to obtain the trained electric field strength optimization strategy basic model.
[0176] In an embodiment of the present invention, each action in the action space includes at least one of the following: an adjustment parameter for adjusting the average dielectric strength, an adjustment parameter for adjusting the average thermal resistance, an adjustment parameter for adjusting the average comprehensive mechanical strength, an adjustment parameter for adjusting the average insulation thickness, and an adjustment parameter for adjusting the average frequency response, wherein the adjustment parameter may be an adjustment amount or an adjustment weight. Exemplarily, the action may be represented in the following form: {(1, adjustment parameter for adjusting the average dielectric strength); (2, adjustment parameter for adjusting the average thermal resistance); (3, adjustment parameter for adjusting the average comprehensive mechanical strength); (4, adjustment parameter for adjusting the average insulation thickness); (5, adjustment parameter for adjusting the average frequency response)}, wherein 1, 2, 3, 4, and 5 are the mapping values corresponding to the electric field strength influencing parameters after the above-mentioned feature mapping encoding.
[0177] Based on the goal of reducing the electric field strength and improving the insulation system margin, the following reward function can be designed:
[0178]
[0179] In the formula, represents the reward value corresponding to taking action a in state s; A represents the average dielectric strength; B represents the average thermal resistance; C represents the average comprehensive mechanical strength; D represents the average insulation thickness; E represents the average frequency response; , , , and are the weights of the parameters affecting the electric field strength, as follows:
[0180] : average dielectric strength, which should be maximized (positive weight);
[0181] : average thermal resistance, which should be minimized (negative weight);
[0182] : average comprehensive mechanical strength, which should be maximized (positive weight);
[0183] : average insulation thickness, which should be maximized (positive weight);
[0184] : Average frequency response, which should be minimized (negative weights).
[0185] See also Fig. 9 , the specific training process of the model is as follows:
[0186] In the embodiment of the present invention, the priority experience replay mechanism can be used to train the initial model of the electric field strength optimization strategy, and each state in the state space is respectively input into the evaluation network for storing and estimating the Q value in the experience replay buffer. The evaluation grid outputs the current action according to the preset dynamic exploration strategy network, i.e., the ε-greedy strategy. In the environment, on the one hand, the state of the environment changes accordingly. become (i.e. update the state), on the other hand, the environment will take corresponding actions according to the reward function The reward is immediately fed back to the agent's experience replay buffer. When a certain amount of experience samples are stored in the experience replay buffer, a small batch of samples is extracted from the experience replay buffer using the priority experience replay mechanism and input into the evaluation network and the target network respectively. The target network outputs the updated state. The corresponding target Q value, which is different from the current reward in the experience sample After addition, the root mean square value is calculated together with the actual Q value output by the evaluation network, which is the loss function value of the network. The loss function value is fed back to the evaluation network, and the network parameters of the evaluation network are updated according to the principle of minimizing loss and the gradient descent method, so that the output action can obtain the maximum cumulative reward. This process will be repeated in each training round until it finally converges or reaches the preset number of training rounds, that is, the preset model training termination condition is met. At the same time, during the model training process, the network parameters of the evaluation network will be regularly copied to the target network, thereby slowing down the frequency of parameter updates, stabilizing the training process, and preventing the model from falling into a local optimal solution.
[0187] The calculation formula of Q value is as follows:
[0188]
[0189] In the formula, represents the Q value corresponding to taking action a in state s; r represents the reward value; represents the discount factor; Indicates updating status (next state) to take the best action The corresponding target Q value.
[0190] By introducing a priority experience replay mechanism, the training frequency of important samples is increased. In the experience replay buffer, the following formula is used:
[0191]
[0192] In the formula, represents the priority of sample j; represents the TD error of sample j (the difference between the target Q value and the current actual Q value); Represents a small constant used to avoid the situation where the priority is zero. By introducing the priority experience replay mechanism, high-priority samples can be trained more frequently, thereby improving the learning efficiency and effect of the model.
[0193] By using the experience replay buffer, past experience samples can be stored and reused to improve sample utilization efficiency:
[0194]
[0195] In the formula, Represents the experience replay buffer; Indicates the current state; Indicates the current action; Indicates the current reward; Represents the updated state (i.e., the next state). By introducing the experience replay buffer, it is possible to ensure that the model can learn from past experience and repeatedly use these samples for training, thereby improving training efficiency.
[0196] By adopting a preset dynamic exploration strategy, using the ε-greedy strategy, and gradually reducing the ε value, from the initial high exploration rate to a low exploration rate, balancing exploration and utilization:
[0197]
[0198] In the formula, ε represents the exploration rate; and denote the minimum and maximum exploration rates, respectively; represents the decay rate; t represents the time step. By adopting a preset dynamic exploration strategy to gradually reduce the exploration rate, the model can explore more different state and action combinations in the early stage, and gradually focus on optimizing using the learned knowledge in the later stage.
[0199] When the training process of the initial model of the electric field strength optimization strategy meets the preset model training termination conditions, such as the number of model iterations reaches the preset number of iterations, or the loss function value reaches the preset threshold, the current initial model of the electric field strength optimization strategy is used as the target model of the electric field strength optimization strategy.
[0200] Then, the electric field strength optimization strategy target model obtained through the above training is verified by using the verification set and preset model verification indicators such as model execution time and model execution success rate. The model parameters are continuously adjusted until the electric field strength optimization strategy target model meets the preset model verification indicators, thereby obtaining the trained electric field strength optimization strategy basic model.
[0201] S360. Use the test set to perform performance testing on the basic model of the electric field strength optimization strategy.
[0202] In the embodiment of the present invention, after obtaining the trained electric field strength optimization strategy basic model, a test set can be used to perform performance testing on it to evaluate its model performance.
[0203] S370: If the test result of the performance test meets the preset model test pass condition, the electric field strength optimization strategy basic model is used as the electric field strength optimization strategy model.
[0204] In an embodiment of the present invention, measurement indicators including but not limited to: reward value change trend, Q value change trend, average reward, cumulative reward, final reward, etc. can be used to determine whether the test results of the performance test of the electric field strength optimization strategy basic model meet the preset model test pass conditions, such as whether the preset average reward threshold is reached, etc. If the performance test passes, the above-mentioned trained electric field strength optimization strategy basic model is used as the final electric field strength optimization strategy model; otherwise, the model is retrained until the retrained electric field strength optimization strategy basic model meets the preset model test pass conditions.
[0205] The technical solution of the embodiment of the present invention uses the multi-dimensional historical data of the target component to train the model, and sets a priority experience playback buffer during the model training process to increase the training frequency of important samples, ensuring that high-priority samples are trained more frequently, thereby improving the learning efficiency and effect of the model; by introducing the attention mechanism, the model's attention to different features is adaptively adjusted to improve the ability to capture key features; by introducing adversarial training technology, the robustness and generalization ability of the model are enhanced. The electric field strength optimization strategy model trained using this technical solution can improve the accuracy of the model's decision output in the face of unknown and changing environments, and improve the model's performance in optimizing electric field strength and improving insulation system margins.
[0206] Embodiment 4
[0207] Fig.10 This is a schematic diagram of the structure of a switch cabinet electric field strength optimization strategy determination device provided by the fourth embodiment of the present invention. Fig.10 As shown, the device comprises:
[0208] The data acquisition module 41 is used to acquire target data of target components in the switch cabinet; wherein the target data includes at least one of the following: average dielectric strength, average thermal resistance, average comprehensive mechanical strength, average insulation thickness and average frequency response; the target component includes at least one of the following switch components: load switch, circuit breaker switch and grounding switch;
[0209] The optimization strategy determination module 42 is used to input the target data into the electric field strength optimization strategy model to obtain the corresponding optimization strategy; wherein the optimization strategy includes at least one of the following: the parameters to be adjusted for the average dielectric strength, the parameters to be adjusted for the average thermal resistance, the parameters to be adjusted for the average comprehensive mechanical strength, the parameters to be adjusted for the average insulation thickness and the parameters to be adjusted for the average frequency response; the electric field strength optimization strategy model is obtained by adversarial training of a dual deep Q network based on an attention mechanism.
[0210] The technical solution of the embodiment of the present invention obtains the target data of the target component in the switch cabinet through the data acquisition module; wherein the target data includes at least one of the following: average dielectric strength, average thermal resistance, average comprehensive mechanical strength, average insulation thickness and average frequency response; the target component includes at least one of the following switch components: load switch, circuit breaker switch and grounding switch; the target data is input into the electric field strength optimization strategy model through the optimization strategy determination module to obtain the corresponding optimization strategy; wherein the optimization strategy includes at least one of the following: the parameters to be adjusted for the average dielectric strength, the parameters to be adjusted for the average thermal resistance, the parameters to be adjusted for the average comprehensive mechanical strength, the parameters to be adjusted for the average insulation thickness and the parameters to be adjusted for the average frequency response; the electric field strength optimization strategy model is obtained by adversarial training of the double deep Q network based on the attention mechanism. This technical solution obtains the target data of the target components in the switch cabinet for different dimensions, and then uses the pre-trained electric field strength optimization strategy model to output the optimization strategy that the switch cabinet can achieve the optimal electric field strength and insulation system margin in the current state, thereby solving the problem of poor electric field strength optimization effect in the existing method, improving the decision accuracy of the electric field strength optimization strategy, and thus improving the operation safety and reliability of the switch cabinet.
[0211] Further, based on the above-mentioned embodiment of the invention, the average dielectric strength is the average value of the dielectric strength corresponding to multiple measurement points on the target component, and the dielectric strength is determined by the breakdown voltage collected at the corresponding measurement points and the thickness of the insulation layer of the target component, and the breakdown voltage and the thickness of the insulation layer of the target component are collected by the electrode testing device and the thickness measuring device respectively;
[0212] The average thermal resistance is the average value of the corresponding thermal resistances of multiple measurement points on the target component. The thermal resistance is determined by the heating power applied at the corresponding measurement point and the collected temperature difference. The heating power is collected by the thermal test device, and the initial temperature and final temperature corresponding to the temperature difference are collected by the infrared thermal imager.
[0213] The average comprehensive mechanical strength is the average value of the comprehensive mechanical strength corresponding to multiple measurement points on the target component. The comprehensive mechanical strength is determined by the actual stress and electric field strength collected at the corresponding measurement points. The actual stress and electric field strength are collected by the stress sensor and the electric field sensor respectively.
[0214] The average insulation thickness is the average value of the insulation thickness of the target component corresponding to multiple measurement points on the target component;
[0215] The average frequency response is the average value of the frequency responses corresponding to multiple measurement points on the target component, and the frequency response is acquired by a frequency response analyzer.
[0216] Further, based on the above-mentioned embodiment of the invention, the electric field strength optimization strategy model includes: a convolutional neural network layer, an attention layer, a target network and an evaluation network; accordingly, the optimization strategy determination module 42 includes:
[0217] A data processing unit is used to preprocess the target data to obtain normalized target data, and input the normalized target data into the convolutional neural network layer to extract the target feature representation corresponding to the normalized target data;
[0218] An attention matrix determination unit, used for inputting the target feature representation into the attention layer and outputting a target attention matrix corresponding to the target feature representation;
[0219] The optimization strategy determination unit is used to process the target attention matrix using the target network and the evaluation network, output the parameters to be adjusted corresponding to each target data, and use each parameter to be adjusted as the optimization strategy.
[0220] Further, based on the above-mentioned embodiment of the invention, the training process of the electric field strength optimization strategy model includes:
[0221] Acquire historical data of the target component; wherein the historical data includes at least one of the following: average dielectric strength historical data, average thermal resistance historical data, average comprehensive mechanical strength historical data, average insulation thickness historical data, and average frequency response historical data;
[0222] Preprocess the historical data to obtain normalized historical data;
[0223] Perform feature mapping encoding on the normalized historical data to obtain a normalized data set;
[0224] The normalized data set is divided into a training set, a validation set, and a test set according to a preset data set segmentation ratio, and an initial model of the electric field strength optimization strategy is initially constructed;
[0225] Add adversarial disturbances to the sample data of the training set and the validation set, and train the initial model of the electric field strength optimization strategy according to the training set and the validation set to obtain the basic model of the electric field strength optimization strategy;
[0226] Use the test set to perform performance testing on the basic model of the electric field strength optimization strategy;
[0227] If the test result of the performance test meets the preset model test pass condition, the electric field strength optimization strategy basic model is used as the electric field strength optimization strategy model.
[0228] Further, based on the above-mentioned embodiment of the invention, the initial model of the electric field strength optimization strategy is trained according to the training set and the verification set to obtain the basic model of the electric field strength optimization strategy, including:
[0229] The sample data in the training set are respectively input into the convolutional neural network layer of the initial model of the electric field strength optimization strategy, and the sample feature representation corresponding to each sample data is extracted;
[0230] Input each sample feature representation into the attention layer of the initial model of the electric field strength optimization strategy to obtain the sample attention matrix corresponding to each sample feature representation;
[0231] Use each sample attention matrix as the state in the state space corresponding to the target network and the evaluation network of the initial model of the electric field strength optimization strategy, and obtain the action space and reward function of the initial model configuration of the corresponding electric field strength optimization strategy;
[0232] Select a state from the state space as the current state, and select the current action in the action space based on the current state using a preset dynamic exploration strategy to obtain an updated state according to the interaction between the current action and the environment;
[0233] Determine an update action based on the update state through the evaluation network; wherein the update action is used to interact with the environment to obtain a new update state;
[0234] Determine the actual Q value corresponding to the current state and current action through the evaluation network and reward function;
[0235] Determine the target Q value corresponding to the update state and update action through the target network;
[0236] Update the network parameters of the evaluation network according to the loss function values corresponding to the actual Q value and the target Q value;
[0237] Copying the network parameters of the evaluation network to the target network at every preset time period;
[0238] When the number of model iterations of the initial model of the electric field strength optimization strategy reaches a preset number of iterations, or the loss function value reaches a preset threshold, the current initial model of the electric field strength optimization strategy is used as the target model of the electric field strength optimization strategy;
[0239] The electric field strength optimization strategy target model is verified using the verification set, and the model parameters are adjusted until the electric field strength optimization strategy target model meets the preset model verification indicators to obtain the trained electric field strength optimization strategy basic model.
[0240] Further, based on the above-mentioned embodiments of the invention, the actions in the action space include at least one of the following: adjustment parameters for adjusting the average dielectric strength, adjustment parameters for adjusting the average thermal resistance, adjustment parameters for adjusting the average comprehensive mechanical strength, adjustment parameters for adjusting the average insulation thickness, and adjustment parameters for adjusting the average frequency response.
[0241] Further, on the basis of the above-mentioned embodiment of the invention, the initial model of the electric field strength optimization strategy is trained according to the training set and the verification set to obtain the basic model of the electric field strength optimization strategy, which also includes:
[0242] The priority of sample data in the training set and validation set is adjusted based on the priority experience replay mechanism.
[0243] The switch cabinet electric field strength optimization strategy determination device provided in the embodiment of the present invention can execute the switch cabinet electric field strength optimization strategy determination method provided in any embodiment of the present invention, and has the corresponding functional modules and beneficial effects of the execution method.
[0244] Embodiment 5
[0245] Fig.11 A schematic diagram of an electronic device 50 that can be used to implement an embodiment of the present invention is shown. The electronic device is intended to represent various forms of digital computers, such as laptop computers, desktop computers, workstations, personal digital assistants, servers, blade servers, mainframe computers, and other suitable computers. The electronic device can also represent various forms of mobile devices, such as personal digital processing, cellular phones, smart phones, wearable devices (such as helmets, glasses, watches, etc.) and other similar computing devices. The components shown herein, their connections and relationships, and their functions are merely examples and are not intended to limit the implementation of the present invention described and / or required herein.
[0246] like Fig.11As shown, the electronic device 50 includes at least one processor 51, and a memory connected to the at least one processor 51 in communication, such as a read-only memory (ROM) 52, a random access memory (RAM) 53, etc., wherein the memory stores a computer program that can be executed by at least one processor, and the processor 51 can perform various appropriate actions and processes according to the computer program stored in the read-only memory (ROM) 52 or the computer program loaded from the storage unit 58 to the random access memory (RAM) 53. In the RAM 53, various programs and data required for the operation of the electronic device 50 can also be stored. The processor 51, the ROM 52, and the RAM 53 are connected to each other through a bus 54. An input / output (I / O) interface 55 is also connected to the bus 54.
[0247] A number of components in the electronic device 50 are connected to the I / O interface 55, including: an input unit 56, such as a keyboard, a mouse, etc.; an output unit 57, such as various types of displays, speakers, etc.; a storage unit 58, such as a disk, an optical disk, etc.; and a communication unit 59, such as a network card, a modem, a wireless communication transceiver, etc. The communication unit 59 allows the electronic device 50 to exchange information / data with other devices through a computer network such as the Internet and / or various telecommunication networks.
[0248] The processor 51 may be any general and / or dedicated processing component with processing and computing capabilities. Some examples of the processor 51 include, but are not limited to, a central processing unit (CPU), a graphics processing unit (GPU), various dedicated artificial intelligence (AI) computing chips, various processors running machine learning model algorithms, a digital signal processor (DSP), and any appropriate processor, controller, microcontroller, etc. The processor 51 executes the various methods and processes described above, such as a method for determining a switch cabinet electric field strength optimization strategy.
[0249] In some embodiments, the switch cabinet electric field strength optimization strategy determination method can be implemented as a computer program, which is tangibly contained in a computer-readable storage medium, such as a storage unit 58. In some embodiments, part or all of the computer program can be loaded and / or installed on the electronic device 50 via the ROM 52 and / or the communication unit 59. When the computer program is loaded into the RAM 53 and executed by the processor 51, one or more steps of the switch cabinet electric field strength optimization strategy determination method described above can be executed. Alternatively, in other embodiments, the processor 51 can be configured to execute the switch cabinet electric field strength optimization strategy determination method in any other appropriate manner (for example, by means of firmware).
[0250] Various implementations of the systems and techniques described above herein can be implemented in digital electronic circuit systems, integrated circuit systems, field programmable gate arrays (FPGAs), application specific integrated circuits (ASICs), application specific standard products (ASSPs), systems on chips (SOCs), load programmable logic devices (CPLDs), computer hardware, firmware, software, and / or combinations thereof. These various implementations can include: being implemented in one or more computer programs that can be executed and / or interpreted on a programmable system including at least one programmable processor, which can be a special purpose or general purpose programmable processor that can receive data and instructions from a storage system, at least one input device, and at least one output device, and transmit data and instructions to the storage system, the at least one input device, and the at least one output device.
[0251] Computer programs for implementing the methods of the present invention may be written in any combination of one or more programming languages. These computer programs may be provided to a processor of a general-purpose computer, a special-purpose computer, or other programmable data processing device, so that when the computer program is executed by the processor, the functions / operations specified in the flow chart and / or block diagram are implemented. The computer program may be executed entirely on the machine, partially on the machine, partially on the machine and partially on a remote machine as a stand-alone software package, or entirely on a remote machine or server.
[0252] In the context of the present invention, a computer-readable storage medium may be a tangible medium that may contain or store a computer program for use by or in conjunction with an instruction execution system, device, or equipment. A computer-readable storage medium may include, but is not limited to, electronic, magnetic, optical, electromagnetic, infrared, or semiconductor systems, devices, or equipment, or any suitable combination of the foregoing. Alternatively, a computer-readable storage medium may be a machine-readable signal medium. A more specific example of a machine-readable storage medium may include an electrical connection based on one or more lines, a portable computer disk, a hard disk, a random access memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory (EPROM or flash memory), an optical fiber, a portable compact disk read-only memory (CD-ROM), an optical storage device, a magnetic storage device, or any suitable combination of the foregoing.
[0253] To provide interaction with a user, the systems and techniques described herein may be implemented on an electronic device having: a display device (e.g., a CRT (cathode ray tube) or LCD (liquid crystal display) monitor) for displaying information to the user; and a keyboard and a pointing device (e.g., a mouse or trackball) through which the user can provide input to the electronic device. Other types of devices may also be used to provide interaction with the user; for example, the feedback provided to the user may be any form of sensory feedback (e.g., visual feedback, auditory feedback, or tactile feedback); and input from the user may be received in any form (including acoustic input, voice input, or tactile input).
[0254] The systems and techniques described herein may be implemented in a computing system that includes backend components (e.g., as a data server), or a computing system that includes middleware components (e.g., an application server), or a computing system that includes frontend components (e.g., a user computer with a graphical user interface or a web browser through which a user can interact with implementations of the systems and techniques described herein), or a computing system that includes any combination of such backend components, middleware components, or frontend components. The components of the system may be interconnected by any form or medium of digital data communication (e.g., a communication network). Examples of communication networks include: a local area network (LAN), a wide area network (WAN), a blockchain network, and the Internet.
[0255] A computing system may include a client and a server. The client and the server are generally remote from each other and usually interact through a communication network. The client and server relationship is generated by computer programs running on the corresponding computers and having a client-server relationship with each other. The server may be a cloud server, also known as a cloud computing server or cloud host, which is a host product in the cloud computing service system to solve the defects of difficult management and weak business scalability in traditional physical hosts and VPS services.
[0256] It should be understood that the various forms of processes shown above can be used to reorder, add or delete steps. For example, the steps described in the present invention can be executed in parallel, sequentially or in different orders, as long as the desired results of the technical solution of the present invention can be achieved, and this document does not limit this.
[0257] The above specific implementations do not constitute a limitation on the protection scope of the present invention. It should be understood by those skilled in the art that various modifications, combinations, sub-combinations and substitutions can be made according to design requirements and other factors. Any modification, equivalent substitution and improvement made within the spirit and principle of the present invention should be included in the protection scope of the present invention.
Claims
1. A method for determining a switch cabinet electric field strength optimization strategy, characterized in that: The method comprises: Obtain target data of a target component in a switch cabinet; wherein the target data includes at least one of the following: average dielectric strength, average thermal resistance, average comprehensive mechanical strength, average insulation thickness, and average frequency response; the target component includes at least one of the following switch components: a load switch, a circuit breaker switch, and a grounding switch; The target data is input into the electric field strength optimization strategy model to obtain a corresponding optimization strategy; wherein the optimization strategy includes at least one of the following: an adjusted parameter of average dielectric strength, an adjusted parameter of average thermal resistance, an adjusted parameter of average comprehensive mechanical strength, an adjusted parameter of average insulation thickness and an adjusted parameter of average frequency response; the electric field strength optimization strategy model is obtained by adversarial training of a dual deep Q network based on an attention mechanism.
2. The method for determining the switch cabinet electric field strength optimization strategy according to claim 1, characterized in that: The average dielectric strength is the average value of the dielectric strengths corresponding to a plurality of measuring points on the target component, and the dielectric strength is determined by the breakdown voltage collected at the corresponding measuring points and the thickness of the insulating layer of the target component, and the breakdown voltage and the thickness of the insulating layer of the target component are collected and obtained by an electrode testing device and a thickness measuring device, respectively; The average thermal resistance is the average value of the thermal resistances corresponding to multiple measuring points on the target component, and the thermal resistance is determined by the heating power applied to the corresponding measuring points and the collected temperature difference, the heating power is collected by a thermal testing device, and the initial temperature and final temperature corresponding to the temperature difference are collected by an infrared thermal imager; The average comprehensive mechanical strength is the average value of the comprehensive mechanical strengths corresponding to a plurality of measuring points on the target component, and the comprehensive mechanical strength is determined by the actual stress and electric field strength collected at the corresponding measuring points, and the actual stress and the electric field strength are collected and obtained by the stress sensor and the electric field sensor respectively; The average insulation thickness is the average value of the insulation layer thickness of the target component corresponding to multiple measurement points on the target component; The average frequency response is an average value of frequency responses corresponding to multiple measurement points on the target component, and the frequency response is acquired by a frequency response analyzer.
3. The method for determining the switch cabinet electric field strength optimization strategy according to claim 1, characterized in that: The electric field strength optimization strategy model includes: a convolutional neural network layer, an attention layer, a target network and an evaluation network; The step of inputting the target data into the electric field strength optimization strategy model to obtain a corresponding optimization strategy includes: Preprocessing the target data to obtain normalized target data, and inputting the normalized target data into the convolutional neural network layer to extract target feature representation corresponding to the normalized target data; Input the target feature representation into the attention layer, and output the target attention matrix corresponding to the target feature representation; The target attention matrix is processed using the target network and the evaluation network, and the parameters to be adjusted corresponding to each of the target data are output, and each of the parameters to be adjusted is used as the optimization strategy.
4. The method for determining the switch cabinet electric field strength optimization strategy according to claim 1, characterized in that: Before acquiring the target data of the target component in the switch cabinet, a training process of the electric field strength optimization strategy model is also included, and the training process includes: Acquire historical data of the target component; wherein the historical data includes at least one of the following: average dielectric strength historical data, average thermal resistance historical data, average comprehensive mechanical strength historical data, average insulation thickness historical data, and average frequency response historical data; Preprocessing the historical data to obtain normalized historical data; Performing feature mapping encoding on the normalized historical data to obtain a normalized data set; Dividing the normalized data set into a training set, a validation set, and a test set according to a preset data set segmentation ratio, and initially constructing an initial model of the electric field strength optimization strategy; Adding adversarial disturbances to the sample data of the training set and the verification set, and training the initial model of the electric field strength optimization strategy according to the training set and the verification set to obtain a basic model of the electric field strength optimization strategy; Using the test set to perform performance testing on the basic model of the electric field strength optimization strategy; If the test result of the performance test meets the preset model test pass condition, the electric field strength optimization strategy basic model is used as the electric field strength optimization strategy model.
5. The method for determining the switch cabinet electric field strength optimization strategy according to claim 4, characterized in that: The electric field strength optimization strategy initial model is trained according to the training set and the verification set to obtain the electric field strength optimization strategy basic model, including: Inputting the sample data in the training set into the convolutional neural network layer of the initial model of the electric field strength optimization strategy respectively, and extracting the sample feature representation corresponding to each of the sample data; Inputting each of the sample feature representations into the attention layer of the initial model of the electric field strength optimization strategy respectively, to obtain a sample attention matrix corresponding to each of the sample feature representations; Using each of the sample attention matrices as the state in the state space corresponding to the target network and the evaluation network of the initial model of the electric field strength optimization strategy, and obtaining the action space and reward function corresponding to the configuration of the initial model of the electric field strength optimization strategy; Selecting a state from the state space as the current state, and selecting a current action in the action space based on the current state using a preset dynamic exploration strategy, so as to obtain an updated state according to the interaction between the current action and the environment; Determining an update action according to the update state through the evaluation network; wherein the update action is used to interact with the environment to obtain a new update state; Determine the actual Q value corresponding to the current state and the current action through the evaluation network and the reward function; Determine the target Q value corresponding to the update state and the update action through the target network; Updating the network parameters of the evaluation network according to the loss function values corresponding to the actual Q value and the target Q value; Copying the network parameters of the evaluation network to the target network at intervals of a preset time period; When the model iteration number of the electric field strength optimization strategy initial model reaches a preset iteration number, or the loss function value reaches a preset threshold, the current electric field strength optimization strategy initial model is used as the electric field strength optimization strategy target model; The electric field strength optimization strategy target model is verified using the verification set, and model parameters are adjusted until the electric field strength optimization strategy target model meets preset model verification indicators to obtain the trained electric field strength optimization strategy basic model.
6. The method for determining the switch cabinet electric field strength optimization strategy according to claim 5, characterized in that: The actions in the action space include at least one of the following: an adjustment parameter for adjusting average dielectric strength, an adjustment parameter for adjusting average thermal resistance, an adjustment parameter for adjusting average comprehensive mechanical strength, an adjustment parameter for adjusting average insulation thickness, and an adjustment parameter for adjusting average frequency response.
7. The method for determining the switch cabinet electric field strength optimization strategy according to claim 4, characterized in that: The training of the electric field strength optimization strategy initial model according to the training set and the verification set to obtain the electric field strength optimization strategy basic model also includes: The priorities of the sample data in the training set and the validation set are adjusted based on a priority experience replay mechanism.
8. A device for determining an optimization strategy for electric field strength of a switch cabinet, characterized in that: The device comprises: A data acquisition module, used to acquire target data of a target component in a switch cabinet; wherein the target data includes at least one of the following: average dielectric strength, average thermal resistance, average comprehensive mechanical strength, average insulation thickness and average frequency response; the target component includes at least one of the following switch components: a load switch, a circuit breaker switch and a grounding switch; An optimization strategy determination module is used to input the target data into an electric field strength optimization strategy model to obtain a corresponding optimization strategy; wherein the optimization strategy includes at least one of the following: an adjusted parameter of average dielectric strength, an adjusted parameter of average thermal resistance, an adjusted parameter of average comprehensive mechanical strength, an adjusted parameter of average insulation thickness, and an adjusted parameter of average frequency response; the electric field strength optimization strategy model is obtained by adversarial training of a dual deep Q network based on an attention mechanism.
9. An electronic device, characterized in that: The electronic device comprises: at least one processor; and a memory communicatively connected to the at least one processor; wherein, The memory stores a computer program executable by the at least one processor, and the computer program is executed by the at least one processor so that the at least one processor can execute the method for determining the switch cabinet electric field strength optimization strategy according to any one of claims 1 to 7.
10. A computer-readable storage medium, characterized in that: The computer-readable storage medium stores computer instructions, and the computer instructions are used to enable a processor to implement the method for determining the switch cabinet electric field strength optimization strategy according to any one of claims 1 to 7 when executed.
Citation Information
Patent Citations
Micro-grid operation optimization scheduling method based on depth deterministic strategy gradient
CN116345578A
Method of training model, device, and storage medium
US20220114343A1
Cited By
Thermal management system of TMBS semiconductor device
CN119538594A
Method for optimizing magnetic thermal stress coupling performance of bus duct in complex electromagnetic environment
CN120706157A
Bus duct magnetic-thermal stress coupling performance optimization method in complex electromagnetic environment
CN120706157B
