An intelligent control method for a medical imaging device power distribution system based on artificial intelligence

Abstract

The application discloses an intelligent control method for a medical imaging equipment power distribution system based on artificial intelligence, relates to the technical field of medical equipment power distribution, and solves the technical problems that the existing medical equipment power distribution method has a lagging response, low monitoring precision of key parameters, insufficient fault recognition accuracy, and potential hidden dangers are difficult to be found in time, and finally leads to inefficient medical power distribution; power diagnosis data is acquired; a plurality of diagnosis data is generated based on the power diagnosis data; the diagnosis data is input into a fault diagnosis model to obtain a diagnosis result, the fault diagnosis model is obtained by constructing a CNN-SVM hybrid model and an LSTM model; the diagnosis result includes a fault diagnosis result and a fault prediction result; an execution strategy is generated based on the diagnosis result; diagnosis and treatment equipment data and real-time power load data are acquired; the execution strategy, the diagnosis and treatment equipment data, and the real-time power load data are input into a load optimization model to obtain a load optimization decision; and efficient power distribution of medical equipment is realized.

Description

Technical Field

[0001] This application belongs to the field of medical equipment power distribution technology, specifically an intelligent control method for medical imaging equipment power distribution system based on artificial intelligence. Background Technology

[0002] With the continuous development and reform of the medical system, the level of medical electrification in major hospitals is increasing day by day, and the number of medical electrical devices used by medical staff in their daily work is also increasing. Ensuring safe electricity use in hospitals has become a key issue of concern to all sectors of society.

[0003] Currently, the mainstream solution in the field of power distribution for medical imaging equipment is based on general building power distribution systems, coupled with manual operation and maintenance inspection modes. This approach meets the basic power distribution needs of the equipment through independent power distribution circuits, basic grounding protection, and manual load adjustment. In some scenarios, simple voltage and current monitoring instruments are also used for basic early warning. Since most medical equipment is directly related to patient safety, the timeliness and accuracy of power distribution adjustments are extremely important. However, existing power distribution methods rely excessively on manual meter reading and experience-based judgment, resulting in low monitoring accuracy for key parameters such as harmonics and grounding resistance in power distribution circuits, insufficient fault identification accuracy, and difficulty in timely detection of potential hazards. Furthermore, the time-consuming manual recording and analysis of large amounts of data leads to a lag in the power distribution system's adjustment response, failing to meet the power distribution needs of large hospitals' medical equipment. Therefore, there is an urgent need for an intelligent control method for medical imaging equipment power distribution systems based on artificial intelligence. Summary of the Invention

[0004] This application provides an intelligent control method for the power distribution system of medical imaging equipment based on artificial intelligence, which solves the technical problems of existing power distribution methods for medical equipment, such as slow response, low accuracy of key parameter monitoring, insufficient fault identification accuracy, and difficulty in timely detection of potential hidden dangers, ultimately leading to inefficient medical power distribution.

[0005] To achieve the above objectives, this application adopts the following technical solution: Firstly, an intelligent power distribution control method for medical imaging equipment based on artificial intelligence is provided, comprising: Acquire power diagnostic data; the power diagnostic data includes power diagnostic data one, power diagnostic data two, and power diagnostic data three; power diagnostic data one consists of several power-related parameters collected by the hospital's main control layer on the power distribution side, including three-phase electrical parameters, active power, reactive power, electrical energy, total harmonic distortion, and other related parameters for each load circuit; power diagnostic data two consists of several power-related parameters collected by the branch circuit layer, including current, voltage, impedance, and power for each branch circuit; power diagnostic data three consists of several power-related parameters collected by the equipment terminal layer, including current, voltage, and power for each medical imaging device; generate several diagnostic data based on the power diagnostic data; The diagnostic data is input into a fault diagnosis model to obtain diagnostic results. The fault diagnosis model is constructed from a CNN-SVM hybrid model and an LSTM model. The diagnostic results include fault diagnosis results and fault prediction results. An execution strategy is generated based on the diagnostic results; data from the diagnostic equipment and real-time power load data are acquired; the execution strategy, diagnostic equipment data, and real-time power load data are input into the load optimization model to obtain a load optimization decision.

[0006] Based on the above technical solution, in the intelligent control method for power distribution system of medical imaging equipment based on artificial intelligence provided in this application, the following steps are taken: acquiring power diagnostic data; generating several diagnostic data based on the power diagnostic data; inputting the diagnostic data into a fault diagnosis model to obtain diagnostic results, the fault diagnosis model being constructed from a CNN-SVM hybrid model and an LSTM model; the diagnostic results include fault diagnosis results and fault prediction results; generating an execution strategy based on the diagnostic results; acquiring medical equipment data and real-time power load data; inputting the execution strategy, medical equipment data, and real-time power load data into a load optimization model to obtain load optimization decisions; by using a fault diagnosis model constructed from a CNN-SVM hybrid model and an LSTM model, the system simultaneously detects and warns of fault types, ensuring timely detection of faults in power distribution equipment; and simultaneously reviewing and optimizing decisions based on the diagnostic results, thus constructing a full-process intelligent power distribution control system. This system specifically addresses pain points such as reliance on manual labor, poor adaptability, and delayed response, achieving efficient power distribution for medical equipment with accurate power distribution data perception, early fault warning, and dynamic load optimization.

[0007] In conjunction with the first aspect above, in one possible implementation, the generation of several diagnostic data based on power diagnostic data includes: Extract Power Diagnostic Data 1, Power Diagnostic Data 2, and Power Diagnostic Data 3 from the power diagnostic data; extract the power parameter data corresponding to each load circuit in Power Diagnostic Data 1; extract the power parameter data corresponding to each branch circuit in Power Diagnostic Data 2; extract the power parameter data corresponding to each terminal device in Power Diagnostic Data 3; The power parameter data includes several parameter items and corresponding parameter value sequences; the parameter items include corresponding current, voltage, and power, etc. The parameter value sequences of several parameter items corresponding to each load loop are integrated to obtain the diagnostic data corresponding to the load loop; the columns of the diagnostic data represent each parameter item, and the rows represent the sampling time of the parameter value corresponding to the parameter item. The parameter value sequences of several parameter items corresponding to each branch loop are integrated to obtain the diagnostic data corresponding to the branch loop; By integrating the parameter value sequences of several parameter items corresponding to each terminal device, the diagnostic data corresponding to the terminal device is obtained. Then set the numbers corresponding to the load circuit, branch circuit, and terminal equipment to the numbers of the corresponding diagnostic data.

[0008] In conjunction with the first aspect above, in one possible implementation, the fault diagnosis model is constructed from a CNN-SVM hybrid model and an LSTM model, including: An input data processing layer is constructed, which includes a diagnostic data input terminal and several branch port groups, and each branch port group is assigned a corresponding number, the number of which is consistent with the matrix number of the diagnostic data; each branch port group includes a fault detection data output terminal, a dynamic adjustment data input terminal, and a fault prediction data output terminal. Connect the fault detection data output terminal corresponding to the branch port group of the input data processing layer to the input terminal of the CNN-SVM hybrid model, and connect the output terminal of the CNN-SVM hybrid model to the dynamic adjustment data input terminal of the input data processing layer. Connect the fault prediction data output of the input data processing layer to the input of the LSTM model; Thus, a fault diagnosis model is constructed; the input end of the fault diagnosis model is the diagnostic data input end corresponding to the input data processing layer, and the output end of the fault diagnosis model includes the output ends of several CNN-SVM hybrid models and the output ends of LSTM models.

[0009] In conjunction with the first aspect above, in one possible implementation, the input data processing layer is used to perform the following steps: S1: Obtain the number corresponding to the diagnostic data, and match the corresponding branch port group based on the number; S2: Obtain the set diagnostic sequence length and the parameter value sequence corresponding to each parameter item in the diagnostic data. Based on the diagnostic sequence length, extract the diagnostic sequence of the corresponding parameter item from the parameter value sequence, and output the diagnostic sequence corresponding to each parameter item through the fault detection data output terminal in the branch port group. S3: Obtain fault diagnosis results; generate parameter acquisition step size based on fault diagnosis results; the fault diagnosis results include the item level corresponding to several fault items and the confidence level corresponding to the item level; the fault diagnosis results are input to the input data processing layer through the dynamic adjustment data input terminal in the branch port group; S4: Extract the parameter value sequence corresponding to each parameter item in the diagnostic data, fit the parameter value in the parameter value sequence into the parameter transformation curve corresponding to the parameter item, extract several parameter values ​​for prediction based on the parameter acquisition step size in the parameter transformation curve, and integrate each parameter value into the prediction sequence corresponding to the parameter item. S5: Output the prediction sequence corresponding to each parameter item through the fault prediction data output terminal in the branch port group.

[0010] In conjunction with the first aspect above, in one possible implementation, the step of generating the parameter acquisition step size based on the fault diagnosis result includes: Extract the item level and confidence level corresponding to several fault items from the fault diagnosis results; Obtain the initial step size, substitute the project level and confidence level corresponding to each fault item, along with the initial step size, into the set step size adjustment function to obtain the parameter acquisition step size; one expression of the step size adjustment function includes:

[0011] in, The adjusted parameter acquisition step size. The initial step size is the time interval between consecutive acquisition times corresponding to the power diagnostic data. This is the initial step size adjustment function; Let n be the project level corresponding to the fault item. This represents the confidence level of the project level corresponding to the fault item number n.

[0012] In conjunction with the first aspect above, one possible implementation method for obtaining the initial step size includes: Acquire several historical diagnostic test data and corresponding test fault diagnosis results for the diagnostic test data; the test fault diagnosis results are obtained by experts based on the diagnostic test data analysis, including the item level corresponding to each fault item; the diagnostic test data includes several diagnostic data groups; each diagnostic data group contains diagnostic data corresponding to each load circuit, branch circuit, and terminal device; the data acquisition step size is different within each diagnostic data group; Input each set of diagnostic data into the fault diagnosis model in sequence to obtain the corresponding diagnostic results; Extract the fault diagnosis results from the diagnostic results; extract the item level corresponding to each fault item in the fault diagnosis results, and the confidence level corresponding to the item level; extract the item level corresponding to each fault item in the test fault diagnosis results, and mark it as the standard item level; calculate the diagnostic accuracy corresponding to the diagnostic data set based on the standard item level, item level, and confidence level corresponding to each fault item. Calculate the average diagnostic accuracy of different diagnostic data groups under the same acquisition step size, and use it as the average diagnostic accuracy corresponding to the acquisition step size. Calculate the average diagnostic accuracy under each acquisition step size in turn; use the acquisition step size corresponding to the maximum average diagnostic accuracy as the initial step size.

[0013] In conjunction with the first aspect above, in one possible implementation, the diagnostic accuracy corresponding to the diagnostic data set is calculated based on the standard item level, item level, and confidence level corresponding to each fault item, including: The standard item levels corresponding to each fault item are integrated into a standard level feature vector according to the set sorting order; The item levels corresponding to each fault item are integrated into a diagnostic level feature vector according to a set sorting order; the confidence levels corresponding to each fault item are integrated into a correction feature vector according to their set sorting order; it can be understood that the set sorting order is the order of the fault items pre-set by relevant personnel, and the sorting order of the standard level feature vector, diagnostic level feature vector, and correction feature vector is the same; the standard level feature vector, diagnostic level feature vector, and correction feature vector are substituted into a set accuracy calculation function to obtain the corresponding diagnostic accuracy; one expression of the accuracy calculation function includes:

[0014] in, For diagnostic accuracy, To correct the feature vector The corresponding element in; This is the standard grade feature vector; This is the feature vector for the diagnostic level.

[0015] In conjunction with the first aspect above, in one possible implementation, an execution strategy is generated based on the diagnostic results, including: Extract the fault diagnosis results and fault prediction results from the diagnostic results; Extract the item level corresponding to each fault item in the fault diagnosis results, and the predicted item level corresponding to each fault item in the fault prediction results; the item level includes 0, 1, 2, and 3; the predicted item level includes 0, 1, 2, and 3. Determine whether each faulty item corresponds to a project level of 2 or 3; If yes, then the execution strategy will be set to the emergency optimization scheduling strategy; If no, determine if the predicted project level corresponding to each faulty project is 2 or 3; if yes, set the execution strategy to emergency optimization scheduling strategy; if no, determine if the project level corresponding to each faulty project is 1; if yes, set the execution strategy to optimization scheduling strategy; if no, determine if the predicted project level corresponding to each faulty project is 1; if yes, set the execution strategy to optimization scheduling strategy; if no, set the execution strategy to maintenance strategy.

[0016] In conjunction with the first aspect above, in one possible implementation, a training method for the load optimization model includes: Acquire several sets of historical data, each set including execution strategy, diagnostic equipment data, and power load data, as well as corresponding load optimization decisions; the load optimization decisions include the optimized load scheduling amounts for each load loop, branch loop, and equipment terminal; the load optimization decisions are power load scheduling decisions determined by experts based on the execution strategy, diagnostic equipment data, and power load data; integrate the execution strategy, diagnostic equipment data, power load data, and corresponding load optimization decisions into several sets of training data and test data; The artificial intelligence model is trained using training data and tested using validation data. The final input consists of execution strategy, diagnostic equipment data, and power load data, and the output is a load optimization model for corresponding load optimization decisions. The artificial intelligence model can be a neural network model.

[0017] Secondly, this application provides a power distribution system for medical imaging equipment based on artificial intelligence, comprising: a data acquisition module, a data processing module, and a load scheduling module; wherein, The data acquisition module is used to obtain an initial step size and to acquire power diagnostic data based on the initial step size. The data processing module includes a data preprocessing unit, a fault diagnosis unit, and a strategy generation unit. The data preprocessing unit is used to generate several diagnostic data based on the power diagnostic data. The fault diagnosis unit is used to input the diagnostic data into the fault diagnosis model to obtain the diagnostic results. The fault diagnosis model is constructed from a CNN-SVM hybrid model and an LSTM model. The diagnostic results include fault diagnosis results and fault prediction results. The strategy generation unit is used to generate an execution strategy based on the diagnostic results, acquire diagnostic equipment data and real-time power load data, and input the execution strategy, diagnostic equipment data and real-time power load data into the load optimization model to obtain a load optimization decision. The load scheduling module is used to perform load scheduling based on the load scheduling amounts corresponding to each load loop, branch loop, and equipment terminal in the load optimization decision.

[0018] This application provides an AI-based intelligent power distribution control method for medical imaging equipment. The method acquires power diagnostic data; generates several diagnostic data points based on the power diagnostic data; inputs the diagnostic data into a fault diagnosis model to obtain diagnostic results. The fault diagnosis model is constructed using a CNN-SVM hybrid model and an LSTM model. The diagnostic results include fault diagnosis results and fault prediction results. An execution strategy is generated based on the diagnostic results. The method acquires data from the medical equipment and real-time power load data; inputs the execution strategy, medical equipment data, and real-time power load data into a load optimization model to obtain load optimization decisions. By using a fault diagnosis model constructed with a CNN-SVM hybrid model and an LSTM model, the method simultaneously detects and warns of fault types, ensuring timely detection of power distribution equipment faults. Furthermore, it reviews and optimizes decisions based on the diagnostic results, constructing a full-process intelligent power distribution control system. This system specifically addresses pain points such as reliance on manual labor, poor adaptability, and delayed response, achieving accurate power distribution data perception, early fault warning, and dynamic load optimization to achieve efficient power distribution for medical equipment.

[0019] It should be understood that the descriptions of technical features, technical solutions, beneficial effects, or similar language in this application do not imply that all features and advantages can be achieved in any single embodiment. Rather, it is understood that the description of a feature or beneficial effect means that a specific technical feature, technical solution, or beneficial effect is included in at least one embodiment. Therefore, the descriptions of technical features, technical solutions, or beneficial effects in this specification do not necessarily refer to the same embodiment. Furthermore, the technical features, technical solutions, and beneficial effects described in this embodiment can be combined in any suitable manner. Those skilled in the art will understand that embodiments can be implemented without one or more specific technical features, technical solutions, or beneficial effects of a particular embodiment. In other embodiments, additional technical features and beneficial effects may be identified in specific embodiments that do not embody all embodiments. Attached Figure Description

[0020] 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.

[0021] Figure 1 This is a schematic diagram illustrating the steps of the intelligent power distribution control method for medical imaging equipment in this application; Figure 2 This is a schematic diagram of the fault diagnosis model in this application; Figure 3 This is a schematic diagram of the module connections of the power distribution system for the medical imaging equipment in this application. Detailed Implementation

[0022] The technical solutions of this application will be clearly and completely described below with reference to the embodiments. Obviously, the described embodiments are only some embodiments of this application, and not all embodiments. Based on the embodiments of this application, all other embodiments obtained by those of ordinary skill in the art without creative effort are within the scope of protection of this application.

[0023] Please see Figure 1 The first aspect of this application provides an intelligent power distribution control method for medical imaging equipment based on artificial intelligence, comprising: Acquire power diagnostic data; the power diagnostic data includes power diagnostic data one, power diagnostic data two, and power diagnostic data three; power diagnostic data one consists of several power-related parameters collected by the hospital's main control layer on the power distribution side, including three-phase electrical parameters, active power, reactive power, electrical energy, total harmonic distortion, and other related parameters for each load circuit; power diagnostic data two consists of several power-related parameters collected by the branch circuit layer, including current, voltage, impedance, and power for each branch circuit; power diagnostic data three consists of several power-related parameters collected by the equipment terminal layer, including current, voltage, and power for each medical imaging device; generate several diagnostic data based on the power diagnostic data; The diagnostic data is input into the fault diagnosis model to obtain the diagnostic results. The fault diagnosis model is constructed from a CNN-SVM hybrid model and an LSTM model. The diagnostic results include fault diagnosis results and fault prediction results. The fault diagnosis results include the fault diagnosis results output by each CNN-SVM hybrid model, and the fault prediction results include the fault prediction results output by each LSTM model. Based on the diagnostic results, an execution strategy is generated, which includes routine load optimization, preventative load optimization, and emergency load optimization. The strategy involves acquiring diagnostic equipment data and real-time power load data. The diagnostic equipment data includes the equipment load and status of each device. The equipment load is the load required for equipment operation. The equipment status is the planned start / stop status of medical equipment corresponding to each department, as input by doctors in each hospital department. For example, if a doctor in the CT department inputs that the CT machine will be used at a certain time, then the machine's start / stop status will be "started" until it needs to be stopped at a certain time. The power load data includes the load scheduling amount corresponding to each load circuit, branch circuit, and device terminal. The execution strategy, diagnostic equipment data, and real-time power load data are input into the load optimization model to obtain a load optimization decision. The load optimization decision includes the load scheduling amount corresponding to each load circuit, branch circuit, and device terminal. When the scheduling amount is zero, it indicates that power is cut off to that load circuit, branch circuit, or device terminal.

[0024] Based on the above technical solution, the intelligent control method for power distribution system of medical imaging equipment based on artificial intelligence provided in this application can acquire power diagnostic data; generate several diagnostic data based on the power diagnostic data; input the diagnostic data into a fault diagnosis model to obtain diagnostic results, the fault diagnosis model being constructed from a CNN-SVM hybrid model and an LSTM model; the diagnostic results include fault diagnosis results and fault prediction results; generate execution strategies based on the diagnostic results; acquire medical equipment data and real-time power load data; input the execution strategy, medical equipment data, and real-time power load data into a load optimization model to obtain load optimization decisions; by using the fault diagnosis model constructed from the CNN-SVM hybrid model and the LSTM model, the system can simultaneously detect and warn of fault types, ensuring timely detection of faults in power distribution equipment; and simultaneously review and optimize decisions based on the diagnostic results, constructing a full-process intelligent power distribution control system to specifically address pain points such as reliance on manual labor, poor adaptability, and delayed response, achieving accurate perception of power distribution data, early warning of faults, and dynamic load optimization, thereby achieving efficient power distribution for medical equipment.

[0025] In one possible implementation, several diagnostic data are generated based on the power diagnostic data, including: extracting power diagnostic data one, power diagnostic data two, and power diagnostic data three from the power diagnostic data; extracting power parameter data corresponding to each load circuit in power diagnostic data one; extracting power parameter data corresponding to each branch circuit in power diagnostic data two; and extracting power parameter data corresponding to each terminal device in power diagnostic data three. Power parameter data contains several parameter items and corresponding parameter value sequences. Parameter items include corresponding current, voltage, and power, etc. It is understandable that the parameter items corresponding to different power parameter data may be the same or different. For example, the parameter items corresponding to power parameter data of different load circuits and branch circuits are generally consistent. However, the parameter items in the power parameter data of different terminal devices are generally inconsistent. This is because different terminal devices have both the same control parameters, such as the basic parameters of input current and voltage, and some device-specific control parameters. The parameter value sequences of several parameter items corresponding to each load circuit are integrated to obtain the diagnostic data corresponding to the load circuit; the columns of the diagnostic data represent each parameter item, and the rows represent the sampling time of the parameter value corresponding to the parameter item. The parameter value sequences of several parameter items corresponding to each branch loop are integrated to obtain the diagnostic data corresponding to the branch loop; By integrating the parameter value sequences of several parameter items corresponding to each terminal device, the diagnostic data corresponding to the terminal device is obtained. Then set the numbers corresponding to the load circuit, branch circuit, and terminal equipment to the numbers of the corresponding diagnostic data.

[0026] It is understood that the numbering of the diagnostic data is used to find the corresponding port branch during subsequent fault detection and simulation.

[0027] With the development of medical imaging technology, high-end medical imaging equipment such as CT and MRI are widely deployed in hospitals at all levels. These devices have high power consumption and extremely high requirements for power supply stability. Moreover, the equipment is distributed in different departments, forming a multi-level power distribution structure with a general control layer, branch circuit layer, and equipment terminal layer. Traditional single-circuit, simple monitoring power distribution methods can no longer adapt to complex power distribution networks. This embodiment classifies power diagnostic data in a progressive multi-level manner and analyzes the classification results separately to adapt to fault analysis and load scheduling of complex power distribution networks. This solves the problem of the traditional dispatching system's extensive management of multi-level power distribution networks and improves the overall management level of the hospital's power distribution system.

[0028] Please see Figure 2 In one possible implementation, the fault diagnosis model is constructed from a CNN-SVM hybrid model and an LSTM model, including: An input data processing layer is constructed, which includes a diagnostic data input terminal and several branch port groups. Each branch port group is assigned a corresponding number, and the number of the branch port group is consistent with the matrix number of the diagnostic data. Each branch port group includes a fault detection data output terminal, a dynamic adjustment data input terminal, and a fault prediction data output terminal. Connect the fault detection data output terminal corresponding to the branch port group of the input data processing layer to the input terminal of the CNN-SVM hybrid model, and connect the output terminal of the CNN-SVM hybrid model to the dynamic adjustment data input terminal of the input data processing layer. Connect the fault prediction data output of the input data processing layer to the input of the LSTM model; Thus, a fault diagnosis model is constructed; the input end of the fault diagnosis model is the diagnostic data input end corresponding to the input data processing layer, and the output end of the fault diagnosis model includes the output ends of several CNN-SVM hybrid models and the output ends of LSTM models; it can be understood that the LSTM model used in this embodiment has a variable-length sequence as input.

[0029] This embodiment integrates the CNN-SVM hybrid model and LSTM model trained with different training data by constructing a branch port group in the input data processing layer, and matches them by the number of different diagnostic data to achieve parallel fault analysis of each load loop, branch loop and terminal device, which effectively improves the efficiency of overall fault analysis. The CNN-SVM hybrid model in this embodiment is a collaborative model of CNN feature extraction and SVM accurate classification. The architecture of the CNN-SVM hybrid model includes a CNN feature extraction unit: adopting the SqueezeNet lightweight architecture to reduce parameter redundancy, containing 4 convolutional layers with 32 / 64 / 128 / 256 kernels, a kernel size of 3×3 main convolutional kernel + 1×5 temporal convolutional kernel combination, 4 hybrid pooling layers, and a batch processing layer. The normalization layer accelerates convergence, and finally, two fully connected layers with 1024 / 128 neurons each map the high-dimensional features into 128-dimensional feature vectors. The SVM classification unit uses the RBF kernel function, and optimizes the hyperparameters using a grid search method. In this embodiment, the penalty factor C=1.2 and the kernel function parameter σ=0.8 to construct a multi-classifier, achieving accurate differentiation of six types of faults. Cross-validation is used to improve the model's generalization ability and avoid overfitting. The interference feature filtering module trains interference feature templates based on a medical electromagnetic interference database, and uses cosine similarity comparison to remove pseudo-features with a similarity ≥85% to the interference template.

[0030] Specifically, the fault diagnosis implementation steps include: dividing the preprocessed standardized dataset into training and test sets in an 8:2 ratio, and simultaneously inputting them into the CNN module; By alternating between CNN convolutional and pooling layers, the hidden fault features in the data are deeply mined, such as the abnormal amplitude of the 3rd and 5th harmonics when the MR gradient amplifier harmonics exceed the standard, the persistently high current when the helium compressor circuit is overloaded, the parameter fluctuation features when the spectrometer circuit has poor contact, the zero-sequence current change when the CT high voltage generator circuit is grounded, the fluctuation features of the MR shim power supply, and the abnormal current features when the CT gantry circuit is overloaded. The 128-dimensional feature vector is output through a fully connected layer, and then the false features are removed by the interference feature filtering module. The purified feature vectors are input into the optimized SVM classifier, with high-weight fault features being prioritized for judgment. The output includes fault type and fault location, accurate to each load circuit and device of MR / CT, such as gradient amplifier, radio frequency amplifier, helium compressor, spectrometer, high voltage generator, rack, reconstruction machine, and fault level. The diagnostic delay is controlled within 64ms. By combining the waveform data of each load circuit recorded in the fault waveform recording, the diagnostic results are verified a second time to ensure that the fault location accuracy is ≥99.98%.

[0031] The CNN-SVM hybrid model in this embodiment is trained using a dataset of 60,000+ PSCAD simulations of 12 different fault scenarios. The training environment is an NVIDIA RTX 3090 GPU, using the Adam optimizer with a learning rate of 0.001 and a decay coefficient of 0.9. After 800 iterations, the loss function value converges to 0.003, and the test set accuracy reaches 99.98%. It can effectively distinguish different causes of the same type of fault, such as excessive harmonics caused by sudden changes in equipment load or component aging. The training early stopping mechanism automatically stops training and saves the current optimal model parameters when the validation set accuracy does not improve for 100 consecutive iterations (i.e., fluctuation ≤ 0.01%), or the loss function value is ≤ 0.005 for 50 consecutive iterations.

[0032] In one possible implementation, the input data processing layer is used to perform the following steps: S1: Obtain the number corresponding to the diagnostic data, and match the corresponding branch port group based on the number; S2: Obtain the set diagnostic sequence length and the parameter value sequence corresponding to each parameter item in the diagnostic data. Based on the diagnostic sequence length, extract the diagnostic sequence for the corresponding parameter item from the parameter value sequence, and output the diagnostic sequence corresponding to each parameter item through the fault detection data output terminal within the branch port group. Specifically, a parameter value sequence length of 1000 indicates that 1000 parameter values ​​have been collected for that parameter item; a set diagnostic sequence length of 100 indicates that the diagnostic data is extracted... ;in ; t = 1, 2, ..., 20; that is, when t = 1, This is the diagnostic sequence corresponding to parameter item number 1. It can be understood that the smaller the number in the row direction of the diagnostic data, the further away the acquisition time corresponding to the parameter value is from the current time. The diagnostic sequence extracted in this embodiment is the set of parameter values ​​closest to the current time. S3: Obtain the fault diagnosis result; it can be understood that the fault diagnosis result here is the fault diagnosis result output by the CNN-SVM hybrid model in a branch port group, and the parameter acquisition step size calculated subsequently is also the parameter acquisition step size of the LSTM model under the corresponding branch port group; generate the parameter acquisition step size based on the fault diagnosis result; the fault diagnosis result includes the item level corresponding to several fault items and the confidence level corresponding to the item level; the fault diagnosis result is input to the input data processing layer through the dynamic adjustment data input terminal in the branch port group; S4: Extract the parameter value sequence corresponding to each parameter item in the diagnostic data, fit the parameter value in the parameter value sequence into the parameter transformation curve corresponding to the parameter item, and use interpolation method for fitting; based on the parameter acquisition step size, extract several parameter values ​​for prediction in the parameter transformation curve, and integrate each parameter value into the prediction sequence corresponding to the parameter item. S5: Output the prediction sequence corresponding to each parameter item through the fault prediction data output terminal in the branch port group.

[0033] In one possible implementation, the parameter acquisition step size is generated based on the fault diagnosis results; including: Extract the item level and confidence level corresponding to several fault items from the fault diagnosis results; Obtain the initial step size, substitute the project level and confidence level corresponding to each fault item, along with the initial step size, into the set step size adjustment function to obtain the parameter acquisition step size; one expression of the step size adjustment function includes:

[0034] in, The adjusted parameter acquisition step size. The initial step size is the time interval between consecutive acquisition times corresponding to the power diagnostic data. This is the initial step size adjustment function; Let n be the project level corresponding to the fault item. The confidence level of the fault item corresponding to the number n is used. In this embodiment, the step size of the data sequence used for prediction is adjusted based on the calculated project level and confidence level of each fault item using the above formula. Under normal operating conditions, the data is generally constant or changes slowly and regularly. Under normal operating conditions, a small amount of data is sufficient to simulate the accuracy result. When the project level corresponding to the fault item is higher, it indicates that the fault item reflects a higher degree of fault severity. If the corresponding confidence level is higher, it indicates that the reliability of the project level is higher. The higher the fault level and the higher the confidence level, the higher the risk of power distribution system faults. The deviation between the current data and the data under normal conditions is larger, generally showing irregular fluctuations. At this time, the data under normal conditions is used. The accuracy of predictions based on data will be low. To ensure the accuracy of predictions, the amount of data needs to be increased adaptively. More data is needed to accurately grasp the operation of the power distribution system, resulting in smaller acquisition step sizes and more intensive parameter sampling. A parameter acquisition step size that dynamically adapts to the current fault state of the power distribution system is also required. Furthermore, in this embodiment, the relationship between the calculated parameter acquisition step size and the initial step size is as follows: the initial step size is an integer multiple of the parameter acquisition step size. This means that the subsequent data sequence input into the LSTM model must contain all the real data from the diagnostic data, as well as the virtual data simulated from the real data. This further ensures the accuracy of the prediction results. This approach achieves both data expansion and guarantees the authenticity of the data source. One way to express the influence quantization function is as follows:

[0035] in, Let n be the project level corresponding to the fault item. The confidence level of the project level corresponding to the fault item with number n; =1, 2, ..., ; This represents the total number of fault items in the fault diagnosis result; To set the confidence threshold, the confidence threshold in this embodiment is set to 78%; The standard step size is set by experts based on experience, and the specific value can be equal to the initial step size. When the confidence level of a project is less than the set confidence threshold, it indicates that the fault diagnosis result corresponding to that fault project level is unreliable. To avoid the influence of these unreliable fault diagnosis results on the step size influence factor, the corresponding single-item step size influence factor is set to 0. When the project level corresponding to the fault project is higher, it indicates that the fault risk of the power distribution system is higher, and the deviation between the current data and the data under normal conditions is greater. At this time, the accuracy of the prediction result using the data under normal conditions will be lower, and it is necessary to reduce the step size and increase the amount of data to ensure the accuracy of the prediction result.

[0036] In this embodiment, the prediction sequence of the LSTM model is dynamically generated based on the fault diagnosis results. The higher the fault level and the greater the confidence level, the higher the fault risk of the power distribution system and the more irregular the parameter fluctuations. At this time, the step size adjustment function will generate a smaller acquisition step size to achieve dense sampling. Dense sampling data can completely capture key features such as parameter mutations and fluctuations under fault conditions, providing richer time-series data for fault prediction by the LSTM model and improving the reliability of the prediction results.

[0037] Confidence threshold filtering for invalid fault information: Fault levels with confidence scores below the threshold are judged as invalid information, and their corresponding step size influence factor is 0. This avoids the interference of unreliable fault diagnosis results on step size adjustment, ensures the rationality and accuracy of step size adjustment, and prevents unnecessary dense sampling caused by invalid information.

[0038] The initial step size is an integer multiple of the final step size: Through mathematical design, it is ensured that the adjusted acquisition step size is an integer multiple of the initial step size, so that the prediction sequence input to the LSTM model not only contains all the real acquisition data, but also generates accurate virtual supplementary data through fitting the real data. This achieves data expansion while ensuring the authenticity of the data source, and further improves the accuracy of fault prediction.

[0039] Under normal conditions, low-density data acquisition is performed using the optimal initial step size, which reduces the amount of data generated, lowers data transmission bandwidth and storage device occupancy, and also reduces the amount of data processed by AI models, thus reducing GPU / CPU computing power consumption. Under fault conditions, intensive sampling is performed only on high-risk circuits / equipment, rather than indiscriminate intensive sampling across the entire system, further reducing the generation of invalid data and ensuring that computing resources are concentrated on fault-related data processing, thereby improving the response speed of fault diagnosis. This dynamic switching significantly reduces the hardware computing power consumption and operation and maintenance costs of the power distribution system.

[0040] In one possible implementation, the initial step size can be obtained by: acquiring several historical diagnostic test data and corresponding test fault diagnosis results; the test fault diagnosis results are obtained by experts based on the diagnostic test data analysis, including the item level corresponding to each fault item; the diagnostic test data includes several diagnostic data groups; each diagnostic data group contains diagnostic data corresponding to each load circuit, branch circuit, and terminal device; the data acquisition step size is different within each diagnostic data group; Input each set of diagnostic data into the fault diagnosis model in sequence to obtain the corresponding diagnostic results; Extract the fault diagnosis results from the diagnostic results; extract the item level corresponding to each fault item in the fault diagnosis results, and the confidence level corresponding to the item level; extract the item level corresponding to each fault item in the test fault diagnosis results, and mark it as the standard item level; calculate the diagnostic accuracy corresponding to the diagnostic data set based on the standard item level, item level, and confidence level corresponding to each fault item. Calculate the average diagnostic accuracy of different diagnostic data groups under the same acquisition step size, and use it as the average diagnostic accuracy corresponding to the acquisition step size. Calculate the average diagnostic accuracy under each acquisition step size in turn; use the acquisition step size corresponding to the maximum average diagnostic accuracy as the initial step size.

[0041] Understandably, since the CNN-SVM model is suitable for analyzing short-term data sequences, if the data sequence is too short, the accuracy of the analysis results will decrease, and if the data sequence is too long, the processing efficiency of the CNN-SVM model will decrease. At the same time, noise may also cause a slight decrease in the accuracy of the analysis results for long data sequences. Therefore, this embodiment ensures that the CNN-SVM model in the fault diagnosis model can comprehensively consider the processing efficiency and noise impact when analyzing the collected data, select the most suitable collected data for fault analysis, and thus obtain a highly accurate diagnostic result.

[0042] In one possible implementation, the diagnostic accuracy of the diagnostic data set is calculated based on the standard item level, item level, and confidence level corresponding to each fault item, including: The standard item levels corresponding to each fault item are integrated into a standard level feature vector according to the set sorting order; The item levels corresponding to each fault item are integrated into a diagnostic level feature vector according to a set sorting order; the confidence levels corresponding to each fault item are integrated into a correction feature vector according to their set sorting order; it can be understood that the set sorting order is the order of the fault items pre-set by relevant personnel, and the sorting order of the standard level feature vector, diagnostic level feature vector, and correction feature vector is the same; the standard level feature vector, diagnostic level feature vector, and correction feature vector are substituted into a set accuracy calculation function to obtain the corresponding diagnostic accuracy; one expression of the accuracy calculation function includes:

[0043] in, For diagnostic accuracy, To correct the feature vector The corresponding element in; This is the standard grade feature vector; Let be the diagnostic level feature vector; the corresponding diagnostic accuracy is calculated using the above formula; the smaller the difference between the item level of each fault item and the standard item level, the higher the accuracy of the fault diagnosis result, and the higher the corresponding diagnostic accuracy; then, the confidence level is used to correct the obtained diagnostic accuracy to obtain the final diagnostic accuracy; when the average confidence level is higher, the reliability of the obtained diagnostic accuracy is higher, and the corresponding corrected diagnostic accuracy is greater than the corrected diagnostic accuracy value with a lower average confidence level; it is understood that the above formula is only one expression of the accurate calculation function, and other formulas that conform to the above logic and achieve the above function can also be used as substitutes; Diagnostic accuracy refers to the degree of matching between the fault diagnosis results in the output diagnostic results and the actual test fault diagnosis results when using the diagnostic data set obtained under the corresponding acquisition step size for fault diagnosis. The higher the consistency between the item level of each fault item in the fault diagnosis results and the item level of the test fault diagnosis results, the higher the degree of matching between the fault diagnosis results and the test fault diagnosis results, that is, the higher the accuracy of fault diagnosis using the model, and the larger the corresponding diagnostic accuracy value should be set. This facilitates the selection of the most suitable acquisition step size and the data acquisition, so that during fault diagnosis, both the accuracy of fault diagnosis can be guaranteed and the amount of data processing can be reduced as much as possible.

[0044] In this embodiment, the initial step size is verified by multiple sets of diagnostic test data with different acquisition step sizes. The step size corresponding to the maximum average diagnostic accuracy is selected as the benchmark. This step size ensures that the CNN-SVM model can acquire enough effective feature points, avoiding the decrease in diagnostic accuracy due to the data sequence being too short, and also avoids the noise and redundancy introduced by the data sequence being too long, thus ensuring the processing efficiency of the model.

[0045] In one possible implementation, an execution strategy is generated based on the diagnostic results, including: extracting fault diagnosis results and fault prediction results from the diagnostic results; Extract the item level corresponding to each fault item in the fault diagnosis results, and the predicted item level corresponding to each fault item in the fault prediction results; the item level includes 0, 1, 2, and 3; the predicted item level includes 0, 1, 2, and 3. Determine whether each faulty item corresponds to a project level of 2 or 3; If yes, then the execution strategy will be set to the emergency optimization scheduling strategy; If no, determine if the predicted project level corresponding to each faulty project is 2 or 3; if yes, set the execution strategy to emergency optimization scheduling strategy; if no, determine if the project level corresponding to each faulty project is 1; if yes, set the execution strategy to optimization scheduling strategy; if no, determine if the predicted project level corresponding to each faulty project is 1; if yes, set the execution strategy to optimization scheduling strategy; if no, set the execution strategy to maintenance strategy.

[0046] In this embodiment, the project levels include four levels: 0, 1, 2, and 3, representing normal, minor anomaly, moderate anomaly, and severe anomaly, respectively. When all project levels corresponding to each fault project are normal, it indicates that the power distribution system is working normally and no adjustment is needed in the short term. When the project levels corresponding to each fault project only include minor anomaly, or minor anomaly and normal, it indicates that the power distribution system has some minor anomalies. In this case, the minor power distribution anomalies can be resolved by optimizing the power distribution. When the project levels corresponding to each fault project include moderate or severe anomalies, it indicates that the power distribution system is working in an abnormal state. It is necessary to stop the power distribution of the fault projects corresponding to the moderate or severe anomalies in a timely manner, issue an alarm, and carry out maintenance promptly.

[0047] In one possible implementation, one training method for the load optimization model includes: Acquire several sets of historical data, each set including execution strategy, diagnostic equipment data, and power load data, as well as corresponding load optimization decisions; the load optimization decisions include the optimized load scheduling amounts for each load loop, branch loop, and equipment terminal; the load optimization decisions are power load scheduling decisions determined by experts based on the execution strategy, diagnostic equipment data, and power load data; integrate the execution strategy, diagnostic equipment data, power load data, and corresponding load optimization decisions into several sets of training data and test data; The artificial intelligence model is trained using training data and tested using validation data. The final result is a load optimization model with the execution strategy, diagnostic equipment data, and power load data as inputs and the corresponding load optimization decision as output. The artificial intelligence model can be a neural network model, and the QDN model is used in this embodiment.

[0048] Please see Figure 3 Secondly, this application provides a power distribution system for medical imaging equipment based on artificial intelligence, including: a data acquisition module, a data processing module, a load scheduling module, and a database; wherein, The data acquisition module is used to obtain an initial step size and to acquire power diagnostic data based on the initial step size. The data processing module includes a data preprocessing unit, a fault diagnosis unit, and a strategy generation unit. The data preprocessing unit is used to generate several diagnostic data based on the power diagnostic data. The fault diagnosis unit is used to input the diagnostic data into the fault diagnosis model to obtain the diagnostic results. The fault diagnosis model is constructed from a CNN-SVM hybrid model and an LSTM model. The diagnostic results include fault diagnosis results and fault prediction results. The strategy generation unit is used to generate an execution strategy based on the diagnostic results, acquire diagnostic equipment data and real-time power load data, and input the execution strategy, diagnostic equipment data and real-time power load data into the load optimization model to obtain a load optimization decision. The load scheduling module is used to perform load scheduling based on the load scheduling amounts corresponding to each load loop, branch loop, and equipment terminal in the load optimization decision. Database: Used to store relevant data in the power distribution system of medical imaging equipment.

[0049] Some of the data in the above formula are calculated by removing dimensions and taking their numerical values. The formula is the closest to the real situation obtained by software simulation of a large amount of collected data. The preset parameters and preset thresholds in the formula are set by those skilled in the art according to the actual situation or obtained through simulation of a large amount of data.

[0050] How this application works: This system can acquire power diagnostic data; generate several diagnostic data based on the power diagnostic data; input the diagnostic data into a fault diagnosis model to obtain diagnostic results. The fault diagnosis model is constructed using a hybrid CNN-SVM model and an LSTM model. The diagnostic results include fault diagnosis results and fault prediction results. An execution strategy is generated based on the diagnostic results. The system acquires data from medical equipment and real-time power load data; inputs the execution strategy, medical equipment data, and real-time power load data into a load optimization model to obtain load optimization decisions. By using a fault diagnosis model constructed using a hybrid CNN-SVM model and an LSTM model, the system can simultaneously detect and warn of fault types, ensuring timely detection of faults in power distribution equipment. Simultaneously, it reviews and optimizes decisions based on the diagnostic results, constructing a full-process intelligent power distribution control system. This system specifically addresses pain points such as reliance on manual labor, poor adaptability, and delayed response, achieving accurate perception of power distribution data, early warning of faults, and dynamic load optimization to achieve efficient power distribution for medical equipment.

[0051] The above embodiments are only used to illustrate the technical methods of this application and are not intended to limit it. Although this application has been described in detail with reference to preferred embodiments, those skilled in the art should understand that modifications or equivalent substitutions can be made to the technical methods of this application without departing from the spirit and scope of the technical methods of this application.

Claims

1. A method for intelligent power distribution control of medical imaging equipment based on artificial intelligence, characterized in that, include: Acquire power diagnostic data; Several diagnostic data points are generated based on power diagnostic data; The diagnostic data is input into a fault diagnosis model to obtain diagnostic results. The fault diagnosis model is constructed from a CNN-SVM hybrid model and an LSTM model. The diagnostic results include fault diagnosis results and fault prediction results. The fault diagnosis model is constructed from a CNN-SVM hybrid model and an LSTM model, including: An input data processing layer is constructed, comprising a diagnostic data input terminal and several branch port groups; each branch port group includes a fault detection data output terminal, a dynamic adjustment data input terminal, and a fault prediction data output terminal; the diagnostic data input terminal is used to input diagnostic data and its corresponding number. The fault detection data output terminal corresponding to the branch port group of the input data processing layer is connected to the input terminal of the CNN-SVM hybrid model, and the output terminal of the CNN-SVM hybrid model is connected to the dynamic adjustment data input terminal of the input data processing layer; the fault detection data output terminal is used to output the diagnostic sequence to the CNN-SVM hybrid model; the dynamic adjustment data input terminal is used to input the fault diagnosis result output by the CNN-SVM hybrid model. The fault prediction data output terminal of the input data processing layer is connected to the input terminal of the LSTM model; the fault prediction data output terminal is used to output the prediction sequence to the LSTM model; The input data processing layer is used to perform the following steps: S1: Obtain the number corresponding to the diagnostic data, and match the corresponding branch port group based on the number; S2: Obtain the set diagnostic sequence length and the parameter value sequence corresponding to each parameter item in the diagnostic data. Based on the diagnostic sequence length, extract the diagnostic sequence of the corresponding parameter item from the parameter value sequence, and output the diagnostic sequence corresponding to each parameter item through the fault detection data output terminal in the branch port group. S3: Obtain fault diagnosis results; generate parameter acquisition step size based on fault diagnosis results; the fault diagnosis results include the item level corresponding to several fault items and the confidence level corresponding to the item level; S4: Extract the parameter value sequence corresponding to each parameter item in the diagnostic data, fit the parameter value in the parameter value sequence into the parameter transformation curve corresponding to the parameter item, extract several parameter values ​​for prediction based on the parameter acquisition step size in the parameter transformation curve, and integrate each parameter value into the prediction sequence corresponding to the parameter item. S5: Output the prediction sequence corresponding to each parameter item through the fault prediction data output terminal in the branch port group; Thus, a fault diagnosis model is constructed; the input end of the fault diagnosis model is the diagnostic data input end corresponding to the input data processing layer, and the output end of the fault diagnosis model includes the output ends of several CNN-SVM hybrid models and the output ends of LSTM models; An execution strategy is generated based on the diagnostic results; data from the diagnostic equipment and real-time power load data are acquired; the execution strategy, diagnostic equipment data, and real-time power load data are input into the load optimization model to obtain a load optimization decision.

2. The intelligent power distribution control method for medical imaging equipment based on artificial intelligence according to claim 1, characterized in that, The generation of several diagnostic data based on power diagnostic data includes: Extract Power Diagnostic Data 1, Power Diagnostic Data 2, and Power Diagnostic Data 3 from the power diagnostic data; extract the power parameter data corresponding to each load circuit in Power Diagnostic Data 1; extract the power parameter data corresponding to each branch circuit in Power Diagnostic Data 2; extract the power parameter data corresponding to each terminal device in Power Diagnostic Data 3; The power parameter data includes several parameter items and the corresponding parameter value sequences for each parameter item; By integrating the parameter value sequences of several parameter items corresponding to each load circuit, the diagnostic data corresponding to the load circuit is obtained. The parameter value sequences of several parameter items corresponding to each branch loop are integrated to obtain the diagnostic data corresponding to the branch loop; By integrating the parameter value sequences of several parameter items corresponding to each terminal device, the diagnostic data corresponding to the terminal device is obtained. Then set the numbers corresponding to the load circuit, branch circuit, and terminal equipment to the numbers of the corresponding diagnostic data.

3. The intelligent power distribution control method for medical imaging equipment based on artificial intelligence according to claim 1, characterized in that, The parameter acquisition step size generated based on the fault diagnosis results includes: Extract the item level and confidence level corresponding to several fault items from the fault diagnosis results; Obtain the initial step size, substitute the project level and confidence level corresponding to each fault item, along with the initial step size, into the set step size adjustment function to obtain the parameter acquisition step size; one expression of the step size adjustment function includes: in, The adjusted parameter acquisition step size. The initial step size is the time interval between consecutive acquisition times corresponding to the power diagnostic data. This is the initial step size adjustment function; Let n be the project level corresponding to the fault item. This represents the confidence level of the project level corresponding to the fault item number n.

4. The intelligent power distribution control method for medical imaging equipment based on artificial intelligence according to claim 3, characterized in that, One method for obtaining the initial step size includes: Acquire a number of historical diagnostic test data and the corresponding test fault diagnosis results for the diagnostic test data; the diagnostic test data includes several diagnostic data groups; each diagnostic data group contains diagnostic data corresponding to each load circuit, branch circuit, and terminal device; the data acquisition step size is different within each diagnostic data group; Input each set of diagnostic data into the fault diagnosis model in sequence to obtain the corresponding diagnostic results; Extract the fault diagnosis results from the diagnostic results; extract the item level corresponding to each fault item in the fault diagnosis results, and the confidence level corresponding to the item level; extract the item level corresponding to each fault item in the test fault diagnosis results, and mark it as the standard item level; calculate the diagnostic accuracy corresponding to the diagnostic data set based on the standard item level, item level, and confidence level corresponding to each fault item. Calculate the average diagnostic accuracy of different diagnostic data groups under the same acquisition step size, and use it as the average diagnostic accuracy corresponding to the acquisition step size. Calculate the average diagnostic accuracy under each acquisition step size in turn; use the acquisition step size corresponding to the maximum average diagnostic accuracy as the initial step size.

5. The intelligent power distribution control method for medical imaging equipment based on artificial intelligence according to claim 4, characterized in that, The diagnostic accuracy of the diagnostic data set is calculated based on the standard item level, item level, and confidence level corresponding to each fault item, including: The standard item levels corresponding to each fault item are integrated into a standard level feature vector according to the set sorting order; The item levels corresponding to each fault item are integrated into a diagnostic level feature vector according to a set order; the confidence levels corresponding to each fault item are integrated into a correction feature vector according to a set order; the standard level feature vector, diagnostic level feature vector, and correction feature vector are substituted into a set accuracy calculation function to obtain the corresponding diagnostic accuracy.

6. The intelligent power distribution control method for medical imaging equipment based on artificial intelligence according to claim 1, characterized in that, Based on the diagnostic results, an execution strategy is generated, including: Extract the fault diagnosis results and fault prediction results from the diagnostic results; Extract the item level corresponding to each fault item in the fault diagnosis results, and the predicted item level corresponding to each fault item in the fault prediction results; the item level includes 0, 1, 2, and 3; the predicted item level includes 0, 1, 2, and 3. Determine whether each faulty item corresponds to a project level of 2 or 3; If yes, then the execution strategy will be set to the emergency optimization scheduling strategy; If no, determine if the predicted project level corresponding to each faulty project is 2 or 3; if yes, set the execution strategy to emergency optimization scheduling strategy; if no, determine if the project level corresponding to each faulty project is 1; if yes, set the execution strategy to optimization scheduling strategy; if no, determine if the predicted project level corresponding to each faulty project is 1; if yes, set the execution strategy to optimization scheduling strategy; if no, set the execution strategy to maintenance strategy.

7. The intelligent power distribution control method for medical imaging equipment based on artificial intelligence according to claim 1, characterized in that, One training method for the load optimization model includes: Acquire several sets of historical data, each set of which includes execution strategies, diagnostic and treatment equipment data, and power load data, as well as corresponding load optimization decisions; the load optimization decisions include the optimized load scheduling amounts for each load loop, branch loop, and equipment terminal; integrate the execution strategies, diagnostic and treatment equipment data, and power load data, as well as the corresponding load optimization decisions, into several sets of training data and test data; The AI ​​model is trained using training data and tested using validation data. The final input consists of execution strategy, diagnostic equipment data, and power load data, and the output is a load optimization model for corresponding load optimization decisions.

8. A power distribution system for medical imaging equipment based on artificial intelligence, characterized in that, include: The module consists of a data acquisition module, a data processing module, and a load scheduling module; among which, The data acquisition module is used to obtain an initial step size and to acquire power diagnostic data based on the initial step size. The data processing module includes a data preprocessing unit, a fault diagnosis unit, and a strategy generation unit. The data preprocessing unit is used to generate several diagnostic data based on the power diagnostic data. The fault diagnosis unit is used to input the diagnostic data into the fault diagnosis model to obtain the diagnostic result. The fault diagnosis model is constructed from a CNN-SVM hybrid model and an LSTM model, and includes: An input data processing layer is constructed, comprising a diagnostic data input terminal and several branch port groups; each branch port group includes a fault detection data output terminal, a dynamic adjustment data input terminal, and a fault prediction data output terminal; the diagnostic data input terminal is used to input diagnostic data and its corresponding number. The fault detection data output terminal corresponding to the branch port group of the input data processing layer is connected to the input terminal of the CNN-SVM hybrid model, and the output terminal of the CNN-SVM hybrid model is connected to the dynamic adjustment data input terminal of the input data processing layer; the fault detection data output terminal is used to output the diagnostic sequence to the CNN-SVM hybrid model; the dynamic adjustment data input terminal is used to input the fault diagnosis result output by the CNN-SVM hybrid model. The fault prediction data output terminal of the input data processing layer is connected to the input terminal of the LSTM model; the fault prediction data output terminal is used to output the prediction sequence to the LSTM model; The input data processing layer is used to perform the following steps: S1: Obtain the number corresponding to the diagnostic data, and match the corresponding branch port group based on the number; S2: Obtain the set diagnostic sequence length and the parameter value sequence corresponding to each parameter item in the diagnostic data. Based on the diagnostic sequence length, extract the diagnostic sequence of the corresponding parameter item from the parameter value sequence, and output the diagnostic sequence corresponding to each parameter item through the fault detection data output terminal in the branch port group. S3: Obtain fault diagnosis results; generate parameter acquisition step size based on fault diagnosis results; the fault diagnosis results include the item level corresponding to several fault items and the confidence level corresponding to the item level; S4: Extract the parameter value sequence corresponding to each parameter item in the diagnostic data, fit the parameter value in the parameter value sequence into the parameter transformation curve corresponding to the parameter item, extract several parameter values ​​for prediction based on the parameter acquisition step size in the parameter transformation curve, and integrate each parameter value into the prediction sequence corresponding to the parameter item. S5: Output the prediction sequence corresponding to each parameter item through the fault prediction data output terminal in the branch port group; Thus, a fault diagnosis model is constructed; the input end of the fault diagnosis model is the diagnostic data input end corresponding to the input data processing layer, and the output end of the fault diagnosis model includes the output ends of several CNN-SVM hybrid models and the output ends of LSTM models; The diagnostic results include fault diagnosis results and fault prediction results; The strategy generation unit is used to generate an execution strategy based on the diagnostic results, acquire diagnostic equipment data and real-time power load data, and input the execution strategy, diagnostic equipment data and real-time power load data into the load optimization model to obtain a load optimization decision. The load scheduling module is used to perform load scheduling based on the load scheduling amount corresponding to each load loop, branch loop, and equipment terminal in the load optimization decision.

Images