A transformer adaptive spraying cooling method, system, device and medium

Abstract

The present application relates to the technical field of transformer cooling, and discloses a transformer adaptive spraying cooling method, system, device and medium. The method comprises the following steps: based on the multi-dimensional working condition data of each main transformer, an oil temperature prediction value of the next moment of the corresponding main transformer is determined by using a machine learning model combining federated learning and gradient boosting machine; a first oil temperature deviation between the oil temperature prediction value and a preset target oil temperature and spraying energy consumption are used to construct a flow control target function corresponding to each main transformer, each flow control target function is solved according to the corresponding flow control constraint condition, and the optimal nozzle flow of the corresponding main transformer is obtained; and based on the optimal nozzle flow of each main transformer, the output power of the water pump of the corresponding main transformer is updated in real time, so that adaptive spraying cooling is performed on each main transformer. The present application realizes the precision, energy saving and safety of transformer spraying cooling.

Description

Technical Field

[0001] This invention relates to the field of transformer cooling technology, and in particular to a transformer adaptive spray cooling method, system, equipment and medium. Background Technology

[0002] Effective heat dissipation is crucial for the normal and safe operation of oil-immersed self-cooled transformers. Due to their inherent structural limitations, the heat dissipation effect of self-cooled transformers depends on the system load and ambient temperature. Especially during hot seasons, high ambient temperatures and a surge in electrical load dramatically increase the heat dissipation pressure on outdoor transformers. If heat dissipation is not timely, it can lead to accelerated aging of the insulation medium, reduced transformer lifespan, and in severe cases, even load shedding, endangering the safe operation of the power grid. Therefore, researching auxiliary cooling devices is of great significance for the safe operation of transformers.

[0003] Traditional transformer auxiliary cooling often employs spray cooling, which is manually controlled to start and stop according to on-site needs. However, existing substations are all unattended, and if transformer cooling is required, there is a delay in personnel arriving on-site: delayed start-up leads to untimely heat dissipation; delayed shutdown results in wasted resources. To address the lag in manually starting and stopping transformer auxiliary cooling devices, existing technologies have proposed setting fixed start-up and stop times for main transformer spray auxiliary cooling. This spraying method lacks timeliness; long start-up times lead to water waste and poor cooling effect, especially during rainy weather. Another approach involves laying PVC pipes around the transformer's outer wall for spray cooling, but in summer, high ambient temperatures cause the main transformer to heat up rapidly, making timely cooling impossible, and the fixed water pressure results in poor cooling performance. Summary of the Invention

[0004] The purpose of this invention is to provide a transformer adaptive spray cooling method, system, equipment, and medium to solve the problems in the prior art where the transformer auxiliary spray cooling strategy cannot be adaptively adjusted according to the transformer operating environment, resulting in resource waste, start-up and shutdown issues, and poor cooling effect.

[0005] In a first aspect, embodiments of the present invention provide a transformer adaptive spray cooling method, comprising:

[0006] Acquire multi-dimensional operating condition data of the main transformer in each substation;

[0007] Based on the multi-dimensional operating condition data of each main transformer, a machine learning model combining federated learning and gradient boosting machine is used to determine the predicted oil temperature value of the corresponding main transformer at the next moment through oil temperature prediction.

[0008] The first oil temperature deviation between the predicted oil temperature and the preset target oil temperature and the spray energy consumption are used to construct the flow control objective function corresponding to each main transformer. Each flow control objective function is solved according to the corresponding flow control constraints to obtain the optimal nozzle flow rate for the corresponding main transformer.

[0009] Based on the optimal nozzle flow rate of each main transformer, the water pump output power of the corresponding main transformer is updated in real time to perform adaptive spray cooling on each main transformer.

[0010] Preferably, after constructing the flow control objective function corresponding to each main transformer based on the first oil temperature deviation between the predicted oil temperature and the preset target oil temperature and the spray energy consumption, and solving each flow control objective function according to the corresponding flow control constraints to obtain the optimal nozzle flow rate for the corresponding main transformer, the method further includes:

[0011] Calculate the second oil temperature deviation between the predicted oil temperature and the actual oil temperature at the next moment for each of the main transformers, and adjust the machine learning model parameters based on the second oil temperature deviation for all the main transformers.

[0012] Preferably, the multi-dimensional operating condition data includes main transformer operation monitoring data, real-time meteorological data, and sprinkler device execution data. The main transformer operation monitoring data includes the main transformer load value and main transformer oil temperature. The real-time meteorological data includes the main transformer area temperature and main transformer area wind speed. The sprinkler device execution data includes the nozzle flow rate and sprinkler water temperature.

[0013] Preferably, the step of determining the predicted oil temperature value for the next moment of the corresponding main transformer by using a machine learning model combining federated learning and gradient boosting machine based on the multi-dimensional operating condition data of each main transformer includes:

[0014] Based on the federated learning architecture, the client is deployed at each substation, and the server is deployed at the regional power dispatch center.

[0015] The pre-trained lightweight gradient booster model is distributed to each client through the server, so that each client can fine-tune the lightweight gradient booster model using local historical main transformer operation data to obtain a personalized gradient booster model for the client.

[0016] The multi-dimensional operating condition data of each main transformer are input into the personalized gradient booster model of the corresponding client to predict the oil temperature, and the predicted oil temperature value of the main transformer at the next moment is output.

[0017] Preferably, the step of constructing a flow control objective function for each main transformer based on the first oil temperature deviation between the predicted oil temperature and the preset target oil temperature and the spray energy consumption, and solving each flow control objective function according to the corresponding flow control constraints to obtain the optimal nozzle flow rate for the corresponding main transformer, includes:

[0018] Based on minimizing the dynamic weighted sum of the absolute value of the first oil temperature deviation and the nozzle flow rate, a flow control objective function is constructed for each of the main transformers.

[0019] The particle swarm optimization algorithm is used to solve each of the flow control objective functions according to the corresponding flow control constraints, including upper and lower flow limits, oil temperature prediction uncertainty constraints, and flow rate change constraints, to obtain the optimal nozzle flow rate for the corresponding main transformer.

[0020] Preferably, the sum of the absolute value of the first oil temperature deviation and the dynamic weighting factor of the nozzle flow rate is 1, and the adjustment strategy for the dynamic weighting factor of the absolute value of the first oil temperature deviation includes:

[0021] When the first oil temperature deviation is within the safe deviation range, the dynamic weighting factor of the absolute value of the first oil temperature deviation is the first value;

[0022] When the first oil temperature deviation is within the controllable deviation range, the dynamic weighting factor of the absolute value of the first oil temperature deviation is the sum of the weighted value of the ratio of the first oil temperature deviation to the length of the controllable deviation range and the first value.

[0023] When the first oil temperature deviation is in the excess deviation range, the dynamic weighting factor of the absolute value of the first oil temperature deviation is the second value.

[0024] Preferably, the step of updating the water pump output power of the corresponding main transformer in real time based on the optimal nozzle flow rate of each main transformer, so as to perform adaptive spray cooling on each main transformer, includes:

[0025] Based on the optimal nozzle flow rate of each main transformer, the water pump power adjustment amount corresponding to the main transformer is derived through a pre-constructed flow-power mapping relationship.

[0026] Based on the water pump power adjustment amount of each main transformer, the water pump output power of the corresponding main transformer is adjusted in real time to perform adaptive spray cooling on each main transformer.

[0027] Secondly, embodiments of the present invention provide a transformer adaptive spray cooling system, comprising:

[0028] The data acquisition module is used to acquire multi-dimensional operating condition data of the main transformer in each substation;

[0029] The oil temperature prediction module is used to determine the predicted oil temperature value of the corresponding main transformer at the next moment by using a machine learning model that combines federated learning and gradient boosting machine, based on the multi-dimensional operating condition data of each main transformer.

[0030] The flow rate calculation module is used to construct the flow control objective function corresponding to each main transformer based on the first oil temperature deviation of the predicted oil temperature and the preset target oil temperature and the spray energy consumption, and to solve each flow control objective function according to the corresponding flow control constraints to obtain the optimal nozzle flow rate of the corresponding main transformer.

[0031] The spray execution module is used to update the water pump output power of the corresponding main transformer in real time based on the optimal nozzle flow rate of each main transformer, so as to perform adaptive spray cooling on each main transformer.

[0032] Thirdly, embodiments of the present invention provide a terminal device, including a processor, a memory, and a computer program stored in the memory and configured to be executed by the processor, wherein the processor executes the computer program to implement the transformer adaptive spray cooling method as described above.

[0033] Fourthly, embodiments of the present invention provide a computer-readable storage medium, the computer-readable storage medium including a stored computer program, wherein, when the computer program is executed, it controls the device where the computer-readable storage medium is located to perform the transformer adaptive spray cooling method as described above.

[0034] Compared with the prior art, the adaptive spray cooling method, system, equipment, and medium for transformers according to embodiments of the present invention have the following advantages at least one point:

[0035] (1) Based on multi-dimensional working condition data, combined with the hybrid model of federated learning and gradient booster, the oil temperature prediction value of each main transformer at the next moment is output in advance, breaking through the traditional passive response mode. When the temperature does not exceed the standard, the active intervention is carried out through flow optimization, which effectively solves the problem of unattended substation cooling start-up and shutdown and heat dissipation not in time, ensuring the stability of transformer insulation medium and extending equipment service life.

[0036] (2) By leveraging the federated learning architecture of “global parameter aggregation + local model fine-tuning”, global knowledge sharing among multiple substations can be achieved, and local adaptation training can be used to adapt to the exclusive heat dissipation characteristics of main transformers of different models and operating environments, avoiding the rough control of the traditional centralized model. At the same time, the original data “does not leave the station”, and only the model parameters are uploaded, strictly protecting the data privacy and security of the power industry.

[0037] (3) The objective function is constructed with minimizing the first oil temperature deviation + spray energy consumption as the core. The optimal nozzle flow rate is solved by combining the flow control constraints. The cooling effect is prioritized when the temperature is high and the load is high, and energy saving and consumption reduction are prioritized when the load is low or the weather is favorable, which greatly reduces the waste of water resources and energy caused by traditional fixed flow spraying. Attached Figure Description

[0038] Figure 1 This is a schematic flowchart of an embodiment of the adaptive spray cooling method for transformers according to the present invention;

[0039] Figure 2 This is a schematic diagram of the oil temperature prediction process according to an embodiment of the present invention;

[0040] Figure 3 This is a schematic diagram of the process for solving the optimal nozzle flow rate according to an embodiment of the present invention;

[0041] Figure 4 This is a schematic diagram of the structure of a transformer adaptive spray cooling system according to an embodiment of the present invention;

[0042] Figure 5 This is a schematic diagram of the structure of a terminal device according to an embodiment of the present invention;

[0043] Figure label:

[0044] 01. Data acquisition module; 02. Oil temperature prediction module; 03. Flow rate calculation module; 04. Spraying execution module; 5000. Terminal device; 5001. Processor; 5002. Bus; 5003. Memory; 5004. Transceiver. Detailed Implementation

[0045] The specific embodiments of the present invention will be described in further detail below with reference to the accompanying drawings and examples. The following examples are for illustrative purposes only and are not intended to limit the scope of the invention.

[0046] In the description of this invention, it should be understood that the terms "first" and "second," etc., are used to distinguish different objects, rather than to describe a specific order.

[0047] In the description of this invention, it should be noted that, unless otherwise defined, all technical and scientific terms used in this invention have the same meaning as commonly understood by those skilled in the art. The terminology used in this specification is for the purpose of describing specific embodiments only and is not intended to limit the invention. Those skilled in the art can understand the specific meaning of the above terms in this invention based on the specific circumstances.

[0048] like Figure 1 The diagram shown is a flowchart illustrating an adaptive spray cooling method for transformers according to an embodiment of the present invention. (Refer to...) Figure 1 An embodiment of the present invention provides a transformer adaptive spray cooling method, comprising the following steps:

[0049] S1. Obtain multi-dimensional operating condition data of the main transformer in each substation;

[0050] The main transformer in each substation refers to the oil-immersed self-cooled main transformer in each unmanned outdoor substation within the same power dispatching area. Multi-dimensional operating data includes main transformer operation monitoring data, real-time meteorological data, and sprinkler system execution data. Specifically, the main transformer operation monitoring data includes the main transformer load value and main transformer oil temperature; the real-time meteorological data includes the temperature and wind speed in the main transformer area; and the sprinkler system execution data includes nozzle flow rate and sprinkler water temperature.

[0051] Specifically, by deploying supporting sensing equipment and data acquisition modules on and around the main transformer in the substation, multi-dimensional operating condition data is obtained: In the main transformer operation monitoring data, the main transformer load value and main transformer oil temperature are calculated and obtained in real time by the background monitoring system; In the real-time meteorological data, the temperature in the main transformer area is collected by a digital temperature sensor that avoids direct sunlight and equipment heat dissipation interference, and the wind speed in the main transformer area is collected by an ultrasonic anemometer or cup anemometer placed in an open area; In the sprinkler device execution data, the nozzle flow rate is monitored in real time by an electromagnetic flow meter or turbine flow meter installed in the main water supply pipeline and branch pipeline of the sprinkler device, and the sprinkler water temperature is collected by a temperature sensor at the outlet of the sprinkler water tank or the main water supply pipeline.

[0052] Furthermore, after the data collection is completed, outlier removal, data normalization, and time alignment preprocessing are performed on the three types of raw data in sequence to ensure the accuracy, consistency, and timeliness of the data, providing reliable data support for subsequent oil temperature prediction and flow optimization control.

[0053] S2. Based on the multi-dimensional operating condition data of each main transformer, a machine learning model combining federated learning and gradient boosting machine is used to determine the predicted oil temperature value of the corresponding main transformer at the next moment through oil temperature prediction.

[0054] It should be noted that the main transformer cooling process is a complex fitting process between main transformer operation monitoring data, real-time meteorological data, and spray device execution data. Therefore, constructing a dynamic model that can accurately predict the main transformer oil temperature is the prerequisite and foundation for realizing closed-loop intelligent control.

[0055] Specifically, this step employs a prediction scheme combining personalized federated learning and a lightweight gradient booster as a dynamic model to predict the oil temperature at the next time step. This model correlates the predicted temperature at the next time step with the current main transformer operation monitoring data, real-time meteorological data, and sprinkler system execution data.

[0056] like Figure 2As shown, this is a flowchart illustrating step S2. (Refer to...) Figure 2 Step S2 includes:

[0057] S201. Based on the federated learning architecture, the client is deployed at each substation, and the server is deployed at the regional power dispatch center.

[0058] The client is deployed as an edge node of the substation, responsible for the preprocessing of local multi-dimensional operating data, training and real-time prediction of personalized gradient booster models, and focusing on learning the specific heat dissipation characteristics of the corresponding main transformer. The server is deployed as a federated server in the regional power dispatch center, responsible for the aggregation and updating of global model parameters, the distribution of personalized models, and the coordination and management of training strategies.

[0059] Through the distributed architecture design of "client-server", it can realize the sharing and interoperability of global knowledge among multiple substations, and ensure that the original data of each substation "does not leave the station", strictly complying with the power industry's data privacy and security standards.

[0060] S202. The pre-trained lightweight gradient booster model is distributed to each client through the server, so that each client can fine-tune the lightweight gradient booster model using local historical main transformer running data to obtain a personalized gradient booster model for the corresponding client.

[0061] The federated server pre-trains a lightweight gradient boosting machine initial model based on historical public datasets. By setting an upper limit on tree depth, controlling the number of trees in a single model, and implementing an early stopping mechanism, it compresses the model size and controls overfitting, ensuring that the model meets the real-time inference requirements of edge computing. The global parameters of the initial model are denoted as follows. .

[0062] After receiving the initial model, each client performs an initial round of local fine-tuning using nearly N (N≥1) months of local historical data on the multi-dimensional operating conditions of the main transformer. The model parameters are optimized by minimizing the mean absolute error loss function, generating an initial personalized model. This model adopts a two-layer structure: a shared feature layer and a local adaptation layer. The shared feature layer is trained by a federated server aggregating data from multiple stations to learn the general features of the load-oil temperature mapping. The local adaptation layer is trained only on the client side, learning the specific heat dissipation characteristics of the corresponding main transformer. This ultimately forms a personalized gradient booster model, whose parameters satisfy the following formula:

[0063]

[0064] in, This represents the client's local model parameters in round t+1. This represents the global parameter weights, used to balance generality and individualization. This represents the global model parameters issued by the federated server in round t+1. This represents the client's local model parameters in round t.

[0065] S203. Input the multi-dimensional operating condition data of each main transformer into the personalized gradient booster model of the corresponding client to predict the oil temperature, and output the predicted oil temperature value of the corresponding main transformer at the next moment.

[0066] The preprocessed multi-dimensional operating condition data of the main transformer is used to construct a real-time feature vector. , Includes the main transformer load values ​​at time t and at n historical times. Main transformer oil temperature and the temperature of the main variable region at time t Regional wind speed Nozzle flow rate Spray water temperature .

[0067] Will The personalized gradient booster model input from the client is used to output the predicted oil temperature for the next time step through ensemble computation of multiple decision trees. The integrated prediction formula is:

[0068]

[0069] in, denoted by , where K represents the total number of decision trees and b represents the bias term.

[0070] Meanwhile, the loss of the local model on the data of that day is calculated as follows:

[0071]

[0072] in, This represents the mean absolute error loss of the local model on the client side, where m represents the local sample size. This represents the predicted oil temperature value for the i-th local sample. This represents the actual oil temperature value of the i-th local sample.

[0073] like Greater than the preset threshold (Based on the accuracy requirements for main transformer oil temperature prediction in the power industry and actual operating conditions), the tree structure weights, splitting thresholds, and other parameters of the model are encrypted and uploaded to the federated server; the model synchronously outputs the 90% confidence interval as follows:

[0074]

[0075] in, This represents the standard deviation of the prediction, providing a reliable basis for adjusting the constraints of subsequent flow optimization.

[0076] Among them, the prediction standard deviation The variance of the leaf node outputs of the lightweight gradient booster is estimated using the following formula:

[0077]

[0078] in, This represents the average value of all decision tree outputs.

[0079] It should be noted that the machine learning model in this step combines the distributed training characteristics of federated learning, the efficient predictive capabilities of lightweight gradient boosting machines, and a personalized modeling approach. Each client's data is processed and trained only on its local node; there is no need to upload raw data to the federated server. The client only uploads model parameters, and this transmission is done via an encrypted protocol to further ensure data privacy and comply with the power industry's data security standards. The federated server's "shared feature layer" learns the common patterns of multiple substations, while the client's "local adaptation layer" focuses on learning the individual characteristics of specific transformers, solving the problem of insufficient personalization caused by the "one-size-fits-all" approach of traditional centralized models. Clients can fine-tune the model in real time based on the latest local data, quickly adapting to local characteristic changes such as equipment aging and seasonal variations.

[0080] S3. Construct the flow control objective function for each main transformer based on the predicted oil temperature and the first oil temperature deviation of the preset target oil temperature and the spray energy consumption. Solve each flow control objective function according to the corresponding flow control constraints to obtain the optimal nozzle flow rate for the corresponding main transformer.

[0081] like Figure 3 As shown, this is a flowchart illustrating step S3. (Refer to...) Figure 3 Step S3 includes:

[0082] S301. Based on minimizing the dynamic weighted sum of the absolute value of the first oil temperature deviation and the nozzle flow rate, construct the flow control objective function corresponding to each main transformer;

[0083] Specifically, the flow control objective function for each main transformer is characterized by the following formula:

[0084]

[0085] in, This represents the objective function for flow control. ( () indicates the predicted oil temperature value With preset target oil temperature First oil temperature deviation, preset target oil temperature It can be set according to the temperature control requirements of each main transformer, and The smaller the value, the closer the cooling effect is to the target. This indicates that nozzle flow rate (a decision variable) is positively correlated with spray energy consumption; The dynamic weighting factor represents the absolute value of the first oil temperature deviation. The dynamic weighting factor representing the nozzle flow rate is 1, and can be dynamically adjusted according to the first oil temperature deviation.

[0086] The following explanation uses the dynamic weighting factor of the absolute value of the first oil temperature deviation as an example to illustrate its adjustment strategy:

[0087] 1) When the first oil temperature deviation is within the safe deviation range, the dynamic weighting factor of the absolute value of the first oil temperature deviation is the first value;

[0088] Safety deviation range is In this embodiment, the first value is set to 0.1. That is, when the first oil temperature deviation... hour, It is 0.1.

[0089] 2) When the first oil temperature deviation is within the controllable deviation range, the dynamic weighting factor of the absolute value of the first oil temperature deviation is the sum of the weighted value of the ratio of the first oil temperature deviation to the length of the controllable deviation range and the first value.

[0090] The controllable deviation range is ( This represents the maximum permissible controllable deviation value. , (This represents the upper limit of the controllable deviation oil temperature). In this embodiment, the weight of the ratio of the first oil temperature deviation to the length of the controllable deviation interval is assigned as 0.85. That is, when the first oil temperature deviation is within... hour, .

[0091] 3) When the first oil temperature deviation is in the over-limit deviation range, the dynamic weighting factor of the absolute value of the first oil temperature deviation is the second value.

[0092] The deviation range is In this embodiment, the second value is set to 0.95. That is, when the first oil temperature deviation... hour, It is 0.95.

[0093] It should be noted that when When the oil temperature reaches the standard, energy saving should be prioritized. Take the minimum value of 0.1. Take the maximum value of 0.9; when At that time, the weight changes linearly with the deviation. The larger, The closer it is to 0.95, the better the temperature control effect; when When the oil temperature approaches the safe threshold, the temperature control weight is forcibly maximized. .

[0094] S302. The particle swarm optimization algorithm is used to solve each flow control objective function according to the corresponding flow control constraints, including upper and lower flow limits, oil temperature prediction uncertainty constraints, and flow rate change constraints, to obtain the optimal nozzle flow rate of the corresponding main transformer.

[0095] The following is a detailed explanation of the flow control constraints, which include upper and lower flow limits, oil temperature prediction uncertainty constraints, and flow rate change constraints:

[0096] 1) Flow upper and lower limit constraints:

[0097] This constraint limits the range of nozzle flow rate to prevent the water film from failing to form effectively due to excessively low flow rate, while also preventing the flow rate from exceeding the pressure resistance and load-bearing capacity of the water pump and pipeline.

[0098] Specifically, the upper and lower flow limits are represented by the following formula:

[0099]

[0100] in, This indicates the minimum flow rate required to ensure an effective water film forms on the surface of the main transformer. This indicates the maximum flow rate within the pressure resistance capacity range of the water pump and pipeline.

[0101] 2) Uncertainty constraints in oil temperature prediction:

[0102] This constraint addresses the uncertainty in machine learning model predictions, ensuring that the transformer oil temperature will not exceed the safe threshold even if the prediction is biased. If the upper limit of the 90% confidence interval for the predicted oil temperature exceeds the transformer's safe threshold, the flow rate limit must be relaxed to prioritize safety.

[0103] Specifically, the following formula is used to characterize the uncertainty constraint of oil temperature prediction:

[0104]

[0105] in, This represents the standard deviation of the forecast; 1.645 is the coefficient for the 90% confidence interval. This indicates the safe oil temperature threshold of the main transformer.

[0106] 3) Flow rate change constraint:

[0107] This constraint limits the magnitude of a single flow rate adjustment, preventing pump overload and motor damage due to sudden power changes, while also preventing sudden flow rate fluctuations from impacting the main transformer's heat dissipation.

[0108] Specifically, the flow rate change constraint is characterized by the following formula:

[0109]

[0110] in, Indicates the current nozzle flow rate. This indicates the maximum allowable adjustment of the flow rate in a single operation, to prevent sudden changes in pump power or impact on the main transformer's heat dissipation.

[0111] Based on this, the particle swarm optimization algorithm is launched to solve the problem. The solution process is explained in detail below:

[0112] N particles are randomly generated, and each particle corresponds to a candidate nozzle flow rate. Substitute each candidate flow rate into the flow control objective function to calculate the fitness; the smaller the objective function value, the higher the fitness.

[0113] Update particle velocity using the following formula:

[0114]

[0115] in, This represents the velocity of the i-th particle in the (k+1)-th iteration. Indicates inertia weight, This represents the velocity of the i-th particle in the k-th iteration. , Represents the learning factor. , Represents a random number in the interval [0,1]. This represents the historical best position of the i-th particle. Indicates the globally optimal position. This represents the position of the i-th particle in the k-th iteration.

[0116] Then update the particle position using the following formula:

[0117]

[0118] After repeating the iterations until the maximum number of iterations is reached, the output global optimal solution is the optimal nozzle flow rate for the corresponding main transformer. .

[0119] It should be noted that after step S3, the method further includes: calculating the second oil temperature deviation between the predicted oil temperature value and the actual oil temperature value of each main transformer at the next moment, and adjusting the machine learning model parameters based on the second oil temperature deviation of all main transformers.

[0120] Specifically, the parameter tuning process for machine learning models includes:

[0121] 1) Calculate the absolute difference between the predicted oil temperature and the actual oil temperature of each main transformer at the next moment as the second oil temperature deviation of the corresponding main transformer at the next moment;

[0122] The actual oil temperature of each main transformer is collected in real time by temperature sensors. Combined with the predicted oil temperature of the corresponding main transformer at the next moment The second oil temperature deviation is calculated using the absolute error formula, which is:

[0123]

[0124] in, This indicates the second oil temperature deviation, reflecting the degree of deviation between the predicted and actual oil temperature values ​​at a single moment.

[0125] 2) Based on the second oil temperature deviation of each main transformer at the next moment, the performance of the machine learning model is evaluated from the two dimensions of instantaneous accuracy and stability, and the evaluation result of whether the machine learning model needs to start parameter adjustment is obtained.

[0126] Set two error thresholds (High-precision threshold) and (Acceptable threshold) ).

[0127] Evaluate single-shot prediction performance from the perspective of instantaneous accuracy (instantaneous performance assessment adapted to a single second oil temperature deviation): If If the prediction accuracy is good, the model does not need adjustment; if If the prediction accuracy is deemed acceptable, the recording error will not be adjusted for the time being; if If so, an early warning will be triggered and the error trend will be tracked.

[0128] Evaluate long-term forecasting performance from a stability perspective (assessment of the cumulative trend of deviations over multiple consecutive periods): calculate the average error over multiple consecutive control periods. (n represents the number of consecutive periods, (representing the second oil temperature deviation in the i-th cycle) and the maximum error ;like or If the model is found to have systematic bias, the output will indicate that parameter adjustments are needed.

[0129] 3) Based on the evaluation results of the machine learning model requiring parameter adjustment, perform hierarchical parameter adjustment on the machine learning model according to the federated learning architecture.

[0130] A tiered adjustment strategy of "local fine-tuning - global update" is adopted, which is explained in detail below:

[0131] First, at the client level, if the stability adjustment condition is triggered, the personalized gradient booster model is incrementally trained based on the latest sets of local multidimensional operating condition data, using the following formula:

[0132]

[0133] in, This represents the new local model parameters on the client side after incremental training. This represents the current local model parameters (including tree weights and split thresholds). This represents the gradient of the local mean absolute error loss function with respect to the parameters. It represents the learning rate, enabling rapid adaptation to recent changes in operating conditions.

[0134] Secondly, at the federated server level, when multiple clients experience errors exceeding the limit, each client calculates the recent loss of its local model and encrypts and uploads the model parameters. The server aggregates global parameters by sample size weighting:

[0135]

[0136] in, This represents the global update parameters generated by the federated server after aggregating parameters from all clients, where C represents the total number of clients participating in the update. This represents the sample size of the c-th client. This represents the local model parameters uploaded by the c-th client.

[0137] After downloading global parameters, the client merges them with local parameters to generate a new personalized model.

[0138]

[0139] in, This represents the final personalized model parameters obtained by the client after integrating the globally updated parameters with the new local parameters.

[0140] Finally, adjust the input feature weights to address persistent bias:

[0141]

[0142] in, Indicates the adjusted features The weight, Representation of features The initial weights, Indicates the deviation from the feature The partial derivatives, This represents an adjustment coefficient that enhances the model's sensitivity to key factors.

[0143] It should be noted that the above parameter tuning process evaluates the model performance from the perspectives of real-time accuracy and stability by calculating the second oil temperature deviation, triggering local incremental training or global parameter updates, and realizing dynamic adaptation of model parameters to ensure that the prediction and control effects continuously match actual working conditions such as equipment aging and seasonal changes.

[0144] S4. Based on the optimal nozzle flow rate of each main transformer, update the water pump output power of the corresponding main transformer in real time to perform adaptive spray cooling for each main transformer.

[0145] Specifically, step S4 includes:

[0146] 1) Based on the optimal nozzle flow rate of each main transformer, the water pump power adjustment amount of the corresponding main transformer is derived through the pre-constructed flow-power mapping relationship;

[0147] A flow-power mapping relationship is constructed, which is based on the calibration of the pump characteristic curve and the pipeline resistance characteristics, and the formula is as follows:

[0148]

[0149] in, This represents the pump output power required to achieve the optimal nozzle flow rate. The coefficient representing the correspondence between power and flow rate (pre-calibrated based on parameters such as pump model, pipe diameter, and length). This indicates the optimal nozzle flow rate.

[0150] Combined with the current nozzle flow rate Calculate flow adjustment amount Then, the pump power adjustment amount is derived through the mapping relationship, and the formula is:

[0151]

[0152] in, This indicates the current output power of the water pump. This refers to the amount of pump power adjustment required to achieve the optimal flow rate (a positive value indicates that the power needs to be increased, and a negative value indicates that the power needs to be decreased).

[0153] 2) Adjust the output power of the water pump of the corresponding main transformer in real time according to the water pump power adjustment amount of each main transformer, so as to perform adaptive spray cooling for each main transformer.

[0154] Based on the water pump power adjustment, the power adjustment is converted into a motor frequency change via a frequency converter. The frequency adjustment formula is as follows:

[0155]

[0156] in, Indicates the operating frequency of the target motor. Indicates the rated frequency of the water pump motor. This indicates the rated power of the water pump.

[0157] The motor speed is changed by adjusting the frequency, thereby updating the water pump output power in real time. After power adjustment, the actual flow rate of the nozzle is collected in real time via an electromagnetic flow meter. The flow deviation is calculated as follows:

[0158]

[0159] in, This indicates flow deviation.

[0160] At the same time, the actual oil temperature of the main transformer in the next data acquisition cycle is recorded. The cooling effect can be evaluated using the following formula:

[0161]

[0162] in, This represents the difference between the current transformer oil temperature and the actual transformer oil temperature in the next data collection cycle. This indicates that the cooling effect is effective; the higher the value, the better the cooling effect.

[0163] like If the threshold is exceeded (usually set to 5%), a secondary fine-tuning is triggered, using the following formula:

[0164]

[0165] in, This indicates the amount of secondary fine-tuning of the flow rate. This represents the fine-tuning coefficient, ensuring that the spray intensity is consistent with the optimization target.

[0166] based on Recalculate power adjustment amount By fine-tuning the motor frequency through a frequency converter until the deviation between the actual flow rate and the optimal flow rate meets the requirements, adaptive and precise spray cooling is achieved for each main transformer.

[0167] It should be noted that this step adjusts the water pump power in real time based on the optimal nozzle flow rate, and with the help of a secondary fine-tuning mechanism, ensures that the spray intensity is precisely matched with the cooling requirements, significantly improves the stability of transformer operation, and reduces the risk of grid load limiting.

[0168] This invention discloses an adaptive spray cooling method for transformers. Relying on multi-dimensional operating condition data and combining a hybrid model of federated learning and gradient boosting machines, it pre-outputs the predicted oil temperature of each main transformer at the next moment, breaking through the traditional passive response mode. It proactively intervenes through flow optimization before the temperature exceeds the limit, effectively solving the problems of delayed heat dissipation and start-up / shutdown in unattended substations, ensuring the stability of the transformer insulation medium and extending equipment lifespan. Utilizing the federated learning architecture of "global parameter aggregation + local model fine-tuning," it achieves global knowledge sharing among multiple substations and adapts to the specific heat dissipation characteristics of different models and operating environments of main transformers through local adaptation training, avoiding the coarse "one-size-fits-all" control of traditional centralized models. Simultaneously, the original data "does not leave the station," only the model parameters are uploaded, strictly protecting data privacy and security in the power industry. The objective function is constructed with minimizing "first oil temperature deviation + spray energy consumption" as the core, and the optimal nozzle flow rate is solved by combining flow control constraints. Under high temperature and high load conditions, priority is given to ensuring cooling effect, while under low load or favorable weather conditions, priority is given to energy saving and consumption reduction, significantly reducing water and energy waste caused by traditional fixed-flow spraying.

[0169] like Figure 4 The diagram shown is a structural schematic of a transformer adaptive spray cooling system according to an embodiment of the present invention. (Refer to...) Figure 4 An embodiment of the present invention provides a transformer adaptive spray cooling system, comprising:

[0170] Data acquisition module 01 is used to acquire multi-dimensional operating condition data of the main transformer in each substation;

[0171] The oil temperature prediction module 02 is used to determine the predicted oil temperature value of the corresponding main transformer at the next moment by using a machine learning model that combines federated learning and gradient boosting machine based on the multi-dimensional operating condition data of each main transformer.

[0172] The flow calculation module 03 is used to construct the flow control objective function corresponding to each main transformer based on the first oil temperature deviation of the predicted oil temperature and the preset target oil temperature and the spray energy consumption. The flow control objective function is solved according to the corresponding flow control constraints to obtain the optimal nozzle flow rate of the corresponding main transformer.

[0173] The spray execution module 04 is used to update the water pump output power of the corresponding main transformer in real time based on the optimal nozzle flow rate of each main transformer, so as to perform adaptive spray cooling for each main transformer.

[0174] It should be noted that the various modules in the aforementioned transformer adaptive spray cooling system can be implemented entirely or partially through software, hardware, or a combination thereof. These modules can be embedded in or independent of the processor in a computer device, or stored in the memory of a computer device as software, so that the processor can call and execute the corresponding operations of each module. For specific limitations regarding the transformer adaptive spray cooling system, please refer to the limitations regarding the transformer adaptive spray cooling method above; both have the same function and effect, and will not be repeated here.

[0175] This invention also provides a terminal device, which includes:

[0176] Processor, memory, and bus;

[0177] The bus is used to connect the processor and the memory;

[0178] The memory is used to store operation instructions;

[0179] The processor is configured to execute operations corresponding to the above-described transformer adaptive spray cooling method by calling the operation instructions.

[0180] In one alternative embodiment, a terminal device is provided, such as Figure 5 As shown, Figure 5 The terminal device 5000 shown includes a processor 5001 and a memory 5003. The processor 5001 and the memory 5003 are connected, for example, via a bus 5002. Optionally, the terminal device 5000 may also include a transceiver 5004. It should be noted that in practical applications, the transceiver 5004 is not limited to one type, and the structure of this terminal device 5000 does not constitute a limitation on the embodiments of the present invention.

[0181] Processor 5001 may be a CPU, a general-purpose processor, a DSP, an ASIC, an FPGA, or other programmable logic device, transistor logic device, hardware component, or any combination thereof. It may implement or execute the various exemplary logic blocks, modules, and circuits described in connection with this disclosure. Processor 5001 may also be a combination that implements computational functions, such as including one or more microprocessor combinations, a combination of a DSP and a microprocessor, etc.

[0182] Bus 5002 may include a path for transmitting information between the aforementioned components. Bus 5002 may be a PCI bus or an EISA bus, etc. Bus 5002 can be divided into address bus, data bus, control bus, etc. For ease of representation, Figure 5 The bus is represented by a single thick line, but this does not mean that there is only one bus or one type of bus.

[0183] The memory 5003 may be a ROM or other type of static storage device capable of storing static information and instructions, RAM or other type of dynamic storage device capable of storing information and instructions, or it may be an EEPROM, CD-ROM or other optical disc storage, optical disc storage (including compressed optical discs, laser discs, optical discs, digital universal optical discs, Blu-ray discs, etc.), magnetic disk storage medium or other magnetic storage device, or any other medium capable of carrying or storing desired program code in the form of instructions or data structures and accessible by a computer, but is not limited thereto.

[0184] The memory 5003 is used to store application code that executes the present invention, and its execution is controlled by the processor 5001. The processor 5001 is used to execute the application code stored in the memory 5003 to implement the content shown in any of the foregoing method embodiments.

[0185] This invention also provides a computer-readable storage medium storing a computer program that, when executed by a processor, implements the above-described transformer adaptive spray cooling method.

[0186] Those skilled in the art will understand that embodiments of the present invention can be provided as methods, systems, or computer program products. Therefore, the present invention can take the form of a completely hardware embodiment, a completely software embodiment, or an embodiment combining software and hardware aspects. Furthermore, the present invention can take the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code. The solutions in the embodiments of the present invention can be implemented using various computer languages, such as the object-oriented programming language Java and the interpreted scripting language JavaScript.

[0187] This invention is described with reference to flowchart illustrations and / or block diagrams of methods, apparatus (systems), and computer program products according to embodiments of the invention. It will be understood that each block of the flowchart illustrations and / or block diagrams, as well as combinations of blocks in the flowchart illustrations and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, special-purpose computer, embedded processor, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, generate instructions for implementing the flowchart. Figure 1 One or more processes and / or boxes Figure 1 A device that provides the functions specified in one or more boxes.

[0188] These computer program instructions may also be stored in a computer-readable storage medium that can direct a computer or other programmable data processing device to function in a particular manner, such that the instructions stored in the computer-readable storage medium produce an article of manufacture including instruction means, which are implemented in a process Figure 1 One or more processes and / or boxes Figure 1 The function specified in one or more boxes.

[0189] These computer program instructions can also be loaded onto a computer or other programmable data processing equipment to cause a series of operational steps to be performed on the computer or other programmable equipment to produce a computer-implemented process, thereby providing instructions that execute on the computer or other programmable equipment for implementing the process. Figure 1 One or more processes and / or boxes Figure 1 The steps of the function specified in one or more boxes.

[0190] In summary, the present invention provides a transformer adaptive spray cooling method, system, equipment, and medium. Relying on multi-dimensional operating condition data and combining a hybrid model of federated learning and gradient boosting machines, it pre-outputs the predicted oil temperature of each main transformer at the next moment, breaking through the traditional passive response mode. It proactively intervenes through flow optimization before the temperature exceeds the limit, effectively solving the problems of unattended substations experiencing start-up and shutdown of cooling and untimely heat dissipation, ensuring the stability of the transformer insulation medium and extending equipment lifespan. Utilizing the federated learning architecture of "global parameter aggregation + local model fine-tuning," it achieves global knowledge sharing among multiple substations and adapts to the specific heat dissipation characteristics of different models and operating environments of main transformers through local adaptation training, avoiding the coarse "one-size-fits-all" control of traditional centralized models. Simultaneously, the original data "does not leave the station," only the model parameters are uploaded, strictly protecting data privacy and security in the power industry. The objective function is constructed with minimizing "first oil temperature deviation + spray energy consumption" as the core, and the optimal nozzle flow rate is solved by combining flow control constraints. Under high temperature and high load conditions, priority is given to ensuring cooling effect, while under low load or favorable weather conditions, priority is given to energy saving and consumption reduction, significantly reducing water and energy waste caused by traditional fixed-flow spraying.

[0191] The various embodiments in this specification are described in a progressive manner. For directly identical or similar parts of the embodiments, refer to each other. Each embodiment focuses on its differences from other embodiments. In particular, the system embodiments are basically similar to the method embodiments, so the description is relatively simple; relevant parts can be referred to the descriptions in the method embodiments. It should be noted that the technical features of the above embodiments can be combined arbitrarily. For the sake of brevity, not all possible combinations of the technical features in the above embodiments are described. However, as long as the combination of these technical features does not contradict each other, it should be considered within the scope of this specification.

[0192] The above description is only a preferred embodiment of the present invention. It should be noted that for those skilled in the art, several improvements and substitutions can be made without departing from the technical principles of the present invention, and these improvements and substitutions should also be considered within the scope of protection of the present invention.

Claims

1. A transformer adaptive spray cooling method, characterized in that, The method comprises: acquiring multi-dimensional working condition data of each main transformer in a substation; based on the multi-dimensional working condition data of each main transformer, using a machine learning model combining federated learning and gradient boosting machine to determine an oil temperature prediction value of each main transformer at the next time point through oil temperature prediction; constructing a flow control target function corresponding to each main transformer with the oil temperature prediction value, a first oil temperature deviation from a preset target oil temperature, and spray energy consumption, and solving each flow control target function according to the corresponding flow control constraint condition to obtain the optimal nozzle flow of the corresponding main transformer; based on the optimal nozzle flow of each main transformer, updating the water pump output power of the corresponding main transformer in real time to adaptively spray and cool each main transformer.

2. The transformer adaptive spray cooling method of claim 1, wherein, After the above steps, the method further comprises: calculating a second oil temperature deviation between the oil temperature prediction value and the actual oil temperature of each main transformer at the next time point, and adjusting the machine learning model parameters according to the second oil temperature deviation of all main transformers.

3. The transformer adaptive spray cooling method of claim 1, wherein, The multi-dimensional working condition data includes main transformer operation monitoring data, real-time weather data, and spray device execution data, wherein the main transformer operation monitoring data includes main transformer load value and main transformer oil temperature, the real-time weather data includes main transformer area temperature and main transformer area wind speed, and the spray device execution data includes nozzle flow and spray water temperature.

4. The transformer adaptive spray cooling method of claim 1, wherein, The method of determining the oil temperature prediction value of each main transformer at the next time point through oil temperature prediction based on the multi-dimensional working condition data of each main transformer comprises: based on a federated learning architecture, deploying a client in each substation and deploying a server in a regional power dispatch center; downloading a pre-trained lightweight gradient boosting machine model to each client through the server, so that each client fine-tunes the lightweight gradient boosting machine model using local historical main transformer operation data to obtain a personalized gradient boosting machine model corresponding to the client; inputting the multi-dimensional working condition data of each main transformer into the personalized gradient boosting machine model corresponding to the client to perform oil temperature prediction, and outputting the oil temperature prediction value of the corresponding main transformer at the next time point.

5. The transformer adaptive spray cooling method of claim 1, wherein, The method of constructing a flow control target function corresponding to each main transformer with the oil temperature prediction value, a first oil temperature deviation from a preset target oil temperature, and spray energy consumption, and solving each flow control target function according to the corresponding flow control constraint condition to obtain the optimal nozzle flow of the corresponding main transformer comprises: based on the dynamic weighted sum of the absolute value of the first oil temperature deviation and the nozzle flow, constructing a flow control target function corresponding to each main transformer; The particle swarm optimization algorithm is used to solve each of the flow control objective functions according to the corresponding flow control constraints, including upper and lower flow limits, oil temperature prediction uncertainty constraints, and flow rate change constraints, to obtain the optimal nozzle flow rate for the corresponding main transformer.

6. The transformer adaptive spray cooling method of claim 5, wherein, The sum of the absolute value of the first oil temperature deviation and the dynamic weighting factor of the nozzle flow rate is 1, and the adjustment strategy for the dynamic weighting factor of the absolute value of the first oil temperature deviation includes: When the first oil temperature deviation is within the safe deviation range, the dynamic weighting factor of the absolute value of the first oil temperature deviation is the first value; When the first oil temperature deviation is within the controllable deviation range, the dynamic weighting factor of the absolute value of the first oil temperature deviation is the sum of the weighted value of the ratio of the first oil temperature deviation to the length of the controllable deviation range and the first value. When the first oil temperature deviation is in the over-limit deviation range, the dynamic weighting factor of the absolute value of the first oil temperature deviation is the second value.

7. The transformer adaptive spray cooling method of claim 1, wherein, The step of updating the water pump output power of the corresponding main transformer in real time based on the optimal nozzle flow rate of each main transformer, so as to perform adaptive spray cooling for each main transformer, includes: Based on the optimal nozzle flow rate of each main transformer, the water pump power adjustment amount corresponding to the main transformer is derived through a pre-constructed flow-power mapping relationship. Based on the water pump power adjustment amount of each main transformer, the water pump output power of the corresponding main transformer is adjusted in real time to perform adaptive spray cooling on each main transformer.

8. A transformer adaptive spray cooling system, characterized in that, include: The data acquisition module is used to acquire multi-dimensional operating condition data of the main transformer in each substation; The oil temperature prediction module is used to determine the predicted oil temperature value of the corresponding main transformer at the next moment by using a machine learning model that combines federated learning and gradient boosting machine, based on the multi-dimensional operating condition data of each main transformer. The flow rate calculation module is used to construct the flow control objective function corresponding to each main transformer based on the first oil temperature deviation of the predicted oil temperature and the preset target oil temperature and the spray energy consumption, and to solve each flow control objective function according to the corresponding flow control constraints to obtain the optimal nozzle flow rate of the corresponding main transformer. The spray execution module is used to update the water pump output power of the corresponding main transformer in real time based on the optimal nozzle flow rate of each main transformer, so as to perform adaptive spray cooling on each main transformer.

9. A terminal device, comprising: The method includes a processor, a memory, and a computer program stored in the memory and configured to be executed by the processor, wherein the processor, when executing the computer program, implements the transformer adaptive spray cooling method as described in any one of claims 1 to 7.

10. A computer-readable storage medium, characterized in that, The computer-readable storage medium includes a stored computer program, wherein, when the computer program is executed, it controls the device containing the computer-readable storage medium to perform the transformer adaptive spray cooling method as described in any one of claims 1 to 7.

Images