Intelligent AI power consumption supervision system for intensive care unit

Abstract

The invention discloses a unit power utilization acquisition module which is used for sliding a current time window and acquiring a plurality of continuous power utilization parameters acquired by each bed of an intensive care unit through an intelligent electric meter in real time. The feature extraction module is used for extracting power consumption time sequence features in the current time window according to a plurality of continuous power consumption parameters obtained in real time; the AI monitoring module is used for inputting the power consumption time sequence characteristics in the current time window into the fault prediction model and outputting abnormal power consumption information; the doctor's advice interaction module is used for obtaining a doctor's advice interaction instruction according to the abnormal electricity utilization information and adjusting an electricity utilization strategy of the intensive care unit; according to the invention, by constructing the abnormal power utilization matrix, the system can simultaneously capture the relevance between the power utilization units and the dynamic change of the time sequence; the matrix-level prediction model significantly improves the perception capability of complex power consumption behaviors, and breaks through the limitation of a traditional scalar or vector method.

Description

Technical Field

[0001] The present invention relates to an electricity consumption monitoring system, and in particular to an intelligent AI electricity consumption monitoring system for an intensive care unit. Background Art

[0002] In the intensive care unit (ICU) environment, a large number of critical medical equipment are installed, such as ventilators, monitors, hemodialysis machines, and ECMO (extracorporeal membrane oxygenation) machines. These devices are highly dependent on a continuous and stable power supply. Once a power outage or failure occurs, the life support system may fail, resulting in serious consequences. For example, the shutdown of the ventilator may prevent the patient from obtaining the necessary respiratory support; the power failure of the continuous renal replacement therapy (CRRT) device may cause the liquid in the pipeline to condense; and the ECMO machine stops working, which will endanger the life safety of patients who rely on the device to maintain circulation and respiratory function.

[0003] At present, the intelligent power management systems on the market are mainly used in industrial or civil scenarios, such as factory power monitoring, school dormitory fire prevention, etc. These systems usually rely on single-point or unit-level fault monitoring technology, such as scalar or vector-based power parameter analysis methods. For example, the patent document with patent publication number CN116434939A discloses a medical equipment supervision method, equipment and storage medium based on AI algorithm, which uses a diagnostic model based on AI algorithm to obtain the operating status of non-core medical equipment, and determines the suspected abnormality of the power quality of the power supply based on the number of non-core medical equipment with abnormal operating status. The power quality is judged based on the current waveform and voltage waveform of the power supply, thereby achieving accuracy and comprehensiveness of medical equipment supervision.

[0004] Although the above-mentioned existing technologies have achieved power consumption anomaly detection, they only analyze power consumption parameters or units at a single time point, ignoring the correlation between power consumption units and dynamic changes in time series, making it difficult to capture global abnormal behaviors in complex scenarios. Most existing systems are based on fixed thresholds or simple trend analysis methods, and are unable to deeply learn the feature evolution laws before power consumption anomalies occur, resulting in alarms only when anomalies occur, lacking the ability to predict in advance. Summary of the invention

[0005] In view of the shortcomings of the prior art, the present invention provides an intelligent AI electricity consumption monitoring system for an intensive care unit, which solves the technical problems raised in the background technology through matrix-level fault prediction.

[0006] To achieve the above objectives, the present invention is implemented through the following technical solutions:

[0007] An intelligent AI power consumption monitoring system for an intensive care unit, comprising:

[0008] The unit power consumption collection module is used to slide the current time window and obtain in real time a number of continuous power consumption parameters collected by the smart meter for each bed in the intensive care unit; wherein the length of the current time window is preset to T;

[0009] A feature extraction module is used to extract the power consumption time sequence features within the current time window based on a number of continuous power consumption parameters obtained in real time;

[0010] The AI ​​monitoring module is used to input the power consumption time series characteristics in the current time window into the fault prediction model and output abnormal power consumption information; wherein the abnormal power consumption information is the abnormal power consumption prediction time point extracted from the abnormal power consumption prediction matrix, and the abnormal power consumption parameters;

[0011] The doctor's order interaction module is used to obtain doctor's order interaction instructions based on abnormal power consumption information and adjust the power consumption strategy of the intensive care unit.

[0012] In some of the embodiments, based on a number of continuous power consumption parameters acquired in real time, power consumption time sequence characteristics within the current time window are extracted, including:

[0013] S2-1, extracting three adjacent continuous power consumption parameters from a number of continuous power consumption parameters obtained in real time; S2-2, calculating the second-order difference value thereof according to the three adjacent continuous power consumption parameters;

[0014] The expression of the second-order difference is: The expression of the second-order difference value is: ;

[0015] in, Represents the second-order difference value of the continuous power consumption parameter, is the tth continuous power consumption parameter, and The power consumption parameters at the first two time points respectively;

[0016] S2-3. The second-order difference value is regarded as the power consumption time series feature at the time point corresponding to the t-th continuous power consumption parameter.

[0017] In some embodiments, the modeling step of the fault prediction model includes:

[0018] S3-1. Obtain the abnormal power consumption parameters of each bed in the intensive care unit, the power consumption time sequence characteristics of the abnormal power consumption parameters, and the normal power consumption parameters through the historical data of the smart meter;

[0019] S3-2, marking the abnormal duration point corresponding to the abnormal power consumption parameter;

[0020] S3-3, constructing an abnormal power consumption matrix according to the abnormal duration points;

[0021] S3-4, obtaining abnormal delay training samples;

[0022] S3-5. Building a fault prediction model based on the abnormal delay training samples.

[0023] In some embodiments, the abnormal power consumption parameters of each bed in the intensive care unit, the power consumption time sequence characteristics of the abnormal power consumption parameters, and the normal power consumption parameters are obtained through the historical data of the smart meter, including:

[0024] S3-1-1. For each smart meter, a historical time window of length T is intercepted on the historical time axis, and the historical power consumption parameters of any power consumption unit in the historical time window at T historical time points are obtained;

[0025] The historical power consumption parameters are characterized as: N historical power consumption parameters of any power consumption unit at a historical time point;

[0026] S3-1-2. Determine abnormal power usage parameters, abnormal power usage time sequence characteristics of the abnormal power usage parameters, and normal power usage parameters from the historical power usage parameters.

[0027] In some embodiments, determining abnormal power usage parameters, abnormal power usage time sequence characteristics of abnormal power usage parameters, and normal power usage parameters from the historical power usage parameters includes:

[0028] S3-1-2-1. Calculate the mean and standard deviation of several historical electricity consumption parameters;

[0029] The mean expression of several historical electricity consumption parameters is: ;

[0030] in, represents the mean of several historical electricity consumption parameters, represents the historical electricity consumption parameters at the tth time point;

[0031] The standard deviation expressions of several historical electricity consumption parameters are as follows: ;

[0032] in, Represents the standard deviation of several historical electricity consumption parameters;

[0033] S3-1-2-2. According to the preset mean weight and standard deviation weight, weighted sum is performed on the means and standard deviations of several historical power consumption parameters to generate historical abnormal thresholds of several historical power consumption parameters;

[0034] The expression of the historical abnormal threshold is: ;

[0035] in, Indicates the historical anomaly threshold, represents the mean weight, is the standard deviation weight;

[0036] S3-1-2-3, comparing the historical power consumption parameters at T historical time points with the historical abnormal thresholds, and screening out abnormal power consumption parameters that exceed the historical abnormal thresholds;

[0037] S3-1-2-4. Calculate the second-order difference of the abnormal power consumption parameters and mark it as the abnormal power consumption time series feature.

[0038] In some of the embodiments, marking the abnormal duration point corresponding to the abnormal power consumption parameter includes:

[0039] S3-2-1. For each abnormal power consumption parameter, locate its corresponding historical time point and mark it as the abnormal start time point;

[0040] S3-2-2, taking the abnormal start time point as the benchmark, expand forward and backward to determine whether the power consumption parameters at adjacent historical time points continuously exceed the abnormal threshold value, until normal power consumption parameters appear, and determine the abnormal end time point;

[0041] S3-2-3 marks the continuous historical time points between the abnormal start time point and the abnormal end time point as an abnormal duration segment; wherein the abnormal duration segment contains K abnormal duration points.

[0042] In some of the embodiments, constructing an abnormal power consumption matrix according to the abnormal duration point includes:

[0043] S3-3-1. Obtain the ID of the power consumption unit at each abnormal duration point to obtain M power consumption unit IDs;

[0044] S3-3-2. At each abnormal duration point, any power consumption unit ID, its corresponding abnormal power consumption parameters and abnormal power consumption time sequence characteristics are standardized and spliced ​​into the abnormal event vector of the power consumption unit;

[0045] S3-3-3. Obtain the abnormal event vectors of M power consumption units, and construct them into an abnormal event sub-matrix for each abnormal duration point; wherein the row vector of the abnormal event sub-matrix is ​​the abnormal event vector, and the column vector is the same type of feature;

[0046] S3-3-4. Concatenate the abnormal event sub-matrices of K abnormal duration points to generate the abnormal power consumption matrix of the intensive care unit in time series.

[0047] In some of the embodiments, obtaining abnormal delay training samples includes:

[0048] S3-4-1. Slide the historical time window to between the last abnormal end time point and the next abnormal start time point, mark K abnormal start time points in the historical time window, and obtain normal power consumption parameters at the abnormal start time points;

[0049] S3-4-2. Construct a normal power consumption matrix of normal power consumption parameters at the time point before the abnormality begins;

[0050] S3-4-3. Define the normal power consumption matrix as input features, define the abnormal power consumption matrix as target labels, and construct abnormal delay training samples.

[0051] In some embodiments, building a fault prediction model based on the abnormal delay training sample includes:

[0052] S3-4-1. Input the abnormal delay training samples into the supervised model;

[0053] S3-4-2, forward propagating the normal power consumption matrix to generate an abnormal power consumption prediction matrix;

[0054] S3-4-3, calculating the matrix loss between the abnormal power consumption prediction matrix and the abnormal power consumption matrix, and obtaining the fault prediction model after minimizing the matrix loss function through back propagation;

[0055] Among them, the matrix loss function is:

[0056] ;

[0057] ;

[0058] ;

[0059] in, represents the abnormal power consumption matrix, represents the abnormal power consumption prediction matrix, K represents the number of abnormal duration points, M represents the number of power consumption units, and R represents the real number domain; represents the matrix element in the i-th row and j-th column of the abnormal power consumption matrix, Represents the matrix element in the i-th row and j-th column in the abnormal power consumption prediction matrix.

[0060] In some embodiments, adjusting the power usage strategy of the intensive care unit according to the abnormal power usage information includes:

[0061] S4-1. Pack the abnormal electricity usage information into a standardized format and send it to the designated medical staff;

[0062] S4-2, medical staff issue medical instructions based on abnormal electricity usage information;

[0063] S4-3. According to the doctor's order interaction instruction, control the power usage strategy of the power unit in the bed.

[0064] The present invention provides an intelligent AI power consumption monitoring system for an intensive care unit, which has the following beneficial effects:

[0065] By constructing an abnormal power consumption matrix, the system can simultaneously capture the correlation between power consumption units (column vectors) and the dynamic changes of time series (row vectors). The matrix-level prediction model significantly improves the perception of complex power consumption behaviors and breaks through the limitations of traditional scalar or vector methods.

[0066] In the model training stage, the present invention introduces a matrix-based loss function to directly calculate the global error between the abnormal power consumption matrix and the prediction matrix. Compared with the traditional scalar or vector loss function, the matrix loss function can simultaneously optimize the deviation of the time dimension and the unit dimension, so that the model has stronger global prediction ability.

[0067] The present invention extracts the normal power consumption parameters (input features) and abnormal power consumption matrix (target label) before the abnormality begins through a sliding time window, and generates abnormal delay training samples. This allows the model to learn the feature evolution law before the abnormality occurs in advance, and achieve accurate prediction of future abnormalities. BRIEF DESCRIPTION OF THE DRAWINGS

[0068] Figure 1 This is a structural block diagram of an intelligent AI power consumption monitoring system for an intensive care unit of the present invention;

[0069] Figure 2 This is a work flow chart of an intelligent AI power consumption monitoring system for an intensive care unit of the present invention;

[0070] Figure 3 The flowchart of the extraction of abnormal power consumption timing characteristics of the present invention;

[0071] Figure 4 This is a flow chart for marking an abnormal duration period of the present invention. DETAILED DESCRIPTION

[0072] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0073] Example 1: Please refer to Figure 1 and Figure 2 The present invention provides an intelligent AI power consumption monitoring system for an intensive care unit, comprising:

[0074] The unit electricity consumption collection module is used to slide the current time window and obtain in real time a number of continuous electricity consumption parameters collected by the smart meter for each bed in the intensive care unit; wherein the length of the current time window is preset to T.

[0075] A feature extraction module is used to extract the power consumption time sequence features within the current time window based on a number of continuous power consumption parameters obtained in real time;

[0076] The extraction step of the feature extraction module includes:

[0077] S2-1. Extract three adjacent continuous power usage parameters from a number of continuous power usage parameters acquired in real time; ensure the continuity of the continuous power usage parameters in time sequence, wherein the continuous power usage parameters include current, voltage, power factor, etc.

[0078] S2-2, according to the three adjacent continuous power consumption parameters, calculate the second-order difference value thereof to capture the change rate and trend of the continuous power consumption parameters;

[0079] The expression of the second-order difference is: The expression of the second-order difference value is: ;

[0080] in, Represents the second-order difference value of the continuous power consumption parameter, is the tth continuous power consumption parameter, and The power consumption parameters at the first two time points respectively; by calculating the second-order difference value, the drastic fluctuation, mutation or trend change of continuous power consumption parameters can be effectively identified, and the change rate and acceleration characteristics of abnormal power consumption can be captured.

[0081] S2-3. The second-order difference value is regarded as the power consumption time series feature at the time point corresponding to the t-th continuous power consumption parameter.

[0082] The AI ​​monitoring module is used to input the power consumption timing characteristics within the current time window into the fault prediction model and output abnormal power consumption information; wherein the abnormal power consumption information is the abnormal power consumption prediction time point and abnormal power consumption parameters extracted from the abnormal power consumption prediction matrix.

[0083] The doctor's order interaction module is used to obtain doctor's order interaction instructions based on abnormal power consumption information and adjust the power consumption strategy of the intensive care unit.

[0084] Specifically, this embodiment sets up independent smart meters on each bed to monitor the continuous power consumption parameters such as current, voltage and power factor of each bed in the intensive care unit in real time and continuously within a preset time window length T. Through the sliding time window mechanism, the continuous power consumption parameters within the current time window can be continuously updated and obtained to ensure the real-time and continuity of the data.

[0085] On this basis, based on the real-time monitoring of continuous power consumption parameters, the system analyzes the power consumption time series characteristics of each bed in the current time window through the fault prediction model, and predicts abnormal power consumption information at future time points, including abnormal power consumption prediction time points and abnormal power consumption parameters. The abnormal power consumption parameters include but are not limited to:

[0086] Abnormal current: The monitored current value exceeds the set abnormal threshold range;

[0087] Abnormal voltage: Voltage fluctuations beyond the normal operating range may cause unstable power supply to the equipment;

[0088] Abnormal load: The load parameters calculated based on current and voltage are abnormal, such as overload or mutation.

[0089] Example 2: See Figures 1 to 4 The technical solution of this embodiment 2 is different from that of embodiment 1 in that the modeling steps of the fault prediction model are disclosed, and the modeling steps include:

[0090] S3-1. Obtain the abnormal power consumption parameters of each bed in the intensive care unit, the power consumption time sequence characteristics of the abnormal power consumption parameters, and the normal power consumption parameters through the historical data of the smart meter;

[0091] Abnormal power consumption parameters: Power consumption parameters that exceed the abnormal statistical threshold represent abnormal performance of power consumption status;

[0092] Electricity consumption time series characteristics: Time series characteristics extracted by calculating second-order difference values ​​and other methods, reflecting the fluctuation rate and trend of parameters;

[0093] Normal power consumption parameters: Parameters that do not exceed the abnormal threshold and serve as the normal state benchmark for model training.

[0094] S3-2, marking the abnormal duration point corresponding to the abnormal power consumption parameter;

[0095] S3-3, constructing an abnormal power consumption matrix according to the abnormal duration points;

[0096] S3-4, obtaining abnormal delay training samples;

[0097] S3-4. Building a fault prediction model based on the abnormal delay training samples.

[0098] Specifically, in this embodiment, based on the historical data collected by the smart meters, combined with abnormal power consumption parameters, power consumption time series characteristics, and normal power consumption parameters, the modeling of the fault prediction model is completed. By processing the historical data, the abnormal duration points are marked, and an abnormal power consumption matrix is constructed based on the abnormal duration points, and further abnormal delay training samples are generated. Finally, the fault prediction model is trained using the abnormal delay training samples.

[0099] In summary, in this embodiment, by extracting the power consumption time series characteristics of abnormal power consumption parameters and constructing abnormal delay training samples, the changing rules of power consumption anomalies can be effectively captured, and an accurate fault prediction model can be trained. This model can realize the abnormal warning ability for the future time points of the power consumption system in the intensive care unit, provide an important basis for the early intervention of power anomalies, thereby effectively reducing the downtime risk of key equipment, and ensuring the stable operation of medical equipment in the intensive care unit and the safety of patients' lives.

[0100] Further, the step S3-1 specifically includes:

[0101] S3-1-1. For each smart meter, intercept a historical time window with a length of T on the historical time axis, and obtain the historical power consumption parameters of any power consumption unit at T historical time points within the historical time window;

[0102] Among them, the historical power consumption parameters are characterized as: N historical power consumption parameters of any power consumption unit at a historical time point; the historical power consumption parameters include: current, voltage, and power factor;

[0103] S3-1-2. Determine the abnormal power consumption parameters, the abnormal power consumption time series characteristics of the abnormal power consumption parameters, and the normal power consumption parameters from the historical power consumption parameters.

[0104] For each smart meter, data is collected on the historical time axis according to a time window with a preset length T. For each time point within the historical time window, historical power consumption parameters such as the current, voltage, and power factor of any power consumption unit are collected to ensure that the data coverage range is complete and has continuity in the time dimension. These historical power consumption parameters can comprehensively reflect the operating state of the power consumption unit at different times.

[0105] Even further, the step S3-1-2 specifically includes:

[0106] S3-1-2-1. Calculate the mean and standard deviation of several historical power consumption parameters;

[0107] The mean expression of several historical power consumption parameters is: ;

[0108] Among them, represents the mean of several historical power consumption parameters, represents the historical electricity consumption parameters at the tth time point;

[0109] The standard deviation expressions of several historical electricity consumption parameters are as follows: ;

[0110] in, Represents the standard deviation of several historical electricity consumption parameters;

[0111] S3-1-2-2. According to the preset mean weight and standard deviation weight, weighted sum is performed on the means and standard deviations of several historical power consumption parameters to generate historical abnormal thresholds of several historical power consumption parameters;

[0112] The expression of the historical abnormal threshold is: ;

[0113] in, Indicates the historical anomaly threshold, represents the mean weight, is the standard deviation weight;

[0114] Specifically, the weighted summation can be performed using the triple standard deviation method, that is, The value is 1. A value of 3 can detect outliers in the power consumption parameters.

[0115] S3-1-2-3, comparing the historical power consumption parameters at T historical time points with the historical abnormal thresholds, and screening out abnormal power consumption parameters that exceed the historical abnormal thresholds;

[0116] S3-1-2-4. Calculate the second-order difference of abnormal power consumption parameters, capture the changing rate and trend of the parameters, and mark them as abnormal power consumption time series characteristics.

[0117] This embodiment generates historical abnormal thresholds by calculating the mean and standard deviation of historical power consumption parameters and weighted summation based on weights, thereby screening out abnormal power consumption parameters. At the same time, the time series characteristics of the abnormal parameters are extracted in combination with the second-order difference values ​​to capture the dynamic change trend of the parameters. Through the above processing method, this embodiment can accurately identify abnormal power consumption behaviors within the historical time period and mark their corresponding change characteristics, providing sufficient data support for the subsequent abnormal duration point marking and abnormal power consumption matrix construction.

[0118] Furthermore, the step S3-2 specifically includes:

[0119] S3-2-1. For each abnormal power consumption parameter, locate its corresponding historical time point and mark it as the abnormal start time point;

[0120] S3-2-2, taking the abnormal start time point as the benchmark, expand forward and backward to determine whether the power consumption parameters at adjacent historical time points continuously exceed the abnormal threshold value, until normal power consumption parameters appear, and determine the abnormal end time point;

[0121] S3-2-3 marks the continuous historical time points between the abnormal start time point and the abnormal end time point as an abnormal duration segment; wherein the abnormal duration segment includes K abnormal duration points.

[0122] Through this embodiment, the accurate marking and demarcation of the time period corresponding to the abnormal power consumption parameters is successfully achieved. By locating the abnormal start time point, determining the abnormal end time point, and marking the abnormal duration period, the time range of the abnormal power consumption parameters and their dynamic evolution process can be fully captured.

[0123] The marking of abnormal duration periods lays a solid foundation for the construction of abnormal power consumption matrix and the training of fault prediction model. At the same time, this method can effectively distinguish the continuous and intermittent characteristics of power consumption anomalies, providing a reliable basis for formulating more accurate power management strategies and abnormal handling solutions, thus significantly improving the safety and stability of the power system operation in the intensive care unit.

[0124] Furthermore, the step S3-3 specifically includes:

[0125] S3-3-1. Obtain the ID of the power consumption unit at each abnormal duration point to obtain M power consumption unit IDs;

[0126] S3-3-2. At each abnormal duration point, any power consumption unit ID, its corresponding abnormal power consumption parameters and abnormal power consumption time sequence characteristics are standardized and spliced ​​into the abnormal event vector of the power consumption unit;

[0127] S3-3-3. Obtain the abnormal event vectors of M power consumption units, and construct them into an abnormal event sub-matrix for each abnormal duration point; wherein the row vector of the abnormal event sub-matrix is ​​the abnormal event vector, and the column vector is the same type of feature;

[0128] S3-3-4. Concatenate the abnormal event sub-matrices of K abnormal duration points to generate the abnormal power consumption matrix of the intensive care unit in time series.

[0129] Through this embodiment, an abnormal power consumption matrix of the power consumption system in the intensive care unit is successfully constructed. The abnormal power consumption matrix not only records the abnormal state of each power consumption unit in the abnormal duration period in a time series form, but also comprehensively reflects the spatial distribution and dynamic change trend of power consumption anomalies by integrating abnormal power consumption parameters, power consumption time series characteristics and related unit IDs.

[0130] The construction of this matrix provides standardized and high-quality training samples for subsequent fault prediction models, significantly improving the accuracy and efficiency of anomaly detection and early warning. At the same time, the structural characteristics of the abnormal power consumption matrix enable the abnormal conditions in the intensive care unit to be clearly displayed.

[0131] Furthermore, the step S3-4 specifically includes:

[0132] S3-4-1. Slide the historical time window to between the last abnormal end time point and the next abnormal start time point, mark K abnormal start time points in the historical time window, and obtain normal power consumption parameters at the abnormal start time points;

[0133] S3-4-2. Construct a normal power consumption matrix of normal power consumption parameters at the time point before the abnormality begins;

[0134] S3-4-3. Define the normal power consumption matrix as input features, define the abnormal power consumption matrix as target labels, and construct abnormal delay training samples.

[0135] Through this embodiment, the accurate construction of abnormal delay training samples is achieved. By sliding the historical time window, the normal power consumption parameters before the abnormality begins are extracted, the normal power consumption matrix is ​​constructed, and the abnormal power consumption matrix is ​​combined as the target label to completely generate the training samples.

[0136] This sample construction method captures the dynamic relationship between normal and abnormal power consumption, providing high-quality data input for the fault prediction model. The application of abnormal delay training samples can help the model learn the feature evolution law before abnormal power consumption occurs, thereby improving the ability to predict future abnormal conditions.

[0137] Furthermore, the step S3-5 specifically includes:

[0138] S3-5-1, input the abnormal delay training sample into the supervised model;

[0139] S3-5-2, forward propagating the normal power consumption matrix to generate an abnormal power consumption prediction matrix;

[0140] S3-5-3. Calculate the matrix loss between the abnormal power consumption prediction matrix and the abnormal power consumption matrix, and obtain the fault prediction model after minimizing the matrix loss function through back propagation.

[0141] Among them, the matrix loss function is:

[0142] ;

[0143] ;

[0144] ;

[0145] in, represents the abnormal power consumption matrix, represents the abnormal power consumption prediction matrix, K represents the number of abnormal duration points, M represents the number of power consumption units, and R represents the real number domain; represents the matrix element in the i-th row and j-th column of the abnormal power consumption matrix, Represents the matrix element in the i-th row and j-th column in the abnormal power consumption prediction matrix;

[0146] Specifically, each row of the matrix corresponds to an abnormal duration point, and the loss calculation includes time series characteristics; each column of the matrix corresponds to a different power consumption unit, and the loss calculation comprehensively considers the correlation between multiple units; calculating the loss at the two-dimensional matrix level can simultaneously capture the comprehensive deviations of the time dimension (row) and the power consumption unit dimension (column); compared with traditional scalar or vector-level loss functions, the matrix-level loss function can better reflect the global performance of the entire system.

[0147] Furthermore, the steps of extracting abnormal power consumption information from the abnormal power consumption prediction matrix formed by the fault prediction model are:

[0148] Define the global abnormal threshold of the abnormal power consumption prediction matrix ;

[0149] Traversing the abnormal power consumption prediction matrix Each element of , to determine whether it exceeds the corresponding global abnormal threshold ;

[0150] like Greater than , then it is considered that the jth power consumption unit at the i-th time point has an abnormality;

[0151] Record the corresponding abnormal prediction time point, power consumption unit, and abnormal power consumption parameter value;

[0152] All abnormal points exceeding the threshold are classified according to time points and power consumption units to generate the abnormal power consumption information.

[0153] Embodiment 3: The technical solution of Embodiment 3 is different from that of Embodiments 1 and 2 in that the embodiment 3 discloses the steps for adjusting the power usage strategy of the intensive care unit according to the abnormal power usage information, including:

[0154] S4-1. Pack the abnormal electricity usage information into a standardized format and send it to the designated medical staff;

[0155] Among them, the abnormal power consumption information packaging refers to organizing and packaging the abnormal power consumption information predicted by the model in a standardized format so as to clearly transmit it to relevant medical staff or power management systems. For example, the abnormal power consumption information includes:

[0156] Abnormal power consumption prediction time point: predict the time point when abnormalities may occur in the future, such as predicting that a certain bed may experience abnormal voltage within the next preset time;

[0157] Abnormal power parameters: specific power parameters involving abnormalities, such as current out of range, excessive voltage fluctuations, abnormal power factor, etc.

[0158] S4-2, medical staff issue medical instructions based on abnormal electricity usage information;

[0159] Medical instructions are instructions issued by medical staff to operate bed power units or medical equipment based on abnormal power consumption information and clinical needs, including:

[0160] Start or stop command: for the switch status of a specific device, such as starting the backup power supply and shutting down the short-circuit device;

[0161] Adjustment of power usage priority: In abnormal situations, priority will be given to ensuring power usage for key equipment (such as ventilators and ECMO);

[0162] Equipment inspection instructions: remind power technicians to check the connection status of specific equipment, circuit operation or equipment power interface.

[0163] Medical instructions can combine the results of power anomaly predictions with actual medical needs to ensure that the adjustment of power usage strategies meets both the safety requirements of equipment operation and the treatment needs of patients.

[0164] S4-3. According to the doctor's order interaction instruction, control the power usage strategy of the power unit in the bed.

[0165] The system matches and executes the preset power usage strategy based on the acquired abnormal power usage information, combined with the type and specific value of the abnormal power usage parameters. The power usage strategy is formulated in advance according to the actual scenario requirements and stored in the system, specifically including:

[0166] Power switching strategy: In case of abnormal current or short circuit in a bed unit, only the power supply of the corresponding unit is cut off to avoid the spread of fault;

[0167] Backup power supply activation strategy: When voltage anomalies or unstable power supply are detected, the system automatically switches to backup power supply to provide continuous power support for key equipment.

[0168] Load adjustment strategy: In response to abnormal load conditions, the power distribution is dynamically adjusted to reduce the power supply load and ensure the normal operation of the equipment;

[0169] Doctor's advice notification strategy: For sudden abnormal power consumption, abnormal power consumption information is immediately generated and sent to the designated medical staff terminal to remind them to check the equipment operation status and take necessary measures in time.

[0170] Through the above method, this embodiment can realize real-time monitoring and prediction of the power status of equipment in the intensive care unit, and quickly formulate and implement corresponding power usage strategies based on abnormal power usage information, thereby effectively ensuring the stable operation of key medical equipment in the intensive care unit and the life safety of patients.

[0171] The above embodiments may be implemented in whole or in part by software, hardware, firmware or any other combination thereof. When implemented using software, the above embodiments may be implemented in whole or in part in the form of a computer program product. The computer program product includes one or more computer instructions or computer programs. When the computer instructions or computer programs are loaded or executed on a computer, the processes or functions described in the embodiments of the present application are generated in whole or in part. The computer may be a general-purpose computer, a special-purpose computer, a computer network, or other programmable device. The computer instructions may be stored in a computer-readable storage medium, or transmitted from one computer-readable storage medium to another computer-readable storage medium, for example, the computer instructions may be transmitted from one website site, computer, server or data center to another website site, computer, server or data center by wired (e.g., infrared, wireless, microwave, etc.) means.

[0172] The computer-readable storage medium may be any available medium that can be accessed by a computer or a data storage device such as a server or a data center that includes one or more available media. The available medium may be a magnetic medium (e.g., a floppy disk, a hard disk, a tape), an optical medium (e.g., a DVD), or a semiconductor medium. The semiconductor medium may be a solid-state hard disk.

[0173] In the several embodiments provided in the present application, it should be understood that the disclosed systems, devices and methods can be implemented in other ways. For example, the device embodiments described above are only schematic. For example, the division of the units is only a logical function division of a waterway underwater terrain change analysis system and method. There may be other division methods in actual implementation. For example, multiple units or components can be combined or integrated into another system, or some features can be ignored or not executed. Another point is that the mutual coupling or direct coupling or communication connection shown or discussed can be through some interfaces, indirect coupling or communication connection of devices or units, which can be electrical, mechanical or other forms.

[0174] The above description is only a specific implementation manner of the present application, but the protection scope of the present application is not limited thereto. Any technician familiar with the technical field can easily think of changes or substitutions within the technical scope disclosed in the present application, which should be included in the protection scope of the present application.

Claims

1. An intelligent AI power consumption monitoring system for intensive care units, characterized in that: include: The unit power consumption collection module is used to slide the current time window and obtain in real time a number of continuous power consumption parameters collected by the smart meter for each bed in the intensive care unit; wherein the length of the current time window is preset to T; A feature extraction module is used to extract the power consumption time sequence features within the current time window based on a number of continuous power consumption parameters obtained in real time; The AI ​​monitoring module is used to input the power consumption time series characteristics in the current time window into the fault prediction model and output abnormal power consumption information; wherein the abnormal power consumption information is the abnormal power consumption prediction time point extracted from the abnormal power consumption prediction matrix, and the abnormal power consumption parameters; The doctor's order interaction module is used to obtain doctor's order interaction instructions based on abnormal power consumption information and adjust the power consumption strategy of the intensive care unit.

2. According to claim 1, an intelligent AI power consumption monitoring system for an intensive care unit is characterized in that: Based on several continuous power consumption parameters obtained in real time, the power consumption time series characteristics in the current time window are extracted, including: S2-1, extracting three adjacent continuous power consumption parameters from a number of continuous power consumption parameters obtained in real time; S2-2, calculating the second-order difference value of three adjacent continuous power consumption parameters; The expression of the second-order difference is: The expression of the second-order difference value is: ; in, Represents the second-order difference value of the continuous power consumption parameter, is the tth continuous power consumption parameter, and The power consumption parameters at the first two time points respectively; S2-3. The second-order difference value is regarded as the power consumption time series feature at the time point corresponding to the t-th continuous power consumption parameter.

3. According to claim 1, the intelligent AI power consumption monitoring system for intensive care units is characterized in that: The modeling steps of the fault prediction model include: S3-1. Obtain the abnormal power consumption parameters of each bed in the intensive care unit, the power consumption time sequence characteristics of the abnormal power consumption parameters, and the normal power consumption parameters through the historical data of the smart meter; S3-2, marking the abnormal duration point corresponding to the abnormal power consumption parameter; S3-3, constructing an abnormal power consumption matrix according to the abnormal duration points; S3-4, obtaining abnormal delay training samples; S3-5. Building a fault prediction model based on the abnormal delay training samples.

4. According to claim 3, an intelligent AI power consumption monitoring system for an intensive care unit is characterized in that: Through the historical data of smart meters, the abnormal power consumption parameters of each bed in the ICU, the power consumption time series characteristics of abnormal power consumption parameters, and normal power consumption parameters are obtained, including: S3-1-1. For each smart meter, a historical time window of length T is intercepted on the historical time axis, and the historical power consumption parameters of any power consumption unit in the historical time window at T historical time points are obtained; The historical power consumption parameters are characterized as: N historical power consumption parameters of any power consumption unit at a historical time point; S3-1-2. Determine abnormal power usage parameters, abnormal power usage time sequence characteristics of the abnormal power usage parameters, and normal power usage parameters from the historical power usage parameters.

5. The intelligent AI power consumption monitoring system for an intensive care unit according to claim 4 is characterized in that: Determining abnormal power usage parameters, abnormal power usage time sequence characteristics of the abnormal power usage parameters, and normal power usage parameters from the historical power usage parameters includes: S3-1-2-1. Calculate the mean and standard deviation of several historical electricity consumption parameters; The mean expression of several historical electricity consumption parameters is: ; in, represents the mean of several historical electricity consumption parameters, represents the historical electricity consumption parameters at the tth time point; The standard deviation expressions of several historical electricity consumption parameters are as follows: ; in, Represents the standard deviation of several historical electricity consumption parameters; S3-1-2-2. According to the preset mean weight and standard deviation weight, weighted sum is performed on the means and standard deviations of several historical power consumption parameters to generate historical abnormal thresholds of several historical power consumption parameters; The expression of the historical abnormal threshold is: ; in, Indicates the historical anomaly threshold, represents the mean weight, is the standard deviation weight; S3-1-2-3, comparing the historical power consumption parameters at T historical time points with the historical abnormal thresholds, and screening out abnormal power consumption parameters that exceed the historical abnormal thresholds; S3-1-2-4. Calculate the second-order difference of the abnormal power consumption parameters and mark it as the abnormal power consumption time series feature.

6. The intelligent AI power consumption monitoring system for an intensive care unit according to claim 5 is characterized in that: Mark the abnormal duration points corresponding to the abnormal power consumption parameters, including: S3-2-1. For each abnormal power consumption parameter, locate its corresponding historical time point and mark it as the abnormal start time point; S3-2-2, taking the abnormal start time point as the benchmark, expand forward and backward to determine whether the power consumption parameters at adjacent historical time points continuously exceed the abnormal threshold value, until normal power consumption parameters appear, and determine the abnormal end time point; S3-2-3 marks the continuous historical time points between the abnormal start time point and the abnormal end time point as an abnormal duration segment; wherein the abnormal duration segment contains K abnormal duration points.

7. The intelligent AI power consumption monitoring system for an intensive care unit according to claim 6 is characterized in that: According to the abnormal duration point, an abnormal power consumption matrix is ​​constructed, including: S3-3-1. Obtain the ID of the power consumption unit at each abnormal duration point to obtain M power consumption unit IDs; S3-3-2. At each abnormal duration point, any power consumption unit ID, its corresponding abnormal power consumption parameters and abnormal power consumption time sequence characteristics are standardized and spliced ​​into the abnormal event vector of the power consumption unit; S3-3-3. Obtain the abnormal event vectors of M power consumption units, and construct them into an abnormal event sub-matrix for each abnormal duration point; wherein the row vector of the abnormal event sub-matrix is ​​the abnormal event vector, and the column vector is the same type of feature; S3-3-4. Concatenate the abnormal event sub-matrices of K abnormal duration points to generate the abnormal power consumption matrix of the intensive care unit in time series.

8. The intelligent AI power consumption monitoring system for an intensive care unit according to claim 7 is characterized in that: Obtain abnormal delay training samples, including: S3-4-1. Slide the historical time window to between the last abnormal end time point and the next abnormal start time point, mark K abnormal start time points in the historical time window, and obtain normal power consumption parameters at the abnormal start time points; S3-4-2. Construct a normal power consumption matrix of normal power consumption parameters at the time point before the abnormality begins; S3-4-3. Define the normal power consumption matrix as input features, define the abnormal power consumption matrix as target labels, and construct abnormal delay training samples.

9. The intelligent AI power consumption monitoring system for an intensive care unit according to claim 8, characterized in that: According to the abnormal delay training sample, a fault prediction model is built, including: S3-4-1. Input the abnormal delay training samples into the supervised model; S3-4-2, forward propagating the normal power consumption matrix to generate an abnormal power consumption prediction matrix; S3-4-3, calculating the matrix loss between the abnormal power consumption prediction matrix and the abnormal power consumption matrix, and obtaining the fault prediction model after minimizing the matrix loss function through back propagation; Among them, the matrix loss function is: ； ； ; in, represents the abnormal power consumption matrix, represents the abnormal power consumption prediction matrix, K represents the number of abnormal duration points, M represents the number of power consumption units, and R represents the real number domain; represents the matrix element in the i-th row and j-th column of the abnormal power consumption matrix, Represents the matrix element in the i-th row and j-th column in the abnormal power consumption prediction matrix.

10. The intelligent AI power consumption monitoring system for an intensive care unit according to claim 9, characterized in that: According to the abnormal power consumption information, adjust the power consumption strategy of the intensive care unit, including: S4-1. Pack the abnormal electricity usage information into a standardized format and send it to the designated medical staff; S4-2, medical staff issue medical instructions based on abnormal electricity usage information; S4-3. According to the doctor's order interaction instruction, control the power usage strategy of the power unit in the bed.

Images