An oil-immersed transformer optimal design method for flexible ring closing and related device

Abstract

The application belongs to the field of transformer optimization design, and discloses an oil-immersed transformer optimization design method for flexible ring closing and related devices, which comprises the following steps: determining the to-be-optimized variables and their value ranges of a target transformer according to the electrical parameter characteristics and thermal performance requirements of the target transformer; constructing an optimization design function of the target transformer based on the to-be-optimized variables of the target transformer and combining a multi-objective optimization theory; taking the minimization of total power consumption as the target while meeting the hotspot temperature constraint; solving the optimization design function of the target transformer by using an improved particle swarm optimization algorithm based on the to-be-optimized variables and their value ranges of the target transformer, and outputting the optimization design result of the target transformer; and the application can quickly obtain a globally optimal design scheme with the lowest total power consumption under the condition of meeting the severe heat dissipation requirement, thereby significantly improving the operation efficiency and safety and reliability of the oil-immersed transformer under the flexible ring closing working condition.

Description

Technical Field

[0001] This invention belongs to the field of transformer optimization design technology, and specifically relates to an optimization design method and related apparatus for oil-immersed transformers with flexible loop closure. Background Technology

[0002] Flexible loop connection is a power conversion device based on power electronics technology. It is used to connect two AC power grids to achieve interconnection of power grids with different voltage levels and frequencies. It can also achieve load balancing, optimize power grid power supply capacity, provide dynamic reactive power support, and improve the reliability of distribution networks and equipment utilization.

[0003] Oil-immersed transformers, as a core component of flexible closed-loop systems, operate under complex conditions involving high-frequency harmonics and voltage fluctuations due to the large number of power electronic converters contained within the flexible closed-loop system. These complex conditions significantly enhance the skin effect and proximity effect, substantially increasing the additional losses in the windings and stray losses in the core, thus raising the risk of transformer overheating. When using traditional design methods to design the parameters of oil-immersed transformers, it is difficult to simultaneously meet stringent heat dissipation requirements and address the efficiency reduction caused by harmonic losses. Therefore, it is necessary to optimize the design parameters of oil-immersed transformers in flexible closed-loop systems.

[0004] Currently, the optimization design of oil-immersed transformers in flexible closed loops mainly adopts trial-and-error methods based on finite element simulation and traditional heuristic algorithms. However, the above-mentioned existing methods generally suffer from problems such as long calculation time for hot spot temperature and easy getting trapped in local optima and slow convergence.

[0005] Specifically, while the trial-and-error method based on finite element simulation has high computational accuracy, the calculation of a single field coupling is extremely time-consuming, making it difficult to support thousands of iterations in the optimization process, resulting in long calculation times for hot spot temperatures. Traditional heuristic algorithms, when dealing with transformer problems with strong nonlinearity and multiple constraints, lack the guidance of prior physical knowledge. The randomly generated initial population contains a large number of infeasible solutions, such as geometric size conflicts and excessive temperature rise, leading to low algorithm search efficiency and a tendency to converge prematurely to local optima, thus failing to obtain the truly globally optimal design scheme. Summary of the Invention

[0006] To address the technical problems existing in the prior art, this invention provides an optimization design method and related apparatus for oil-immersed transformers with flexible loop closure, in order to solve the technical problems that existing methods generally suffer from long hot spot temperature calculation time and are prone to getting trapped in local optima and slow convergence.

[0007] To achieve the above objectives, the technical solution adopted by the present invention is as follows: This invention provides an optimized design method for oil-immersed transformers with flexible loop closure, comprising: Based on the electrical parameter characteristics and thermal performance requirements of the target transformer, the variables to be optimized and their value ranges for the target transformer are determined. Based on the variables to be optimized for the target transformer, and combined with multi-objective optimization theory, an optimization design function for the target transformer is constructed. The optimization design function for the target transformer aims to minimize the total power consumption while satisfying the hot spot temperature constraint. Based on the variables to be optimized of the target transformer and their value range, the optimization design function of the target transformer is solved using an improved particle swarm optimization algorithm, and the optimization design result of the target transformer is output. In the improved particle swarm optimization algorithm, a population initialization strategy guided by physical constraints is used to generate the particle swarm, and the feasibility of current density constraints and magnetic flux density constraints is checked and repaired for each particle in the particle swarm. In addition, the improved particle swarm optimization algorithm also embeds a pre-built transformer hot spot temperature prediction model, and designs an adaptive inertia weight and a dynamic time-varying learning factor based on the current iteration progress and population diversity.

[0008] Furthermore, the variables to be optimized for the target transformer include at least one of the following: core diameter, number of turns in the low-voltage winding, thickness of the high-voltage conductor, width of the high-voltage conductor, thickness of the low-voltage conductor, and width of the low-voltage conductor.

[0009] Furthermore, the optimization design function for the target transformer is as follows:

[0010]

[0011]

[0012] in, The optimization design function for the target transformer; The variables to be optimized are the target transformer; The objective function is the total loss. For hotspot temperature constraint functions; The highest permissible hot spot temperature; This refers to the no-load loss of the iron core; For winding copper loss; For a pre-built transformer hotspot temperature prediction model; The x-axis represents the location of the hotspot monitoring point; The vertical coordinate represents the location of the hotspot monitoring point; The intensity of the iron core heat source; The heat source intensity of the high-voltage winding; The heat source intensity for the low-voltage winding.

[0013] Furthermore, the pre-constructed transformer hotspot temperature prediction model is a pre-trained physical information neural network model that incorporates an attention mechanism. Among them, the pre-trained physical information neural network model with attention mechanism combines data-driven learning and physical constraints, and incorporates the residual of the steady-state heat conduction equation into the loss function through automatic differentiation technology.

[0014] Furthermore, the process of performing a feasibility check on current density constraints and magnetic flux density constraints for each particle in the particle swarm is as follows:

[0015] in, This serves as a feasibility check and determination marker for particles; The variables to be optimized are the target transformer; This represents the high-voltage winding current density corresponding to the particle. The maximum allowable current density; This represents the low-voltage winding current density corresponding to the particle. The magnetic flux density of the iron core corresponding to the particle; This represents the maximum permissible magnetic flux density.

[0016] Furthermore, the repair process for each particle in the particle swarm is as follows: If the magnetic flux density of the iron core exceeds the preset magnetic flux density saturation value, then the diameter of the iron core is increased. If the cross-sectional area of ​​the high-voltage or low-voltage conductor is insufficient, the cross-sectional area of ​​the high-voltage or low-voltage conductor shall be enlarged proportionally.

[0017] Furthermore, the adaptive inertia weights based on the current iteration progress and population diversity are as follows:

[0018] in, This is an adaptive inertial weight based on the current iteration progress and population diversity; For the weighting factor; Assigning weights to schedules; For diversity weight; The dynamic time-varying learning factor is as follows:

[0019] in, For individual learning factors; As a social learning factor; This is the initial value of the individual learning factor; This is the termination value of the individual learning factor; This represents the current iteration number; This represents the maximum number of iterations. This is the initial value for the social learning factor; This is the termination value for the social learning factor.

[0020] This invention also provides an optimized design system for oil-immersed transformers with flexible loop closure, comprising: The variable determination module is used to determine the variables to be optimized for the target transformer and their value range based on the electrical parameter characteristics and thermal performance requirements of the target transformer. The function construction module is used to construct the optimization design function of the target transformer based on the variables to be optimized of the target transformer and combined with multi-objective optimization theory. The optimization design function of the target transformer aims to minimize the total power consumption while satisfying the hot spot temperature constraint. The optimization solution module is used to solve the optimization design function of the target transformer based on the variables to be optimized and their value range, using an improved particle swarm optimization algorithm, and outputs the optimization design result of the target transformer. In the improved particle swarm optimization algorithm, a population initialization strategy guided by physical constraints is used to generate the particle swarm, and the feasibility of current density constraints and magnetic flux density constraints is checked and repaired for each particle in the particle swarm. In addition, the improved particle swarm optimization algorithm also embeds a pre-built transformer hot spot temperature prediction model, and designs an adaptive inertia weight and a dynamic time-varying learning factor based on the current iteration progress and population diversity.

[0021] The present invention also provides an electronic device, comprising: A processor is used to execute computer programs; A computer-readable storage medium storing a computer program, which, when executed by the processor, performs the aforementioned optimized design method for oil-immersed transformers with flexible loop closure.

[0022] The present invention also provides a computer-readable storage medium storing a computer program that, when executed by a processor, implements the aforementioned optimized design method for oil-immersed transformers with flexible loop closure.

[0023] Compared with the prior art, the beneficial effects of the present invention are as follows: The optimization design method for oil-immersed transformers in flexible loop conditions provided by this invention can achieve a synergistic improvement in hotspot temperature calculation efficiency, optimization convergence speed, and global optimal solution acquisition capability while meeting stringent heat dissipation requirements. This allows for the rapid acquisition of the globally optimal design scheme with the lowest total power consumption, thereby significantly improving the operating efficiency and safety reliability of oil-immersed transformers under flexible loop conditions. Specifically, by determining the variables to be optimized and their value ranges based on the electrical parameter characteristics and thermal performance requirements of the target transformer, a reasonable search boundary is defined for the optimization process, reducing invalid searches from the source. Secondly, the optimization design function aims to minimize total power consumption while taking into account hotspot temperature constraints, precisely meeting the core requirement of balancing heat dissipation and efficiency for oil-immersed transformers in flexible loop conditions under high-frequency harmonics and voltage fluctuations. This thoroughly solves the problem of traditional designs struggling to balance heat dissipation and efficiency from the optimization objective level. Furthermore, the improved particle swarm optimization algorithm breaks through the limitations of traditional methods in principle. Overcoming the limitations of existing algorithms, its physically constrained population initialization strategy effectively eliminates infeasible solutions in the initial population through feasibility checks and repairs of current density and magnetic flux density, significantly improving search efficiency and avoiding geometric size conflicts and excessive temperature rise issues. The embedded pre-built hotspot temperature prediction model replaces the time-consuming finite element field coupling calculation, significantly shortening the hotspot temperature calculation time and supporting high-frequency iterations in the optimization process. The adaptive inertial weights and dynamic time-varying learning factors designed based on iteration progress and population diversity can dynamically adjust the particle search strategy, effectively preventing the algorithm from converging to local optima too early, ensuring that the final output optimization design result is globally optimal. This satisfies the stringent heat dissipation requirements under the complex working conditions of flexible closed loops, minimizes the efficiency decline caused by harmonic losses, and improves the convergence speed of the optimization algorithm, thereby significantly improving the operational reliability, efficiency, and design rationality of oil-immersed transformers in flexible closed loops.

[0024] Furthermore, the core diameter, number of turns in the high and low voltage windings, thickness and width of the high voltage conductor, and thickness and width of the low voltage conductor are taken as the variables to be optimized for the target transformer. This ensures that the variables to be optimized cover the core influencing parameters of oil-immersed transformer loss, temperature rise, magnetic flux, and current density. The optimization dimensions are comprehensive and highly targeted, enabling coordinated control of core loss, winding copper loss, and hot spot temperature from the source, and ensuring that the optimization results can directly support the implementation of engineering design.

[0025] Furthermore, an optimization design function is constructed with the goal of minimizing total power consumption while satisfying hot spot temperature constraints. This function combines the no-load loss of the iron core, the copper loss of the winding, and the pre-constructed transformer hot spot temperature prediction model to achieve a multi-objective balance between optimal loss and safe temperature rise. This function not only accurately adapts to the operating characteristics of high harmonic loss and high overheating risk under the complex working conditions of flexible closed loop, but also maximizes the transformer operating efficiency while meeting stringent heat dissipation and safety requirements. This effectively solves the technical contradiction that traditional designs cannot take into account both loss and heat dissipation.

[0026] Furthermore, a pre-trained physical information neural network model with an attention mechanism is used as a pre-built transformer hotspot temperature prediction model. Combining data-driven learning and physical constraints, the residual of the steady-state heat conduction equation is incorporated into the loss function through automatic differentiation. Compared with finite element simulation, this significantly shortens the hotspot temperature calculation time while ensuring calculation accuracy. It can support tens of thousands of iterations of optimization algorithms and can balance computational efficiency and reliability.

[0027] Furthermore, by using the feasibility judgment rules of the current density of the high-voltage winding, the low-voltage winding, and the magnetic flux density of the iron core, it is possible to quickly identify whether the particle solution meets the electromagnetic safety constraints, achieve efficient screening of infeasible solutions, significantly reduce invalid iterations, improve the overall optimization efficiency of the algorithm, and avoid safety hazards such as current overload and iron core magnetic saturation during the optimization process.

[0028] Furthermore, a repair strategy is adopted that increases the core diameter when the magnetic flux density exceeds the standard and proportionally enlarges the conductor cross-sectional area when it is insufficient. This achieves directional repair based on physical mechanisms, ensuring that the particle solution quickly returns to the feasible region, maintaining the rationality and engineering feasibility of the parameter design, and avoiding parameter conflicts and design failures caused by random repair.

[0029] Furthermore, by adopting an adaptive inertial weight that integrates iterative progress and population diversity, the algorithm can dynamically balance its global search and local development capabilities. Combined with dynamically time-varying individual learning factors and social learning factors, the algorithm can autonomously adjust its search strategy according to the optimization process, effectively avoiding premature convergence and getting stuck in local optima. This significantly improves the convergence speed and the ability to obtain the global optimal solution, ensuring the optimality and stability of the optimization results.

[0030] The oil-immersed transformer optimization design system, electronic device, and computer-readable storage medium provided by this invention possess all the advantages of the aforementioned oil-immersed transformer optimization design method for flexible loop closure. Attached Figure Description

[0031] To more clearly illustrate the technical solutions in the embodiments of this application or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0032] Figure 1 A flowchart of the optimized design method for a flexible closed-loop oil-immersed transformer provided by the present invention; Figure 2 A flowchart of the optimized design method for a flexible loop-closed oil-immersed transformer provided in Example 1; Figure 3 This is an architecture diagram of the physical information neural network model incorporating an attention mechanism in Example 1; Figure 4 This is a structural block diagram of the oil-immersed transformer optimization design system for flexible loop closure provided in Example 2; Figure 5 This is a structural block diagram of the electronic device provided in Example 3. Detailed Implementation

[0033] To make the technical problems, technical solutions, and beneficial effects solved by this application clearer, the technical solutions in the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only a part of the embodiments of this application, and not all of them. All other embodiments obtained by those skilled in the art based on the embodiments of this application without creative effort are within the scope of protection of this application.

[0034] As attached Figure 1 As shown, this invention provides an optimized design method for an oil-immersed transformer with flexible loop closure, comprising the following steps: Step 100: Based on the electrical parameter characteristics and thermal performance requirements of the target transformer, determine the variables to be optimized for the target transformer and their value ranges.

[0035] Step 200: Based on the variables to be optimized of the target transformer, and combined with multi-objective optimization theory, construct the optimization design function of the target transformer; wherein, the optimization design function of the target transformer aims to minimize the total power consumption while satisfying the hot spot temperature constraint.

[0036] Step 300: Based on the variables to be optimized of the target transformer and their value range, the optimization design function of the target transformer is solved using the improved particle swarm optimization algorithm, and the optimization design result of the target transformer is output.

[0037] It is worth noting that in the improved particle swarm optimization algorithm, a population initialization strategy guided by physical constraints is used to generate the particle swarm, and the feasibility of current density constraints and magnetic flux density constraints is checked and repaired for each particle in the particle swarm. Furthermore, the improved particle swarm optimization algorithm also incorporates a pre-built transformer hotspot temperature prediction model, and designs adaptive inertia weights and dynamic time-varying learning factors based on the current iteration progress and population diversity.

[0038] In the above embodiments, by embedding a pre-built transformer hotspot temperature prediction model to replace the traditional time-consuming finite element field coupling calculation, the time cost of a single iteration is significantly reduced from the perspective of computational principle, effectively solving the problem of long calculation time for hotspot temperature. At the same time, by using a population initialization strategy guided by physical constraints and the feasibility check and repair of current density and magnetic flux density, physical prior knowledge is integrated into the algorithm search process, reducing the generation of infeasible solutions from the source and significantly improving search efficiency. In addition, by combining adaptive inertial weights based on iteration progress and population diversity with dynamic time-varying learning factors, the algorithm's global exploration and local development capabilities are dynamically balanced, avoiding the defects of traditional heuristic algorithms that are prone to getting trapped in local optima and slow convergence. In summary, the oil-immersed transformer optimization design method for flexible loop closure described in this invention can quickly obtain the globally optimal design scheme with the lowest total power consumption while meeting stringent heat dissipation requirements, significantly improving the operating efficiency and safety reliability of oil-immersed transformers under flexible loop closure conditions.

[0039] The following specific embodiments further explain the optimization design method for oil-immersed transformers with flexible loop closure provided by the present invention: Example 1 As attached Figure 2 As shown in the figure, this embodiment 1 provides an optimized design method for an oil-immersed transformer with flexible loop closure, including the following steps: Step 1: Based on the electrical parameter characteristics and thermal performance requirements of the target transformer, determine the variables to be optimized and their value ranges. The variables to be optimized include at least one of the following: core diameter, number of turns in the low-voltage winding, thickness of the high-voltage conductor, width of the high-voltage conductor, thickness of the low-voltage conductor, and width of the low-voltage conductor.

[0040] Taking a typical oil-immersed transformer as an example, the range of values ​​for the variable to be optimized is as follows: The core diameter ranges from [150mm, 200mm], the number of turns in the low-voltage winding ranges from [40, 60], the thickness of the high-voltage conductor ranges from [1.9mm, 2.2mm], the width of the high-voltage conductor ranges from [2.0mm, 2.2mm], the thickness of the low-voltage conductor ranges from [2.2mm, 3.0mm], and the width of the low-voltage conductor ranges from [7.0mm, 7.5mm].

[0041] Step 2: Based on the variables to be optimized for the target transformer, and in conjunction with multi-objective optimization theory, construct the optimization design function for the target transformer; wherein, the optimization design function for the target transformer aims to minimize the total power consumption while satisfying the hot spot temperature constraint.

[0042] Specifically, considering the dual requirements of performance and safety for the target transformer in actual operation, based on the variables to be optimized of the target transformer, minimizing total loss and meeting hot spot temperature constraints are taken as the core optimization objectives. According to the multi-objective optimization theory, a multi-objective optimization function is established to obtain the optimization design function of the target transformer.

[0043] The optimal design function for the target transformer is as follows:

[0044] in, The optimization design function for the target transformer; The variables to be optimized are the target transformer; The objective function is the total loss. The hot spot temperature constraint function is used to predict the hot spot temperature of the transformer using a pre-built transformer hot spot temperature prediction model. The highest permissible hot spot temperature.

[0045] Total loss objective function The mathematical description of is as follows:

[0046] in, This refers to the no-load loss of the iron core; This refers to the copper loss in the winding.

[0047] Core no-load loss The calculation formula is as follows:

[0048] in, The unit loss of the silicon steel sheets in the iron core is taken as 2.0 W / Kg; For the quality of the iron core; This is the cross-sectional area of ​​the iron core; The diameter of the iron core; The length of the iron core; This represents the core density.

[0049] Winding copper loss This includes high-voltage winding losses and low-voltage winding losses, i.e., winding losses. The calculation formula is as follows:

[0050] in, This refers to high-voltage winding losses; This refers to the low-voltage winding loss.

[0051] The mathematical expression for the hotspot temperature constraint function is as follows:

[0052] in, For a pre-built transformer hotspot temperature prediction model; The x-axis represents the location of the hotspot monitoring point; The vertical coordinate represents the location of the hotspot monitoring point; The intensity of the iron core heat source; The heat source intensity of the high-voltage winding; The heat source intensity for the low-voltage winding.

[0053] It should be noted that the optimized design of the target transformer needs to meet performance constraints, material constraints, and process constraints. Specifically, the performance constraints are: the total loss must be within a preset reasonable range, and the hot spot temperature must not exceed the design limit; the material constraints are: the conductor current and the core magnetic flux density must be within the safe operating range; and the process constraints are: the number of winding turns must be an integer, and the conductor specifications must meet the process requirements.

[0054] Step 3: Train the physical information neural network model with attention mechanism to obtain a pre-trained physical information neural network model with attention mechanism, which serves as a pre-constructed transformer hotspot temperature prediction model. The pre-trained physical information neural network model with attention mechanism combines data-driven learning and physical constraints, and incorporates the residuals of the steady-state heat conduction equation into the loss function through automatic differentiation techniques.

[0055] Specifically, the process of training a physical information neural network model that incorporates an attention mechanism is as follows: Step 31: Construct a physical information neural network model incorporating an attention mechanism. The network input of this model is a five-dimensional feature vector, including the coordinates of the hotspot monitoring location, the core heat source intensity, the high-voltage winding heat source intensity, and the low-voltage winding heat source intensity. The coordinates of the hotspot monitoring location include the horizontal and vertical coordinates. The network output of this model is the predicted hotspot temperature, i.e., the predicted temperature value at the hotspot monitoring location.

[0056] In this embodiment 1, the network structure of the physical information neural network model incorporating an attention mechanism includes an input layer, a pre-residual block, a lightweight attention module, a post-residual block, and an output layer, as shown in the attached figure. Figure 3 As shown.

[0057] The input layer receives a five-dimensional feature vector containing the coordinates of the hotspot monitoring location, the core heat source intensity, the high-voltage winding heat source intensity, and the low-voltage winding heat source intensity. Through feature mapping, the received five-dimensional feature vector is mapped to a high-dimensional space to obtain a high-dimensional feature input. The high-dimensional feature input is then normalized and activated by a GELU (Gaussian Error LinearUnit, a nonlinear activation function based on Gaussian error function) layer to perform standardization and nonlinear transformation, obtaining preprocessed features to accelerate network convergence.

[0058] The pre-processed residual block is used to perform preliminary deep feature extraction on the preprocessed features to obtain the preliminary extracted feature tensor; the pre-processed residual block includes a fast residual block a and a fast residual block b in series.

[0059] The lightweight attention module is used to capture global dependencies between features based on the initially extracted feature tensors through a multi-head mechanism, obtaining attention-weighted features. The lightweight attention module includes a multi-head attention block and a Softmax (soft maximum value function) normalization layer. Specifically, the process of the lightweight attention module is as follows:

[0060]

[0061] in, Features that are weighted for attention It is a soft maximum function; This is a query matrix containing query information for the current feature point; This is the key matrix, used to match the query matrix to calculate relevance weights; This is the transpose symbol for a matrix; Key matrix transpose; This is a value matrix containing the actual feature information to be weighted; For the attention head dimension, we set it to 64; For the initially extracted feature tensors; This is the joint transformation matrix.

[0062] The post-residual block is used to perform a second deep feature extraction on the attention-weighted features to obtain the second-extracted feature tensor. The post-residual block includes a fast residual block a and a fast residual block b in series. The output layer is used to map the second-extracted feature tensor to the temperature prediction value of the hotspot monitoring location by using the fully connected network dimensionality reduction block a, the GELU activation layer and the fully connected network dimensionality reduction block b in sequence, and outputs the hotspot temperature prediction value.

[0063] Step 32: Based on the predetermined transformer design variables and their value ranges, generate 50,000 training samples using the Latin hypercube sampling method; calculate the loss value and heat source intensity corresponding to each training sample; obtain the actual temperature field data of the transformer using the finite element simulation method; construct a training dataset including the coordinates of the hotspot monitoring location, the heat source intensity, and the temperature value of the hotspot monitoring location based on the loss value and heat source intensity corresponding to each training sample and the actual temperature field data of the transformer; divide the training dataset in an 8:1:1 ratio to obtain the training set, validation set, and test set.

[0064] Step 33: Combining data-driven learning and physical constraints, the residuals of the steady-state heat conduction equation are incorporated into the loss function through automatic differentiation technology, constructing a composite loss function that includes data loss and physical loss as the total loss function; wherein, through automatic differentiation technology, the partial derivative of temperature with respect to the coordinates of the temperature monitoring location is calculated to ensure that the prediction results of the physical information neural network model with attention mechanism conform to the physical laws of heat conduction.

[0065] The total loss function is as follows:

[0066] in, This is the total loss function; For data loss; The weighting coefficient for physical loss is set to 0.15; This is a physical loss.

[0067] It should be noted that in the total loss function, data loss is used to measure the deviation between the predicted hotspot temperature and the actual hotspot temperature, while physical loss is used to describe the mean square value of the parameters of the steady-state heat conduction equation.

[0068] Data loss is represented by mean squared error, as follows:

[0069] in, The number of samples in the training dataset; For the training dataset, the first Predicted hotspot temperatures for each sample; For the training dataset, the first The true value of the hotspot temperature for each sample.

[0070] The physical loss is calculated based on the residual of the steady-state heat conduction equation; taking a two-dimensional temperature field as an example, the steady-state heat conduction equation corresponding to the two-dimensional temperature field is as follows:

[0071]

[0072] in, For a two-dimensional temperature field, use the Laplace operator. The total heat source intensity of the two-dimensional temperature field; Thermal conductivity; This is a temperature field used to describe the temperature values ​​at different locations within a two-dimensional plane. The x-coordinate in the two-dimensional plane; The vertical coordinate is the coordinate in a two-dimensional plane.

[0073] Based on the steady-state heat conduction equation, the first and second partial derivatives of temperature with respect to the coordinates of the temperature monitoring location are calculated using PyTorch's automatic differentiation technique (an open-source deep learning framework) to obtain the physical loss. Taking a two-dimensional temperature field as an example, the physical loss corresponding to the two-dimensional temperature field is as follows:

[0074] in, This represents the number of physical constraint sampling points in the two-dimensional physical field. For the first Temperature field at each physically constrained sampling point; For the first The total heat source intensity of each physically constrained sampling point; The unit conversion factor for heat source intensity; For the first The x-coordinate of each physical constraint sampling point in a two-dimensional plane; For the first The horizontal coordinates of each physical constraint sampling point in a two-dimensional plane.

[0075] Step 34: Using the backpropagation algorithm and the Adam (Adaptive Moment Estimation, a widely used deep learning optimization algorithm) optimizer, combined with the training and validation sets, train the physical information neural network model with an attention mechanism. During training, monitor changes in the total loss function and update parameters accordingly. Training stops when the loss corresponding to the validation set no longer decreases significantly, and the optimal model parameters are saved, resulting in the pre-trained physical information neural network model with an attention mechanism.

[0076] Preferably, when using the Adam optimizer for training, the training epochs are set to 300, the initial learning rate is set to 0.0005, and the batch size is set to 512. Among them, a learning rate scheduling strategy is adopted, in which the learning rate decays by 30% when the loss corresponding to the validation set does not decrease for 25 consecutive epochs. The early stopping strategy is set to 60, that is, if the performance of the model does not improve within 60 epochs, training is stopped to save computing resources.

[0077] Step 35: Evaluate the performance of the pre-trained physical information neural network model with attention mechanism on the test set. Specifically, the performance of the pre-trained physical information neural network model with attention mechanism is verified by calculating the mean absolute error, root mean square error, and coefficient of determination to ensure that the model's prediction accuracy meets the requirements of engineering applications and achieves high-precision and rapid prediction of transformer hotspot temperatures. It should be noted that the smaller the values ​​of the mean absolute error and root mean square error, the better the model's performance; the closer the coefficient of determination is to 1, the better the model's performance.

[0078] Mean absolute error is defined as follows:

[0079] in, Mean absolute error; The number of samples in the test set; For the test set Predicted hotspot temperatures for each sample; For the test set The true value of the hotspot temperature for each sample.

[0080] The root mean square error is defined as follows:

[0081] in, This is the root mean square error.

[0082] The coefficient of determination is defined as follows:

[0083] in, The coefficient of determination.

[0084] Step 4: Based on the variables to be optimized of the target transformer and their value range, use the improved particle swarm optimization algorithm to solve the optimization design function of the target transformer and output the optimization design result of the target transformer.

[0085] Specifically, the steps are as follows: Step 41: Based on the variables to be optimized of the target transformer and their value range, a population initialization strategy guided by physical constraints is used to generate a particle swarm. Feasibility checks and repairs of current density constraints and magnetic flux density constraints are performed on each particle in the particle swarm to obtain the initialized population.

[0086] Specifically, based on the current density constraints and magnetic flux density constraints, particles are randomly generated within the range of values ​​of the variables to be optimized, forming a particle swarm; a feasibility check of the current density constraints and magnetic flux density constraints is performed on each particle in the particle swarm to obtain the feasibility check judgment mark of the particle; based on the feasibility check judgment mark of the particle, each particle in the particle swarm is repaired to obtain the initial population.

[0087] In this embodiment 1, the current density constraint is defined as follows:

[0088] in, This represents the current density of the high-voltage winding. This is the rated current of the high-voltage winding; This refers to the width of the high-voltage conductor; This refers to the thickness of the high-voltage conductor. This refers to the current density of the low-voltage winding. This is the rated current of the low-voltage winding; This refers to the width of the low-voltage conductor. This refers to the thickness of the low-voltage conductor.

[0089] In this embodiment 1, the magnetic flux density constraint is defined as follows:

[0090] in, The magnetic flux density of the iron core; This represents the main magnetic flux of the iron core. This represents the effective cross-sectional area of ​​the iron core.

[0091] In this embodiment 1, the process of performing feasibility checks on current density constraints and magnetic flux density constraints for each particle in the particle swarm, and obtaining the feasibility check judgment mark of the particle, is as follows:

[0092] in, This serves as a feasibility check and determination marker for particles; The variables to be optimized are the target transformer; This represents the high-voltage winding current density corresponding to the particle. The maximum allowable current density; This represents the low-voltage winding current density corresponding to the particle. The magnetic flux density of the iron core corresponding to the particle; This represents the maximum permissible magnetic flux density.

[0093] In this embodiment 1, the process of repairing each particle in the particle swarm based on the feasibility check and determination identifier of the particles to obtain the initial population includes: determining whether each particle in the particle swarm needs to be repaired according to the feasibility check and determination identifier of the particles; if so, determining whether the particles meet the preset physical constraints, and repairing the particles according to the determination result to ensure the high feasibility of the initial population; wherein, the preset physical constraints include current density constraints and magnetic flux density constraints.

[0094] It should be noted that if the current density constraint exceeds the limit, the cross-sectional area of ​​the high-voltage or low-voltage conductor is increased proportionally; if the magnetic flux density constraint exceeds the limit, the core diameter is increased. Specifically, if the magnetic flux density of the core exceeds the preset magnetic flux density saturation value, the core diameter is increased; if the cross-sectional area of ​​the high-voltage or low-voltage conductor is insufficient, the cross-sectional area of ​​the high-voltage or low-voltage conductor is enlarged proportionally. For example, if the magnetic flux density of the core exceeds the preset magnetic flux density saturation value, the core diameter is increased by 10% each time. After 50 attempts, 50 particles that meet the physical constraints are successfully initialized. Compared to the 40% feasible solution rate corresponding to traditional random initialization of particles, its feasible solution rate can be increased to 100%.

[0095] Step 42: Embed the pre-built transformer hotspot temperature prediction model into the particle swarm optimization algorithm, and design an adaptive inertia weight and a dynamic time-varying learning factor based on the current iteration progress and population diversity to obtain an improved particle swarm optimization algorithm.

[0096] In this Example 1, the calculation process of the adaptive inertia weight based on the current iteration progress and population diversity is as follows:

[0097] in, This is an adaptive inertial weight based on the current iteration progress and population diversity; As a weighting factor, we take 0.7; Assigning weights to schedules; For diversity weights.

[0098] Schedule weight The definition is as follows:

[0099] in, To achieve the maximum weight, we set it to 0.9; To minimize the weight, we set it to 0.4; This represents the current iteration number; The maximum number of iterations is set to 100.

[0100] Diversity weight The definition is as follows:

[0101]

[0102] in, For population diversity; It is a mean function; It is a function of standard deviation; To initialize the particles in the population.

[0103] In this embodiment 1, the dynamic learning factor is designed using a linear decreasing or increasing strategy to achieve a smooth transition from global exploration to local development in the algorithm; specifically, the definition of the dynamic learning factor is as follows:

[0104] in, For individual learning factors; As a social learning factor; The initial value for the individual learning factor is set to 2.5; The final value for the individual learning factor is 0.5. The initial value for the social learning factor is set to 0.5. The termination value for the social learning factor is 2.5.

[0105] It should be noted that in the optimization iteration of the improved particle swarm optimization algorithm, the individual learning factor decreases linearly from 2.5 to 0.5, while the social learning factor increases linearly from 0.5 to 2.5, in order to achieve the shift from emphasizing individual experience to emphasizing group experience.

[0106] Step 43: Using the improved particle swarm optimization algorithm, solve the optimization design function of the target transformer and output the optimization design result of the target transformer. Specifically, the process is as follows: Step 431: For each particle in the initial population, calculate the core loss, high-voltage winding loss and low-voltage winding loss corresponding to each particle.

[0107] Step 432: Based on the pre-built transformer hot spot temperature prediction model, predict the hot spot temperature at the hot spot monitoring location to obtain the predicted hot spot temperature value.

[0108] Step 433: Calculate the fitness value based on the optimization design function of the target transformer, and record the individual optimal position and the global optimal position.

[0109] Step 434: Based on the adaptive parameter velocity and position update method, update the velocity and position of each particle, and perform physical constraint checks and repairs on the updated particles to obtain the updated particles; repeat the operations of steps 431-433 for the updated particles.

[0110] The adaptive parameter speed and position update methods are as follows:

[0111]

[0112] in, for At this moment The velocity of each particle; for At this moment The velocity of each particle; A random number in the interval [0,1]. For the first The individual historical best position of each particle; for At this moment The position of each particle; A random number in the interval [0,1]. The globally optimal position; for At this moment The position of each particle.

[0113] It should be noted that the process of checking and repairing the physical constraints of the updated particles is basically the same as the operation in step 41 above, and will not be repeated here.

[0114] Step 435: Repeat the operation of step 434 until the preset termination condition is met, and output the optimized design result of the target transformer.

[0115] Example explanation: In this embodiment 1, an oil-immersed transformer for flexible loop closure is used as an example. The above-mentioned optimization design method for oil-immersed transformers for flexible loop closure is used for optimization design, and the trial-and-error method based on finite element simulation is used as a comparison.

[0116] The main parameters of the example oil-immersed transformer are shown in Table 1.

[0117] Table 1 shows the main parameters of an example oil-immersed transformer.

[0118] The optimized design results are explained below: (1) Calculate the mean absolute error, root mean square error and coefficient of determination of the pre-constructed transformer hot spot temperature prediction model; the calculation results show that the mean absolute error is 0.132K, the root mean square error is 0.185K and the coefficient of determination is 0.9987. It can be seen that the pre-constructed transformer hot spot temperature prediction model in the above embodiment 1 can meet the high precision requirements of engineering applications; in addition, the optimized running time of the pre-constructed transformer hot spot temperature prediction model in embodiment 1 and the finite element simulation software in the trial and error method based on finite element simulation is statistically analyzed, as shown in Table 2.

[0119] Table 2. Optimized running time of transformer hotspot temperature prediction model and finite element simulation software

[0120] As can be seen from Table 2 above, under the premise of meeting high accuracy, the transformer hot spot temperature prediction model pre-built in Example 1 has a significant advantage in running speed and can save a lot of computing resources.

[0121] (2) In the iterative optimization process of the improved particle swarm optimization algorithm, the optimal solution set of the variables to be optimized and the performance index of the example transformer is obtained as the optimization scheme; the optimization scheme is compared and analyzed with the original design scheme provided by the manufacturer, as shown in Table 3 below.

[0122] Table 3 Comparison of the optimized solution and the original design provided by the manufacturer

[0123] As can be seen from Table 3, under the condition of satisfying temperature constraints, the optimization scheme has a better optimization effect. Compared with the original design scheme provided by the manufacturer, the total loss of the optimization scheme is reduced by 911.69W, a reduction of 22.3%, which is suitable for multi-objective optimization design of high-performance transformers. Therefore, from the perspective of comprehensive development, the optimization scheme in this example is better than the original design scheme provided by the manufacturer and is more suitable for the development of power systems.

[0124] The oil-immersed transformer optimization design method for flexible loop closure described in Embodiment 1 determines the variables to be optimized and their value ranges for the target transformer based on the electrical parameter characteristics and thermal performance requirements of the target transformer. This defines a reasonable search boundary for the optimization process, reducing invalid searches from the source. By constructing a multi-objective optimization function with minimizing total loss as the main objective and satisfying hot spot temperature constraints, the optimization design function aims to minimize total power consumption while taking into account hot spot temperature constraints. This precisely meets the core requirement of balancing heat dissipation and efficiency for oil-immersed transformers in flexible loop closure under high-frequency harmonic and voltage fluctuation conditions. This completely solves the problem of traditional designs struggling to balance heat dissipation and efficiency from the perspective of optimization objectives.

[0125] In this embodiment 1, a physical information neural network model with an attention mechanism is trained as a transformer hotspot temperature prediction model and embedded into a particle swarm optimization algorithm to form an improved particle swarm optimization algorithm. Secondly, in the improved particle swarm optimization algorithm, a population initialization strategy guided by physical constraints is used to generate a particle swarm, and the feasibility of current density constraints and magnetic flux density constraints is checked and repaired for each particle in the particle swarm to ensure the high feasibility of initializing the population. In addition, the improved particle swarm optimization algorithm is designed with adaptive inertia weights and dynamic time-varying learning factors based on the current iteration progress and population diversity to improve optimization performance.

[0126] In this embodiment 1, the improved particle swarm optimization algorithm breaks through the limitations of existing algorithms in principle. Its physical constraint-guided population initialization strategy effectively eliminates infeasible solutions in the initial population through feasibility checks and repairs of current density and magnetic flux density, significantly improving search efficiency and avoiding geometric size conflicts and excessive temperature rise. Secondly, the embedded pre-built hot spot temperature prediction model significantly shortens the hot spot temperature calculation time and supports high-frequency iteration in the optimization process. In addition, the adaptive inertial weight and dynamic time-varying learning factor designed based on iteration progress and population diversity can dynamically adjust the particle search strategy, effectively avoiding the algorithm from converging to local optima too early, ensuring that the final output optimization design result is globally optimal. This satisfies the stringent heat dissipation requirements under the complex working conditions of flexible closed loop, minimizes the efficiency decline caused by harmonic losses, and improves the convergence speed of the optimization algorithm, thereby significantly improving the operational reliability, efficiency, and design rationality of oil-immersed transformers in flexible closed loops.

[0127] It should be noted that the oil-immersed transformer designed based on the optimized design method for flexible loop closing of the oil-immersed transformer described in Embodiment 1 is a flexible control type transformer. The flexible loop closing device constructed by it has significant advantages in terms of structural rationality and performance. In particular, in the oil-immersed transformer, transformer oil is used as the insulating medium filling material, which can effectively isolate oxygen to delay the oxidation process of the insulating paperboard and optimize the heat conduction efficiency. While improving equipment stability and simplifying the manufacturing process, it significantly extends the service life of the transformer in the flexible loop closing device.

[0128] Example 2 As attached Figure 4 As shown in the figure, this embodiment 2 provides an optimization design system for oil-immersed transformers with flexible loop closure, including a variable determination module, a function construction module, and an optimization solution module.

[0129] The variable determination module determines the variables to be optimized and their value ranges for the target transformer based on its electrical parameter characteristics and thermal performance requirements. The function construction module constructs the optimization design function for the target transformer based on the variables to be optimized and in conjunction with multi-objective optimization theory. The optimization design function aims to minimize total power consumption while satisfying hotspot temperature constraints. The optimization solution module solves the optimization design function for the target transformer using an improved particle swarm optimization algorithm, based on the variables to be optimized and their value ranges, and outputs the optimization design results for the target transformer.

[0130] It is worth noting that in the improved particle swarm optimization algorithm, a population initialization strategy guided by physical constraints is used to generate the particle swarm, and the feasibility of current density constraints and magnetic flux density constraints is checked and repaired for each particle in the particle swarm. Furthermore, the improved particle swarm optimization algorithm also incorporates a pre-built transformer hotspot temperature prediction model, and designs adaptive inertia weights and dynamic time-varying learning factors based on the current iteration progress and population diversity.

[0131] Example 3 As attached Figure 5 As shown, this embodiment 3 provides an electronic device, including: a memory for storing a computer program; a processor for executing the computer program to implement the steps of the optimization design method for a flexible loop-closed oil-immersed transformer; or, the processor for executing the computer program to implement the functions of each module in the above-mentioned optimization design system for a flexible loop-closed oil-immersed transformer.

[0132] For example, the computer program may be divided into one or more modules / units, which are stored in the memory and executed by the processor to complete the present invention. The one or more modules / units may be a series of computer program instruction segments capable of performing a preset function, the instruction segments describing the execution process of the computer program in the electronic device.

[0133] The electronic device may be a desktop computer, laptop, handheld computer, or cloud server, etc. The electronic device may include, but is not limited to, a processor and memory. Those skilled in the art will understand that the above are examples of electronic devices and do not constitute a limitation on the electronic device. It may include more components than described above, or combine certain components, or different components. For example, the electronic device may also include a communication interface, input / output devices, network access devices, and a bus.

[0134] The processor can be a central processing unit, or other general-purpose processors, digital signal processors, application-specific integrated circuits (ASICs), off-the-shelf programmable gate arrays (OPGs) or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. A general-purpose processor can be a microprocessor, or any conventional processor. The processor is the control center of the electronic device, connecting various parts of the electronic device via various communication interfaces and lines.

[0135] The memory can be used to store the computer program and / or module. The processor implements various functions of the electronic device by running or executing the computer program and / or module stored in the memory and by calling the data stored in the memory.

[0136] The memory may primarily include a program storage area and a data storage area. The program storage area may store the operating system and at least one application program required for a given function (such as sound playback, image playback, etc.). The data storage area may store data created based on the use of the mobile phone (such as audio data, phonebook, etc.). Furthermore, the memory may include high-speed random access memory and non-volatile memory, such as hard disks, RAM, plug-in hard disks, smart memory cards, secure digital cards, flash memory cards, at least one disk storage device, flash memory device, or other volatile solid-state storage devices.

[0137] Example 4 This embodiment 4 also provides a computer-readable storage medium storing a computer program that, when executed by a processor, implements the steps of the optimization design method for a flexible loop-closed oil-immersed transformer.

[0138] If the modules / units integrated in the optimized design system for oil-immersed transformers for flexible loop closure are implemented as software functional units and sold or used as independent products, they can be stored in a computer-readable storage medium.

[0139] Based on this understanding, the present invention can implement all or part of the processes in the above-described optimization design method for oil-immersed transformers with flexible loop closure, or it can be accomplished by a computer program instructing related hardware. The computer program can be stored in a computer-readable storage medium. When executed by a processor, the computer program can implement the steps of the above-described optimization design method for oil-immersed transformers with flexible loop closure. The computer program includes computer program code, which can be in the form of source code, object code, executable file, or a preset intermediate form, etc.

[0140] The computer-readable storage medium may include any entity or device capable of carrying the computer program code, recording media, USB flash drive, portable hard drive, magnetic disk, optical disk, computer memory, read-only memory, random access memory, electrical carrier signal, telecommunication signal, and software distribution medium, etc.

[0141] The optimization design method for oil-immersed transformers with flexible loop closure described in this invention generates a particle swarm using a population initialization strategy guided by physical constraints. Each particle in the swarm undergoes feasibility checks and repairs for current density and magnetic flux density constraints to directly eliminate and correct invalid design schemes that violate electromagnetic principles. This allows the improved particle swarm optimization algorithm to directly locate its search starting point within the physically feasible domain, avoiding redundant calculations and wasted computing power in invalid spaces. Consequently, the proportion of feasible solutions in the initialization population is increased from 40% in traditional methods to 100%, significantly reducing invalid calculations and improving optimization efficiency.

[0142] In this invention, an improved particle swarm optimization algorithm is formed by embedding a pre-built transformer hotspot temperature prediction model into the particle swarm optimization algorithm. The pre-built transformer hotspot temperature prediction model adopts a pre-trained physical information neural network model with an attention mechanism to predict the transformer hotspot temperature. Compared with finite element simulation in the trial-and-error method based on finite element simulation, the pre-built transformer hotspot temperature prediction model improves the single prediction speed by more than 1000 times, while maintaining a high accuracy level of less than 0.15K average absolute error, meeting the needs of engineering applications.

[0143] In this invention, by designing adaptive inertial weights and dynamic time-varying learning factors based on the current iteration progress and population diversity, the algorithm parameters can be dynamically adjusted according to the optimization progress and population state, thereby improving the convergence speed of the algorithm by 35%. High-quality solutions can usually be obtained within 70 iterations, which is significantly reduced compared to the 150 iterations required by traditional fixed parameter methods.

[0144] The above embodiments are merely one of the implementation methods for achieving the technical solution of the present invention. The scope of protection claimed by the present invention is not limited to this embodiment, but also includes any variations, substitutions and other implementation methods that can be easily conceived by those skilled in the art within the scope of the technology disclosed in the present invention.

Claims

1. An optimized design method for oil-immersed transformers with flexible loop closure, characterized in that, include: Based on the electrical parameter characteristics and thermal performance requirements of the target transformer, the variables to be optimized and their value ranges for the target transformer are determined. Based on the variables to be optimized for the target transformer, and combined with multi-objective optimization theory, an optimization design function for the target transformer is constructed. The optimization design function for the target transformer aims to minimize the total power consumption while satisfying the hot spot temperature constraint. Based on the variables to be optimized of the target transformer and their value range, the optimization design function of the target transformer is solved using an improved particle swarm optimization algorithm, and the optimization design result of the target transformer is output. In the improved particle swarm optimization algorithm, a population initialization strategy guided by physical constraints is used to generate the particle swarm, and the feasibility of current density constraints and magnetic flux density constraints is checked and repaired for each particle in the particle swarm. In addition, the improved particle swarm optimization algorithm also embeds a pre-built transformer hot spot temperature prediction model, and designs an adaptive inertia weight and a dynamic time-varying learning factor based on the current iteration progress and population diversity.

2. The optimized design method for an oil-immersed transformer with flexible loop closure according to claim 1, characterized in that, The variables to be optimized for the target transformer include at least one of the following: core diameter, number of turns in the low-voltage winding, thickness of the high-voltage conductor, width of the high-voltage conductor, thickness of the low-voltage conductor, and width of the low-voltage conductor.

3. The optimized design method for an oil-immersed transformer with flexible loop closure according to claim 1, characterized in that, The optimization design function for the target transformer is as follows: in, The optimization design function for the target transformer; The variables to be optimized are the target transformer; The objective function is the total loss. For hotspot temperature constraint functions; The highest permissible hot spot temperature; This refers to the no-load loss of the iron core. For winding copper loss; For a pre-built transformer hotspot temperature prediction model; The x-axis represents the location of the hotspot monitoring point; The vertical coordinate represents the location of the hotspot monitoring point; The intensity of the iron core heat source; The heat source intensity of the high-voltage winding; The heat source intensity for the low-voltage winding.

4. The optimized design method for an oil-immersed transformer with flexible loop closure according to claim 1, characterized in that, The pre-built transformer hotspot temperature prediction model is a pre-trained physical information neural network model that incorporates an attention mechanism; Among them, the pre-trained physical information neural network model with attention mechanism combines data-driven learning and physical constraints, and incorporates the residual of the steady-state heat conduction equation into the loss function through automatic differentiation technology.

5. The optimized design method for an oil-immersed transformer with flexible loop closure according to claim 1, characterized in that, The process of performing feasibility checks on current density constraints and magnetic flux density constraints for each particle in the particle swarm is as follows: in, This serves as a feasibility check and determination marker for particles; The variables to be optimized are the target transformer; This represents the high-voltage winding current density corresponding to the particle. The maximum allowable current density; This represents the low-voltage winding current density corresponding to the particle. The magnetic flux density of the iron core corresponding to the particle; This represents the maximum permissible magnetic flux density.

6. The optimized design method for an oil-immersed transformer with flexible loop closure according to claim 2, characterized in that, The repair process for each particle in the particle swarm is as follows: If the magnetic flux density of the iron core exceeds the preset magnetic flux density saturation value, then the diameter of the iron core is increased. If the cross-sectional area of ​​the high-voltage or low-voltage conductor is insufficient, the cross-sectional area of ​​the high-voltage or low-voltage conductor shall be enlarged proportionally.

7. The optimized design method for an oil-immersed transformer with flexible loop closure according to claim 1, characterized in that, The adaptive inertia weights, based on the current iteration progress and population diversity, are as follows: in, This is an adaptive inertial weight based on the current iteration progress and population diversity; For the weighting factor; Assigning weights to schedules; For diversity weight; The dynamic time-varying learning factor is as follows: in, For individual learning factors; As a social learning factor; This is the initial value of the individual learning factor; This is the termination value of the individual learning factor; This represents the current iteration number; This represents the maximum number of iterations. This is the initial value for the social learning factor; This is the termination value for the social learning factor.

8. An optimized design system for oil-immersed transformers with flexible loop closure, characterized in that, include: The variable determination module is used to determine the variables to be optimized for the target transformer and their value range based on the electrical parameter characteristics and thermal performance requirements of the target transformer. The function construction module is used to construct the optimization design function of the target transformer based on the variables to be optimized of the target transformer and combined with multi-objective optimization theory. The optimization design function of the target transformer aims to minimize the total power consumption while satisfying the hot spot temperature constraint. The optimization solution module is used to solve the optimization design function of the target transformer based on the variables to be optimized and their value range, using an improved particle swarm optimization algorithm, and outputs the optimization design result of the target transformer. In the improved particle swarm optimization algorithm, a population initialization strategy guided by physical constraints is used to generate the particle swarm, and the feasibility of current density constraints and magnetic flux density constraints is checked and repaired for each particle in the particle swarm. In addition, the improved particle swarm optimization algorithm also embeds a pre-built transformer hot spot temperature prediction model, and designs an adaptive inertia weight and a dynamic time-varying learning factor based on the current iteration progress and population diversity.

9. An electronic device, characterized in that, include: A processor is used to execute computer programs; A computer-readable storage medium storing a computer program, which, when executed by the processor, performs the oil-immersed transformer optimization design method for flexible loop closure as described in any one of claims 1-7.

10. A computer-readable storage medium storing a computer program, characterized in that, When the computer program is executed by the processor, it implements the oil-immersed transformer optimization design method for flexible loop closing as described in any one of claims 1-7.

Images