Operating room purification state prediction and early warning system based on multi-source data fusion

Abstract

The invention discloses an operating room purification state prediction and early warning system based on multi-source data fusion, and belongs to the technical field of medical environment control. An operating room purification state prediction and early warning system based on multi-source data fusion comprises a data acquisition module, a data processing module, a hybrid prediction module and an early warning control module. The operating room environment control system solves the problems that an existing operating room environment control system lags in response to sudden and non-linear heat and humidity disturbance and is difficult to maintain stability of environment parameters, the hybrid prediction model is constructed by fusing multi-source disturbance data and time sequence environment data, dynamic prospective prediction of temperature and humidity of a core area can be achieved, and the operating room environment control system has a good application prospect. The air conditioning system is driven to carry out self-adaptive prospective regulation and control and initiative maintenance based on the predicted value, and meanwhile, accurate information prediction support is provided for preventive maintenance of key equipment, so that an environment control mode is converted from passive response to combination of active prevention and active maintenance; and the stability, the safety and the control efficiency of the operating room environment are obviously improved.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of medical environment control, in particular to a surgical room purification state prediction and early warning system based on multi-source data fusion. BACKGROUND

[0002] The surgical room is the core place for high-precision and high-risk surgery in the hospital, and the cleanliness, temperature and humidity and other parameters of the internal environment must be strictly controlled within a certain range to ensure the safety of the operation, reduce the risk of postoperative infection and protect the comfort of medical staff. The traditional surgical room environment control mainly relies on the real-time feedback of the limited sensors deployed in the fixed area according to the air conditioning purification system (HVAC) to respond to the control. However, the surgical room environment is a complex system dynamically affected by many factors. In addition to the running state of the air conditioning unit itself, the intensity of the operation schedule, the frequent entry and exit of medical staff and patients, and the start and stop and operation of various medical equipment (such as anesthetic machines, extracorporeal circulation machines, and high-frequency electric knives) will introduce significant thermal and humidity disturbances and air flow field changes. These disturbances have the characteristics of suddenness, intermittence and nonlinearity. The existing feedback control system based on real-time monitoring has obvious hysteresis, and often starts adjustment after the environmental parameters have deviated, which is difficult to achieve precise forward-looking control to maintain the stability of the environmental parameters; therefore, it cannot meet the existing needs. For this reason, we propose a surgical room purification state prediction and early warning system based on multi-source data fusion. SUMMARY

[0003] The purpose of the present application is to provide a surgical room purification state prediction and early warning system based on multi-source data fusion, which fuses real-time monitoring data and dynamic disturbance data of the surgical room environment, uses a hybrid prediction model to capture the time sequence rule of the environmental parameters and the nonlinear influence of the disturbance events, and thus realizes precise dynamic prediction of the future environmental state of the core area. Based on the prediction results, the system can issue a graded early warning before the environmental parameters actually deviate from the threshold, and send control instructions to the air conditioning control system in advance to make it actively adjust the operating parameters to offset the predicted disturbance, thereby solving the problems raised in the background art.

[0004] To achieve the above purpose, the present application provides the following technical scheme: a surgical room purification state prediction and early warning system based on multi-source data fusion, comprising: A data acquisition module configured to acquire real-time surgical room environment data and environmental disturbance data, wherein the environmental data at least includes time sequence pressure and temperature and humidity data acquired by pressure and temperature and humidity sensors deployed in air conditioning units, refrigerated water inlet and outlet pipes, air supply pipes, return air pipes and multiple positions inside the surgical room, and the environmental disturbance data at least includes operation scheduling information, personnel entry and exit logs and medical equipment start and stop states; a data processing module configured to timestamp align, missing value process and feature extract the environment data and the environment disturbance data to generate a fusion feature dataset; a hybrid prediction module including a coupled time series analysis module and a machine learning module, configured to receive the fusion feature dataset and output dynamic prediction values of the pressure and the temperature and humidity in the core area of the operating room for a future set period, wherein the time series analysis module is used to capture periodicity and trend of the pressure and the temperature and humidity data, and the machine learning module is used to learn a nonlinear mapping relationship between the environment disturbance data and the dynamic changes of the pressure and the temperature and humidity; a pre-warning control module configured to compare the dynamic prediction values with preset threshold values of a purified environment, and generate a graded pre-warning signal before the prediction values deviate from the threshold values, and feed back the dynamic prediction values to an operating room air conditioning automatic control system for adaptive regulation and control of the operating room air conditioning automatic control system.

[0005] Further, the data acquisition module includes: an environment sensing module configured to acquire operating room environment data in real time through distributed intelligent sensors, the environment data including time series pressure and temperature and humidity data acquired by pressure and temperature and humidity sensors of air conditioning units, incoming and outgoing refrigerated water pipes, air supply pipes, return air pipes and multiple positions inside the operating room; a disturbance perception module configured to acquire environment disturbance data in real time, the environment disturbance data including at least operating scheduling information, personnel access logs and medical equipment start-stop states; a quality verification module configured to receive raw data of the environment sensing module and the disturbance perception module, to perform physical rationality verification and cross-sensor consistency verification on the sensor data, to automatically generate device calibration and fault check prompts when data anomalies are found, and to perform interpolation repair on short-time abnormal data based on historical normal data.

[0006] Further, the data acquisition module adopts an adaptive acquisition method, specifically including: receiving operating scheduling information, the operating scheduling information including at least operating room number, planned start time, expected duration and operating type; determining a current operating phase of a target operating room based on the operating scheduling information and current time, the operating phase including preoperative preparation period, operating period or postoperative cleaning period; generating and issuing data acquisition strategy instructions dynamically according to the determined current operating phase and the operating type, the data acquisition strategy instructions including at least: adjusting sampling frequency of at least part of pressure and temperature and humidity sensors of the environment sensing module deployed in the core area of the operating room; and adjusting monitoring parameters of components in the disturbance perception module for monitoring personnel access and for monitoring equipment start-stop, respectively, the adjustments of the personnel access components including monitoring sensitivity and reporting threshold, and the adjustments of the equipment start-stop components including state determination conditions and response delay; wherein the adjustment strategy follows: when determining as the surgery operation period, the core area pressure and temperature and humidity sensors adopt a first sampling frequency for data acquisition, when determining as the postoperative cleaning period, the same batch of sensors adopt a second sampling frequency for data acquisition, and the first sampling frequency is higher than the second sampling frequency.

[0007] Further, the data processing module comprises: a synchronization alignment module configured to receive asynchronous data streams from the data acquisition module, perform clock correction and sampling point alignment on the environment data and the environment disturbance data based on a unified time reference, and generate a multi-source data sequence with a unified time axis; a missing value processing module configured to identify and distinguish missing values in the multi-source data sequence, including for random missing values caused by intermittent failure of conventional sensors, using linear interpolation based on adjacent valid data for filling, and for long-term continuous missing values caused by system maintenance or equipment replacement, using the most relevant historical same-period data of the same type for pattern filling and adding a data missing label; a feature construction module configured to analyze the environment disturbance data, and based on a predefined rule base, mark the start and end timestamps and event types of disturbance events on the unified time axis; integrate the time axis marked with disturbance events with the processed multi-source data sequence, and extract key features for constructing the fusion feature data set, wherein the key features include pressure and temperature and humidity statistical features in the current and historical windows, encoding features of the current disturbance event type, and time intervals from the start of the last same type of disturbance event.

[0008] Further, the hybrid prediction module further comprises: a feature allocation module configured to divide the fusion feature data set into a first data subset and a second data subset according to the types and sources of data features, and perform intelligent distribution; wherein the time series analysis module receives the first data subset with continuous time series characteristics allocated by the feature allocation module, and the first data subset at least contains historical and current pressure and temperature and humidity time series data; the machine learning module receives the second data subset representing discrete events allocated by the feature allocation module, and the second data subset at least contains surgery scheduling, personnel access event and equipment state change data; The weight allocation module is configured to dynamically determine the fusion weight of the first prediction sub-result and the second prediction sub-result based on the activity analysis of current and historical environmental disturbance data. Among them, the fusion weight is adjusted in reverse correlation based on the level of business disturbance in which the operating room is located.

[0009] Furthermore, the weight allocation module adjusts the fusion weights according to the level of service disturbance, specifically as follows: When environmental disturbance data analysis indicates that the operating room is at a high disturbance level, the weight allocation module increases the fusion weight of the second prediction sub-result. The high disturbance level is triggered by the following conditions: In the future, the system will be designed to detect intensive surgical schedules within a set time period, monitor personnel entry and exit frequencies exceeding thresholds in real time, and trigger planned or unplanned start-ups and shutdowns of high-power medical equipment. When the environmental disturbance data analysis indicates that the operating room is in a low-disturbance steady state, the weight allocation module increases the fusion weight of the first prediction sub-result. The low-disturbance steady state corresponds to a period in which there are no surgical appointments, little personnel activity, and no high-power equipment operating in the operating room.

[0010] Furthermore, the early warning control module includes: The graded early warning module is configured to pre-store early warning thresholds and alarm thresholds associated with the purification environment threshold. It compares the received dynamic prediction value with the early warning threshold and alarm threshold in real time. When the dynamic prediction value does not exceed the early warning threshold, the status is determined to be normal. When the dynamic prediction value exceeds the early warning threshold but does not reach the alarm threshold, a first-level early warning signal is generated and issued. When the dynamic prediction value reaches the alarm threshold, a second-level alarm signal is generated and issued. Among them, the first-level warning signal and the second-level alarm signal are converted into mutually distinguishable audio-visual alarm forms and structured status messages, and sent to the local monitoring terminal in the operating room and the remote central monitoring platform for display; The forward-looking control module is configured to generate forward-looking control instructions based on dynamic predicted values ​​and the current operating parameters of the air conditioning system. The forward-looking control instructions are used to instruct the air conditioning automatic control system to adjust the chilled water side parameters, supply air temperature, supply air volume and unit operating mode in advance before the dynamic predicted values ​​reach the warning threshold, in order to offset the predicted environmental fluctuations.

[0011] Furthermore, the forward-looking control instructions include intervention measures for the air conditioning automatic control system, the intervention measures including: If the dynamic forecast values ​​indicate that the pressure, temperature and humidity in the core area of ​​the operating room will change in a direction that deviates from the preset comfort range, the forward-looking control command specifically includes a coordinated adjustment strategy for the supply air temperature, supply air volume and unit operating frequency, which is used to proactively input a reverse environmental parameter to the operating room environment to compensate before the predicted deviation occurs. If the dynamic forecast value combined with environmental disturbance data analysis indicates that a known type of disturbance event will occur in the future, the forward-looking control instructions will further include a standard response plan pre-bound to the type of disturbance event. The standard response plan specifies the target operating mode and parameter benchmarks of the air conditioning unit before the event is expected to begin.

[0012] Furthermore, the feature construction module includes: Build submodules for: The environmental disturbance data is parsed based on a predefined rule base, and the event types are labeled and the event levels are divided. The event intensity value is calculated and normalized based on the event duration, scope of impact and event weight coefficient; the co-occurrence frequency of the two types of events within 30 minutes in historical data is statistically analyzed, normalized to obtain the event correlation degree, and an event correlation table is constructed. Using a unified timestamp as an index, construct a structured timeline of timestamp-event ID-event type-event level-intensity value-related event ID-related degree of association; Integration submodule, used for: Based on the structured timeline, different stages of the event are determined; processed multi-source data sequences corresponding to different stages of the event are obtained, and the change characteristics of each type of environmental data in the processed multi-source data sequences corresponding to different stages of the event are calculated to obtain the environmental data responsiveness to the event and normalize it; based on the absolute value of the responsiveness, the event hierarchy coefficient, and the event intensity value, event association weights are assigned to the processed multi-source data for each timestamp and normalized. ; in, The responsiveness of environmental data to events; The rate of change of data during the event; The rate of change of the baseline data before the event; This represents the event intensity value. An integrated data sequence of timestamp-environmental data value-event information-response rate-association weight is generated based on timestamp, environmental data value, event information, responsiveness, and correlation weight. The feature extraction submodule is used for: Based on multi-scale historical windows, the mean, variance, maximum, minimum, median, range, first-order difference mean, second-order difference mean, and number of consecutive increases / decreases of environmental data in the integrated data sequence are calculated. The event types in the integrated data sequence are one-hot encoded, the event levels are binary encoded, the instantaneous values ​​and 30-minute moving averages of the event intensity values ​​are extracted, and the time intervals from the previous similar events and first-level events, the related event type codes, the correlation values, and the average intensity of the related events are extracted. Calculate the instantaneous value of responsivity, 30-minute moving average, peak responsivity and its occurrence time in the integrated data sequence; the coupling value between responsivity and event intensity and the total coupling during the event; the difference between the current data value and the baseline value before the event; the percentage of offset; and the duration of offset; calculate the impact of preceding events and residual offset values. The fusion submodule is used for: Calculate the correlation coefficient between each feature and the event type and the trend of environmental data change, and filter features with correlation coefficients greater than or equal to a preset correlation coefficient threshold to obtain the first feature set; measure the similarity between feature vectors in the first feature set based on cosine similarity, and filter features with similarity coefficients less than or equal to a preset similarity coefficient threshold to obtain the second feature set; divide the features in the second feature set into core features and auxiliary features, add source labeling, validity labeling, and applicable scenario labeling to each feature, and organize the data based on the format of core feature-auxiliary feature-event metadata-data quality label using timestamp as index to generate a fused feature dataset.

[0013] Furthermore, the forward-looking regulation module includes: The preprocessing submodule is used for: The dynamic forecast values ​​are divided into three levels according to the urgency of time: emergency control period, regular control period, and preparatory control period. The deviation difference and deviation rate between the temperature and humidity forecast values ​​and the purification environment threshold are calculated for each period, and the corresponding real-time trend type is marked. Collect real-time operating parameters of the air conditioning system and determine the real-time status of the unit by combining the rated parameter range of the unit. Acquire real-time environmental disturbance data to determine the real-time disturbance scenario; Real-time trend types, real-time unit status, and real-time disturbance scenarios are used as real-time scenario data. A submodule is constructed to acquire historical data on the control of the purification status of the operating room. Based on the historical data, a dynamic strategy pool is constructed for trend type, unit status, and disturbance scenario. Each strategy subset contains core adjustment parameters, auxiliary adjustment parameters, adjustment range, and execution constraints. The filtering submodule is used for: Select a subset of policies from the dynamic policy pool that match the real-time data of the scenario. If there are multiple matching subsets, determine the optimal target policy through the first dual-objective optimization mechanism. When no policy subset is matched, the policy adaptive supplementation mechanism is triggered, which generates a temporary control policy based on historical successful control cases. The decomposition submodule is used to break down the optimal target strategy or temporary control strategy into specific instruction entries for each time period based on three control periods: emergency, normal, and preparation. These initial time-segment instruction entries are divided into core parameter instructions and auxiliary parameter instructions. Each initial time-segment instruction entry includes the parameter adjustment type, the target adjustment value, the execution start time, and the execution duration. The encoding submodule is used to encode the initial time-segmented instruction entries with three levels of priority based on the urgency of the control period. The first priority is to encode all parameter adjustment instructions during the emergency period, the second priority is to encode all instruction entries during the regular control period, and the third priority is to encode all parameter adjustment instructions during the preparation period, thus obtaining the encoded initial time-segmented instruction entries. The simulation submodule is used to construct a virtual operation scenario consistent with the actual scenario based on the current air conditioning system operating parameters and environmental disturbance information. The encoded initial time-segmented instruction entries are input into the virtual operation scenario to simulate the change trajectory of air conditioning system operating parameters and temperature and humidity in the core area of ​​the operating room during the execution of the encoded initial time-segmented instruction entries. The verification submodule is used to perform load safety verification and control effect verification on the operating parameters of the air conditioning system and the temperature and humidity change trajectory of the core area of ​​the operating room during the execution of the instruction. When it is determined that the load safety verification result and the control effect verification result both meet the requirements, the encoded initial time-segmented instruction entry is used as a forward-looking control instruction.

[0014] Compared with the prior art, the beneficial effects of the present invention are: This invention integrates environmental disturbance data such as surgical scheduling, personnel activities, and equipment start-up and shutdown, and deeply fuses them with time-series data from densely deployed sensors to construct a fusion feature set capable of characterizing the complex dynamic processes in the operating room. Based on this, a hybrid prediction model that couples time-series analysis and machine learning is used to not only capture the inherent cycle and trend of environmental parameters, but also to accurately learn the nonlinear mapping relationship between various disturbance events and environmental fluctuations. This enables dynamic and forward-looking prediction of pressure, temperature, and humidity in the core area within a future set time period, while providing accurate information prediction support for the preventive maintenance of key equipment. This helps to more reliably maintain the clean environment required for surgery and reduce the risk of infection. Attached Figure Description

[0015] Fig. 1 This is a schematic diagram of the operating room purification status prediction and early warning system based on multi-source data fusion according to the present invention; Fig. 2 This is an execution diagram of the operating room purification status prediction and early warning system based on multi-source data fusion according to the present invention. Detailed Implementation

[0016] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0017] To address the technical problem of existing operating room environmental control systems, which rely on delayed feedback from fixed sensors and are unable to effectively handle sudden, nonlinear disturbances such as surgical scheduling, personnel movement, and equipment start-up and shutdown, resulting in large fluctuations in environmental parameters and low control accuracy, please refer to [link to relevant documentation]. Figs. 1-2 This embodiment provides the following technical solution: The operating room purification status prediction and early warning system based on multi-source data fusion includes: The data acquisition module is configured to collect operating room environmental data and environmental disturbance data in real time. The environmental data includes at least the time-series pressure and temperature and humidity data obtained by pressure and temperature and humidity sensors deployed in multiple locations inside the operating room, such as the air conditioning unit, inlet and outlet chilled water pipes, air supply ducts, return air ducts, and the operating room itself. The environmental disturbance data includes at least surgical scheduling information, personnel entry and exit logs, and the start and stop status of medical equipment. The data processing module is configured to perform timestamp alignment, missing value processing, and feature extraction on the environmental data and environmental disturbance data to generate a fused feature dataset. The hybrid prediction module includes a coupled time series analysis module and a machine learning module, configured to receive the fused feature dataset and output dynamic prediction values ​​of pressure and temperature and humidity in the core area of ​​the operating room within a future set time period. The time series analysis module is used to capture the periodicity and trend of pressure and temperature and humidity data, and the machine learning module is used to learn the nonlinear mapping relationship between environmental disturbance data and dynamic changes in pressure and temperature and humidity. The early warning control module is configured to compare the dynamic predicted value with a preset purification environment threshold, generate a graded early warning signal before the predicted value deviates from the threshold, and simultaneously feed the dynamic predicted value back to the operating room air conditioning automatic control system to drive the operating room air conditioning automatic control system to perform adaptive regulation.

[0018] The technical effects of the above solution are as follows: the data acquisition module enables comprehensive real-time perception of the operating room environment and disturbance factors, building a high-dimensional data foundation. The data processing module ensures the spatiotemporal consistency and availability of multi-source heterogeneous data, providing a high-quality fused feature dataset for subsequent analysis. Based on this, the hybrid prediction module combines time series analysis and machine learning to effectively capture the inherent variation patterns of pressure and temperature / humidity, and accurately learn the complex nonlinear relationship between external disturbances and the internal environment. This achieves high-precision dynamic prediction of pressure, temperature, and humidity in the core area for future periods. The early warning control module generates tiered early warnings by comparing predicted values ​​with thresholds and achieves closed-loop feedback from predicted values ​​to the air conditioning automatic control system. Ultimately, this achieves a transformation in the operating room purification status from post-event monitoring to pre-event prediction and early warning, and from static control to dynamic adaptive regulation. It also provides accurate information prediction support for the preventive maintenance of key equipment, thereby improving the operating room's environmental safety assurance capabilities and the level of intelligent energy management.

[0019] The data acquisition module includes: The environmental sensing module is configured to collect real-time operating room environmental data through distributed intelligent sensors. This environmental data includes time-series pressure and temperature / humidity data acquired by pressure and temperature / humidity sensors at multiple locations within the operating room, including the air conditioning unit, inlet and outlet chilled water pipes, supply air ducts, return air ducts, and other components. Specifically: The first temperature sensor and the first humidity sensor, which are placed before and after the surface cooling section of the air conditioning unit in the operating room, are used to monitor the initial state of the air handling process and the cooling and dehumidification effect. Second temperature and second humidity sensors are installed in the main air supply duct of the operating room and in each high-efficiency air supply ceiling static pressure box to monitor key air supply parameters after processing and delivery to the surgical area. A three-dimensional temperature and humidity distribution field inside the operating room is constructed by using a third temperature sensor and a third humidity sensor array arranged at multiple different heights in the core surgical area, the surrounding area, and the return air vent. Pressure sensors and temperature sensors are also installed on the inlet and outlet chilled water pipes of the air conditioning unit to monitor the working status of the chilled water system. The disturbance sensing module is configured to collect environmental disturbance data in real time. This environmental disturbance data includes at least surgical scheduling information, personnel entry and exit logs, and the start / stop status of medical equipment, specifically: The surgical scheduling information is actively obtained or received from the hospital information system. The surgical scheduling information includes at least the operating room number, planned start time, estimated duration, surgical type, and number of participating medical staff. And real-time monitoring and recording of personnel and equipment dynamics within the operating room, including: The personnel access monitoring component automatically records the time of personnel entering and exiting and the approximate changes in the number of people by using access control sensors or video analysis units deployed in the operating room buffer room and the main entrance; The equipment status monitoring component obtains real-time start / stop status and operating power change data by connecting to the power monitoring interface or equipment network of key medical equipment (such as anesthesia machines, surgical shadowless lamps, and high-frequency electrosurgical units). The quality verification module is configured to receive raw data from the environmental sensing module and the disturbance sensing module, and perform physical rationality verification (such as whether the temperature value is within an extremely impossible range) and cross-sensor consistency verification (such as whether there are contradictions in the readings of the same type of sensor at different locations). When the verification finds abnormal data, it automatically generates equipment calibration and fault check prompts, and performs interpolation repair on short-term abnormal data based on historical normal data to ensure the continuity and reliability of the data flow input to downstream modules.

[0020] In this embodiment, raw data streams from the environmental sensing module and the disturbance sensing module are received and appended with a unified timestamp and source identifier. Physical rationality verification is performed on the sensor data, i.e., filtering is performed based on preset physical limit thresholds (e.g., operating room temperature should be between 15-30℃, humidity between 30%-60%, and pressure between 5 Pa-25 Pa), and abnormal data points exceeding the reasonable range are marked and temporarily stored. Cross-sensor consistency verification is performed, based on the sensor network topology and physical model (e.g., supply air temperature should be lower than return air temperature, and core area temperature should be higher than surrounding area), analyzing the expected relationship between related sensor readings. If real-time data deviates significantly from the expected relationship (e.g., the difference exceeds the dynamic threshold), it is determined to be a consistency anomaly. Abnormal data detected by the verification is classified and processed: for short-term random missing or abrupt changes caused by transient interference, linear interpolation based on time series or the mean of nearby valid data is used for repair; for data caused by sensor failure, communication... In the event of a prolonged period of data loss due to interruption or system maintenance, a data filling process is initiated. This involves combining the normal data patterns from the same historical period (e.g., the same week or time period) of the sensor with the most relevant normal sensor data to perform joint estimation and filling, and adding explicit data loss labels to the filled data. During processing, if an abnormal pattern indicates possible sensor drift or permanent failure, a device calibration or fault check prompt containing the abnormal location, abnormal type, and suggested actions is automatically generated and pushed to the maintenance interface. After completing all verification and repair steps, a continuous and reliable normalized data stream is output to the data processing module.

[0021] The technical effects of the above solution are as follows: By deploying high-density, multi-layered, and three-dimensional sensors throughout the entire air handling chain (including inlet and outlet chilled water pipes) and the three-dimensional space of the operating room, the environmental sensing module enables panoramic and accurate perception and synchronous monitoring of air conditions (temperature and humidity) and system dynamic conditions (pressure), thus providing more comprehensive and high-quality baseline environmental and system operation data for the predictive model. The disturbance perception module deeply integrates information systems and IoT monitoring to accurately capture multi-dimensional dynamic disturbance factors such as surgical scheduling, personnel flow, and equipment operation, fundamentally establishing a traceable link between environmental changes and disturbance sources. The quality verification module, through its built-in intelligent verification and repair mechanism, ensures the physical rationality and logical consistency of multi-source heterogeneous data at the data source, thereby improving the reliability and robustness of the original data. The data acquisition module, through the collaborative work of the three sub-modules, jointly constructs an integrated data acquisition system with broader sensing dimensions (temperature and humidity + pressure), clear sources, and reliable data, laying a more solid and reliable data foundation for subsequent accurate prediction and intelligent control.

[0022] The data acquisition module employs an adaptive acquisition method, specifically including: Receive surgical scheduling information, which includes at least the operating room number, planned start time, estimated duration, and surgical type; Based on the surgical scheduling information and the current time, the current operational phase of the target operating room is determined, including the preoperative preparation phase, the surgical procedure phase, or the postoperative cleaning phase. Based on the determined current operational stage and the type of surgery, data acquisition strategy instructions are dynamically generated and issued, including at least the following: The sampling frequency of at least some of the pressure and temperature / humidity sensors deployed in the core area of ​​the operating room in the environmental sensing module is adjusted. The monitoring parameters of the components used to monitor personnel entry and exit and the components used to monitor equipment start and stop in the disturbance sensing module are adjusted respectively. The adjustment of the personnel entry and exit component includes monitoring sensitivity and reporting threshold, and the adjustment of the equipment start and stop component includes status judgment conditions and response delay. The adjustment strategy follows this principle: when the surgery is in progress, data is collected from the pressure and temperature / humidity sensors in the core area using a first sampling frequency; when the postoperative cleaning period is in progress, data is collected from the same batch of sensors using a second sampling frequency, with the first sampling frequency being higher than the second sampling frequency.

[0023] The technical effects of the above solution are as follows: By introducing an adaptive acquisition method based on surgical scheduling and operational phases, the system can intelligently sense the working rhythm and critical periods of the operating room, thereby dynamically optimizing the resource allocation and execution strategy for data acquisition. By making differentiated adjustments to the sampling frequency and disturbance monitoring parameters of the core sensors in different operational phases (such as the surgical period and the postoperative cleaning period), it ensures that high spatiotemporal resolution environmental and disturbance data are obtained during the critical surgical period to support accurate prediction and safety warning, while reducing the density of data acquisition and system load during non-critical periods. Thus, while ensuring the core performance of the system, it achieves the unity of intelligent data acquisition, efficient resource utilization, and economical system operation.

[0024] The data processing module includes: The synchronization alignment module is configured to receive asynchronous data streams from the data acquisition module, perform clock correction and sampling point alignment on environmental data and environmental disturbance data based on a unified time reference, and generate a multi-source data sequence with a unified time axis. The missing data processing module is configured to identify and differentiate missing values ​​in multi-source data sequences. For random missing values ​​caused by intermittent failures of regular sensors, it uses linear interpolation based on adjacent valid data to fill in the missing data. For long-term continuous missing values ​​caused by system maintenance or equipment replacement, it uses the most relevant historical data of the same type from the same period to fill in the missing data and adds a missing data label. The feature construction module is configured to parse environmental disturbance data and, based on a predefined rule base, mark the start and end timestamps and event types of disturbance events on a unified time axis. The time axis marked with disturbance events is integrated with the processed multi-source data sequence, and key features are extracted to construct the fused feature dataset. The key features include the statistical features of pressure and temperature and humidity within the current and historical windows, the encoding features of the current disturbance event type, and the time interval since the start of the last similar disturbance event.

[0025] The event types include the surgical preparation stage, the surgical procedure stage, the interoperative period, the period of high population density, and the period of high-temperature equipment activation.

[0026] The predefined rule base is used to automatically identify and label key events with clear physical meaning and environmental impact from structured environmental disturbance data (such as surgical scheduling, personnel access logs, and equipment status). The rule base mainly includes: Surgical phase event rules: Based on the planned start time, expected duration and surgical type of the surgical schedule, define the start and end conditions for the preoperative preparation phase (e.g., 30 minutes before start), the surgical procedure phase (from the planned start to the expected end) and the surgical interval phase (the interval between two consecutive surgeries). Personnel disturbance event rules: Based on access control, define the criteria for determining periods of dense personnel entry and exit, such as when the total number of people entering and exiting exceeds N people within 5 minutes or when the estimated number of people in the room in real time exceeds X people; Equipment disturbance event rules: Based on equipment status and power data, define the criteria for determining the activation period of high-heat equipment, such as an anesthesia machine or high-frequency electrosurgical unit entering a high-power operation state from standby state for more than T minutes.

[0027] In this embodiment, the synchronization alignment module receives asynchronous raw data streams with different sampling periods (such as high-frequency sensor data and low-frequency disturbance event data) from the data acquisition module. The synchronization alignment module maintains a high-precision, unified system time base internally. For each incoming data, it first compares and corrects the clock offset with the system time base based on its own timestamp and data source identifier to eliminate errors caused by device clock asynchrony. The synchronization alignment module defines a unified resampling time point sequence for all data sources according to a preset target synchronization frequency (such as 1 minute / time). For continuous sensor data (such as temperature and humidity), it uses the mean or end value within the sampling interval for resampling. For discrete event data (such as personnel entry and exit, equipment start and stop), it determines whether the occurrence time falls within the current synchronization time window. If so, it is marked as a valid event. All data that has undergone clock correction and resampling is arranged and packaged according to a unified time axis to generate a multi-source data sequence with strictly aligned timestamps and synchronized data channels, which is then output to the downstream module.

[0028] In this embodiment, the missing data processing module receives a unified timeline multi-source data sequence from the synchronization alignment module. The module iterates through each time point and each data channel of the data sequence to accurately identify the location of missing data. It classifies missing values ​​according to the missing pattern (e.g., missing duration, type of missing data, status of adjacent data points). For random missing values ​​caused by transient interference or communication jitter and with a short duration (e.g., less than 3 consecutive sampling points), it quickly fills the missing data using linear interpolation based on adjacent valid data points. For continuous missing values ​​with a long duration caused by planned maintenance, equipment replacement, or long-term failures, it initiates an advanced filling strategy: based on the sensor network topology and physical correlation (e.g., spatial proximity, functional relevance), it automatically selects several normal channels with the strongest correlation to the missing data channel as references; it retrieves historical data from the historical database that has the same temporal context as the current missing time period (e.g., the same day of the week, the same time of day, similar surgical stages); and it constructs a regression or similarity model by combining the current data and the patterns of historical data from the reference channels to estimate and fill the missing values. All data generated through imputation are labeled with missing data to record the imputation type and confidence level, ultimately outputting a continuous, complete data sequence with integrity metadata.

[0029] The technical effects of the above solution are as follows: the synchronization alignment module ensures the accurate fusion of multi-source heterogeneous data under a unified time benchmark, thus establishing a reliable time series foundation for subsequent analysis; the missing data handling module adopts intelligent and differentiated imputation strategies for different types of missing data, ensuring data integrity and continuity to the maximum extent while retaining the meta-information of data quality by adding labels; the feature labeling module transforms discrete disturbance logs into structured events with clear time series and category information, thereby revealing the potential driving force of environmental changes; and the feature construction module integrates clean time series data with labeled disturbance events to extract a high-dimensional feature set that contains both the statistical regularity of the environment itself and the correlation of external disturbances and the historical dependence of events. Based on the above processing flow, a fundamental guarantee can be provided for the accuracy of subsequent predictions.

[0030] The hybrid prediction module also includes: The feature allocation module is configured to divide the fused feature dataset into a first data subset and a second data subset and perform intelligent distribution based on the type and source of the data features; The time series analysis module receives a first data subset with continuous time series characteristics allocated by the feature allocation module. The first data subset includes at least historical and current pressure and temperature and humidity time series data. The machine learning module receives a second data subset representing discrete events assigned by the feature allocation module. The second data subset includes at least surgical scheduling, personnel entry and exit events, and equipment status change data. The weight allocation module is configured to dynamically determine the fusion weight of the first prediction sub-result and the second prediction sub-result based on the activity analysis of current and historical environmental disturbance data. The fusion weight is adjusted inversely based on the level of business disturbance in the operating room. The determination of the business disturbance level is based on real-time and comprehensive analysis of current and historical environmental disturbance data, specifically: The system continuously monitors and summarizes the type, frequency, intensity and duration of various disturbance events from the feature labeling module. For example, it can count the number and level of planned and ongoing surgeries per unit time (such as the next 30 minutes), calculate the frequency and peak number of personnel entering and leaving in real time, and monitor the start-up and shutdown status and operating power level of high-fever medical equipment. The system compares these quantitative indicators with preset thresholds and performs a comprehensive score based on a set of predefined grading rules, which are typically set as follows: When the surgical schedule is dense in the future (such as two consecutive major surgeries), the real-time monitoring shows that the frequency of personnel entering and exiting exceeds the high threshold, or there is a concentrated start-up and shutdown of high-power medical equipment, whether planned or unplanned, the system will determine it as a high disturbance level. When there are no surgeries scheduled in the operating room, there is little activity among personnel, and no high-power equipment is running, it is determined to be a low-disturbance steady-state level. Cases falling between these two categories are classified as medium disturbance levels based on the weighted scores of specific indicators. The rating process is dynamic and is updated in real time as new events occur and old events end.

[0031] The weight allocation module adjusts the fusion weights according to the level of business disturbance, specifically as follows: When environmental disturbance data analysis indicates that the operating room is at a high disturbance level, the weight allocation module increases the fusion weight of the second prediction sub-result. The high disturbance level is triggered by the following conditions: In the future, the system will be designed to detect intensive surgical schedules within a set time period, monitor personnel entry and exit frequencies exceeding thresholds in real time, and trigger planned or unplanned start-ups and shutdowns of high-power medical equipment. When the environmental disturbance data analysis indicates that the operating room is in a low-disturbance steady state, the weight allocation module increases the fusion weight of the first prediction sub-result. The low-disturbance steady state corresponds to a period in which there are no surgical appointments, little personnel activity, and no high-power equipment operating in the operating room.

[0032] The technical effects of the above solution are as follows: By intelligently distinguishing and distributing the source and type of fused features through the feature allocation module, it is possible to ensure that the time series analysis module can focus on learning the inherent continuous change patterns of pressure and temperature and humidity data, while allowing the machine learning module to more purely explore the complex mapping relationship between discrete disturbance events and environmental responses, thus giving full play to the advantages of the two different technical paths; The weight allocation module introduces a dynamic weight mechanism based on the activity of environmental disturbances, which can intelligently adjust the emphasis ratio of dependence on historical trends and immediate event responses according to the intensity of the actual business state of the operating room (such as the period of intensive surgery and the period of calm recovery), thereby achieving flexible matching between the prediction logic and the real physical scene, and thus improving the overall adaptability and prediction accuracy of the hybrid prediction model in different operating stages.

[0033] The early warning and control module includes: The graded early warning module is configured to pre-store early warning thresholds and alarm thresholds associated with the purification environment threshold. It compares the received dynamic prediction value with the early warning threshold and alarm threshold in real time. When the dynamic prediction value does not exceed the early warning threshold, the status is determined to be normal. When the dynamic prediction value exceeds the early warning threshold but does not reach the alarm threshold, a first-level early warning signal is generated and issued. When the dynamic prediction value reaches the alarm threshold, a second-level alarm signal is generated and issued. Among them, the first-level warning signal and the second-level alarm signal are converted into mutually distinguishable audio-visual alarm forms and structured status messages, and sent to the local monitoring terminal in the operating room and the remote central monitoring platform for display; The forward-looking control module is configured to generate forward-looking control instructions based on dynamic predicted values ​​and the current operating parameters of the air conditioning system. The forward-looking control instructions are used to instruct the air conditioning automatic control system to adjust the chilled water side parameters (such as flow rate, pressure, valve opening), supply air temperature, supply air volume and unit operating mode in advance before the dynamic predicted values ​​reach the warning threshold, in order to offset the predicted environmental fluctuations.

[0034] The technical effects of the above solution are as follows: the hierarchical early warning module upgrades the single threshold comparison into a predictive hierarchical early warning system, realizing the transformation from passive response to active intervention, enabling managers to intervene in potential risks earlier and more flexibly. The signal release module ensures that early warning information can be accurately, timely and unambiguously perceived and received through multimodal and multi-terminal release methods, forming an effective alarm path from the system to personnel. The forward-looking control module realizes the closed loop of prediction and control, and improves the stability, accuracy and energy efficiency of operating room environmental control by generating and issuing control instructions in advance based on predicted values.

[0035] Forward-looking control directives include intervention measures for air conditioning automatic control systems, including: If the dynamic forecast values ​​indicate that the pressure, temperature and humidity in the core area of ​​the operating room will change in a direction that deviates from the preset comfort range, the forward-looking control command specifically includes a coordinated adjustment strategy for the supply air temperature, supply air volume and unit operating frequency, which is used to proactively input a reverse environmental parameter to the operating room environment to compensate before the predicted deviation occurs. If the dynamic forecast value combined with environmental disturbance data analysis indicates that a known type of disturbance event will occur in the future, the forward-looking control instructions will further include a standard response plan pre-bound to the type of disturbance event. The standard response plan specifies the target operating mode and parameter benchmarks of the air conditioning unit before the event is expected to begin.

[0036] The technical effects of the above-mentioned technical solution are as follows: It achieves accurate pre-compensation for environmental fluctuations and proactive response to known disturbances through forward-looking control commands. Specifically, for predicted deviations in pressure and temperature and humidity, the system applies reverse adjustment in advance through a coordinated strategy of supply air temperature, supply air volume and unit operating frequency, transforming the lagging response of traditional feedback control into a feedforward proactive intervention, thereby effectively suppressing the amplitude and duration of fluctuations in environmental parameters. At the same time, by dynamically binding specific disturbance events (such as the start of surgery or equipment activation) with preset standard response plans, the air conditioning system can automatically switch to the optimal operating mode before the disturbance occurs, thereby directly converting the predicted information into preventive control actions.

[0037] Working Principle: The data acquisition module collects real-time data on operating room environmental pressure, temperature, and humidity, as well as disturbance data such as surgical scheduling, personnel entry and exit, and equipment start-up and shutdown. The data processing module performs time alignment, feature extraction, and labeling to construct a unified dataset that integrates time series and event features. Based on this, the hybrid prediction module captures the inherent trends and cycles of pressure, temperature, and humidity through the sequence analysis module, while the machine learning module learns the complex nonlinear relationship between disturbance events and changes in environmental parameters. Together, they generate dynamic predictions of pressure, temperature, and humidity for future periods. On this basis, the early warning control module compares these predictions with preset thresholds and issues tiered warnings before the parameters actually exceed the limits. At the same time, the predictions are converted into forward-looking control instructions and fed back to the air conditioning automatic control system, realizing the transformation from "monitoring-response" to "prediction-prevention." By actively offsetting the predicted disturbances, environmental parameter fluctuations are significantly reduced, thereby better ensuring surgical safety and personnel comfort.

[0038] The feature construction module includes: Build submodules for: The environmental disturbance data is parsed based on a predefined rule base, and the event types are labeled and the event levels are divided. The event intensity value is calculated and normalized based on the event duration, scope of impact and event weight coefficient; the co-occurrence frequency of the two types of events within 30 minutes in historical data is statistically analyzed, normalized to obtain the event correlation degree, and an event correlation table is constructed. Using a unified timestamp as an index, construct a structured timeline of timestamp-event ID-event type-event level-intensity value-related event ID-related degree of association; Integration submodule, used for: Based on the structured timeline, different stages of the event are determined; processed multi-source data sequences corresponding to different stages of the event are obtained, and the change characteristics of each type of environmental data in the processed multi-source data sequences corresponding to different stages of the event are calculated to obtain the environmental data responsiveness to the event and normalize it; based on the absolute value of the responsiveness, the event hierarchy coefficient, and the event intensity value, event association weights are assigned to the processed multi-source data for each timestamp and normalized. ; in, The responsiveness of environmental data to events; The rate of change of data during the event; The rate of change of the baseline data before the event; This represents the event intensity value. An integrated data sequence of timestamp-environmental data value-event information-response rate-association weight is generated based on timestamp, environmental data value, event information, responsiveness, and correlation weight. The feature extraction submodule is used for: Based on multi-scale historical windows, the mean, variance, maximum, minimum, median, range, first-order difference mean, second-order difference mean, and number of consecutive increases / decreases of environmental data in the integrated data sequence are calculated. The event types in the integrated data sequence are one-hot encoded, the event levels are binary encoded, the instantaneous values ​​and 30-minute moving averages of the event intensity values ​​are extracted, and the time intervals from the previous similar events and first-level events, the related event type codes, the correlation values, and the average intensity of the related events are extracted. Calculate the instantaneous value of responsivity, 30-minute moving average, peak responsivity and its occurrence time in the integrated data sequence; the coupling value between responsivity and event intensity and the total coupling during the event; the difference between the current data value and the baseline value before the event; the percentage of offset; and the duration of offset; calculate the impact of preceding events and residual offset values. The fusion submodule is used for: Calculate the correlation coefficient between each feature and the event type and the trend of environmental data change, and filter features with correlation coefficients greater than or equal to a preset correlation coefficient threshold to obtain the first feature set; measure the similarity between feature vectors in the first feature set based on cosine similarity, and filter features with similarity coefficients less than or equal to a preset similarity coefficient threshold to obtain the second feature set; divide the features in the second feature set into core features and auxiliary features, add source labeling, validity labeling, and applicable scenario labeling to each feature, and organize the data based on the format of core feature-auxiliary feature-event metadata-data quality label using timestamp as index to generate a fused feature dataset.

[0039] In this embodiment, a submodule is constructed to parse the environmental disturbance data based on a predefined rule base, mark event types, and divide event levels. The event types include the surgical preparation stage, the surgical procedure stage, the surgical interval stage, the period of dense personnel entry and exit, and the period of high-fever equipment activation. The event levels include primary events and secondary events. Primary events are the surgical procedure stage, the period of dense personnel entry and exit, and the period of high-fever equipment activation. Secondary events are the surgical preparation stage and the surgical interval stage.

[0040] In this embodiment, the event intensity value is calculated based on the event duration, impact range, and event weight coefficient. The event intensity value is calculated as follows: Event Intensity Value = (Event Duration / Baseline Duration) × (Impact Range / Total Area of ​​Monitoring Area) × Event Weight Coefficient. The baseline duration is 60 minutes for Level 1 events and 30 minutes for Level 2 events. The event weight coefficients are: 1.0 for the surgical procedure phase, 0.9 for periods of high personnel traffic, 0.8 for periods of high-temperature equipment activation, 0.5 for the surgical preparation phase, and 0.4 for the surgical interval phase. The event intensity value is a comprehensive quantitative indicator that measures the potential impact of an event on the monitoring environment from three dimensions: time, space, and business importance. The time dimension reflects the event's duration (longer duration, higher intensity) by dividing the event duration by the baseline duration. The spatial dimension reflects the event's spatial coverage (wider coverage, higher intensity). The business dimension reflects the event's business importance through the event weight coefficient (e.g., the surgical procedure phase has the highest weight, indicating the most critical impact on the environment). In simple terms, a higher event intensity value indicates a greater overall impact of the event on the monitoring environment; conversely, a lower value indicates a smaller impact. Changes in the event intensity value directly affect subsequent response calculations, weight allocation, and event analysis processes, influencing the stability and sensitivity of response calculations, the magnitude of event-related weights, and the priority of event-level verification and data association.

[0041] In this embodiment, the co-occurrence frequency of two types of events within 30 minutes in historical data is statistically analyzed, normalized to obtain the event correlation degree, and an event correlation table is constructed. The co-occurrence frequency is calculated by statistically analyzing the frequency of consecutive occurrence of two types of events within 30 minutes in historical data and normalizing it to obtain the correlation degree. The correlation rules are stored, such as the correlation degree of 0.85 for "high fever equipment activation period - dense personnel entry and exit period" and 0.92 for "surgery in progress stage - surgery preparation stage", forming the event correlation table.

[0042] In this embodiment, the coupling value between responsiveness and event intensity and the total coupling amount during the event are defined as follows: the coupling value is the product of responsiveness and event intensity; the total coupling amount is the cumulative sum of the coupling values.

[0043] In this embodiment, the change characteristics of each type of environmental data at different stages of the event are calculated to obtain the responsiveness of the environmental data to the event. The different stages of the event include the 30 minutes before the event, the event in progress, and the 30 minutes after the event. ;in, The responsiveness of environmental data to events; The rate of change of data during the event; The rate of change of the baseline data before the event; The event intensity value; the data change rate = (maximum data value - minimum data value) / event duration; the baseline change rate is the average change rate of the 30 minutes prior to the event.

[0044] In this embodiment, based on the absolute value of responsiveness, the event hierarchy coefficient, and the event intensity value, an event association weight is assigned to the multi-source data for each timestamp. Specifically, the event hierarchy coefficient is: 1.0 for first-level events and 0.6 for second-level events; the association weight = absolute value of responsiveness × event hierarchy coefficient (1.0 for first-level events and 0.6 for second-level events) × event intensity value.

[0045] In this embodiment, the influence of preceding events and the residual offset value are calculated, wherein the influence of preceding events = correlation of the previous event × responsiveness of the previous event × 0.3; and the residual offset value = current data correction amount × influence of preceding events.

[0046] In this embodiment, the correlation coefficient between each feature and the event type and the trend of environmental data change is calculated, where the correlation coefficient is the Pearson correlation coefficient and the threshold is 0.3.

[0047] In this embodiment, the core features include the mean and rate of change in the basic statistical features, the event coding features, the coupling value and peak value in the responsiveness features, and the auxiliary features include the time interval in the spatiotemporal correlation features and the influence of preceding events in the cross-event transmission features.

[0048] The working principle and beneficial effects of the above technical solution are as follows: By constructing sub-modules to analyze environmental disturbance data, label event types, divide levels, calculate event intensity values ​​and correlations, and construct a structured timeline, the key elements of events can be clearly organized and quantified, providing a standardized and structured data foundation for subsequent analysis, facilitating a comprehensive and detailed understanding of events; the integration sub-module determines the event stage based on the structured timeline, analyzes the changing characteristics of environmental data at different stages of the event, calculates responsibility and assigns event correlation weights, generating an integrated data sequence. This helps to deeply understand the response patterns of environmental data to events and provides data support for studying the intrinsic relationship between events and the environment; the feature extraction sub-module extracts rich and diverse features from the integrated data sequence, covering environmental data statistical features, event coding features, responsibility-related features, and preceding event impact features, reflecting the characteristics of events and environmental data from multiple dimensions, providing sufficient information for more accurate event analysis and prediction; the fusion sub-module filters features by calculating correlation and similarity, obtains an effective feature set, classifies and labels the features, and organizes and generates a fused feature dataset. This not only removes redundant and irrelevant features, but also improves the usability and interpretability of the data, making the analysis results more accurate and reliable, which is beneficial for subsequent modeling and predictive applications.

[0049] The forward-looking regulation module includes: The preprocessing submodule is used for: The dynamic forecast values ​​are divided into three levels according to the urgency of time: emergency control period, regular control period, and preparatory control period. The deviation difference and deviation rate between the temperature and humidity forecast values ​​and the purification environment threshold are calculated for each period, and the corresponding real-time trend type is marked. Collect real-time operating parameters of the air conditioning system and determine the real-time status of the unit by combining the rated parameter range of the unit. Acquire real-time environmental disturbance data to determine the real-time disturbance scenario; Real-time trend types, real-time unit status, and real-time disturbance scenarios are used as real-time scenario data. A submodule is constructed to acquire historical data on the control of the purification status of the operating room. Based on the historical data, a dynamic strategy pool is constructed for trend type, unit status, and disturbance scenario. Each strategy subset contains core adjustment parameters, auxiliary adjustment parameters, adjustment range, and execution constraints. The filtering submodule is used for: Select a subset of policies from the dynamic policy pool that match the real-time data of the scenario. If there are multiple matching subsets, determine the optimal target policy through the first dual-objective optimization mechanism. When no policy subset is matched, the policy adaptive supplementation mechanism is triggered, which generates a temporary control policy based on historical successful control cases. The decomposition submodule is used to break down the optimal target strategy or temporary control strategy into specific instruction entries for each time period based on three control periods: emergency, normal, and preparation. These initial time-segment instruction entries are divided into core parameter instructions and auxiliary parameter instructions. Each initial time-segment instruction entry includes the parameter adjustment type, the target adjustment value, the execution start time, and the execution duration. The encoding submodule is used to encode the initial time-segmented instruction entries with three levels of priority based on the urgency of the control period. The first priority is to encode all parameter adjustment instructions during the emergency period, the second priority is to encode all instruction entries during the regular control period, and the third priority is to encode all parameter adjustment instructions during the preparation period, thus obtaining the encoded initial time-segmented instruction entries. The simulation submodule is used to construct a virtual operation scenario consistent with the actual scenario based on the current air conditioning system operating parameters and environmental disturbance information. The encoded initial time-segmented instruction entries are input into the virtual operation scenario to simulate the change trajectory of air conditioning system operating parameters and temperature and humidity in the core area of ​​the operating room during the execution of the encoded initial time-segmented instruction entries. The verification submodule is used to perform load safety verification and control effect verification on the operating parameters of the air conditioning system and the temperature and humidity change trajectory of the core area of ​​the operating room during the execution of the instruction. When it is determined that the load safety verification result and the control effect verification result both meet the requirements, the encoded initial time-segmented instruction entry is used as a forward-looking control instruction.

[0050] In this embodiment, the dynamic forecast value is divided into three levels according to the urgency of time: emergency control period, regular control period, and preparatory control period. Specifically, the emergency control period is the next 5-10 minutes, the regular control period is the next 10-30 minutes, and the preparatory control period is the next 30-60 minutes.

[0051] In this embodiment, the corresponding trend type is marked, where the trend type is, for example, "emergency period - rapid heating - deviation difference 0.8℃ - deviation rate 0.2℃ / minute"; "normal period - slow humidification - deviation difference 5%RH - deviation rate 0.5%RH / minute".

[0052] In this embodiment, real-time operating parameters of the air conditioning system are collected, including supply air temperature, supply air volume, fan speed, opening degree of chilled water valve / hot water valve, filter pressure difference, and unit operating mode (cooling / heating / dehumidification / air supply).

[0053] In this embodiment, real-time environmental disturbance data is acquired to determine the disturbance scenario. The real-time environmental disturbance data includes the current surgical stage (preoperative preparation / intraoperative operation / postoperative cleaning), the number of people in the operating room, and the start / stop status of large medical equipment (such as electrosurgical unit, ultrasound equipment). The disturbance scenario is, for example, a surgical scenario with intraoperative operation + dense personnel + all equipment in operation.

[0054] In this embodiment, each strategy subset includes core adjustment parameters, auxiliary adjustment parameters, adjustment range, and execution constraints. Core adjustment parameters are key parameters that directly affect temperature and humidity changes, such as supply air temperature and supply air volume. Auxiliary adjustment parameters are supporting parameters that optimize control effects, such as supply air angle and filter pressure difference threshold. The adjustment range is a safe adjustment range that matches the unit's status. Execution constraints include, for example, that the supply air temperature adjustment should not exceed 2°C in a single operation under high-load critical conditions. For example, the strategy subset for "emergency rapid heating + low-load steady state + intensive intraoperative scenario" includes the following: "Core parameters: supply air temperature reduced by 2-3°C, supply air volume increased by 10%-15%; Auxiliary parameters: adjust the angle of the top air outlet to vertically downward; Range: fan speed adjustment not exceeding 20% ​​of the rated value; Constraints: avoid unit load exceeding 50%."

[0055] In this embodiment, the optimal objective strategy is determined through a first dual-objective optimization mechanism. The first dual-objective optimization mechanism has the objective of maximizing the control response speed as the first objective and minimizing the unit load loss as the second objective. Pareto front analysis is used to select a compromise solution, and the objective weights are dynamically allocated according to the current unit status: when the unit load rate is below 30%, the response speed weight is 0.8 and the load loss weight is 0.2; when the load rate is above 70%, the response speed weight is 0.3 and the load loss weight is 0.7.

[0056] In this embodiment, a temporary control strategy is generated based on successful historical control cases. The core logic of the temporary strategy is to make small adjustments in stages and correct them in real time. For example, the supply air temperature is reduced by 1°C first, and the temperature and humidity changes are observed within 5 minutes before deciding whether to continue adjusting, so as to avoid unit failure due to large adjustments.

[0057] In this embodiment, the optimal target strategy or temporary control strategy is broken down into specific time-segmented instruction entries based on three control periods: emergency, normal, and preparation, serving as initial time-segmented instruction entries. These initial time-segmented instruction entries are divided into core parameter instructions and auxiliary parameter instructions. Each initial time-segmented instruction entry includes the parameter adjustment type, the target adjustment value, the execution start time, and the execution duration, specifically: Emergency Control Period Command Item Breakdown: Within the next 5-10 minutes, with the goal of quickly offsetting impending temperature and humidity deviations, the priority is set as core parameters over auxiliary parameters; Core Parameter Command Breakdown: For core adjustment parameters in the optimal strategy (such as supply air temperature and supply air volume), based on the deviation rate of dynamic predicted values ​​and the low / medium load status of the unit (the adjustment range needs to be reduced under the critical state of high load), quantify the commands: Command Item 1 (Supply Air Temperature): Parameter Type = Supply Air Temperature; Adjustment Type = Decrease; Adjustment Target Value = Current Supply Air Temperature - 2.5℃; Execution Start Time = Immediate; Execution Duration = 10 minutes; Safety Constraint Clause = Single Adjustment Range ≤ 3℃, Unit Load Rate ≤ 50%; Command Item 2 (Supply Air Volume): Parameter Type = Supply Air Volume; Adjustment Type = Increase; Adjustment Target Value = Current Supply Air Volume + 15%; Execution Start Time = Immediate; Execution Duration = 10 minutes Minutes; Safety constraint clause = Fan speed adjustment ≤ 20% of rated value, avoid airflow velocity > 0.3m / s; Auxiliary parameter instruction breakdown: For the auxiliary adjustment parameters (such as air supply angle, air outlet opening) in the optimal strategy, with the goal of enhancing the core parameter control effect, accurately match the core parameter adjustment logic: Instruction item 3 (air supply angle): Parameter type = top air supply outlet angle; Adjustment type = adjustment; Adjustment target value = vertically downward (0° angle); Execution start time = immediately; Execution duration = 10 minutes; Safety constraint clause = no airflow directly hitting the surgical area during angle adjustment; Instruction item 4 (filter pressure difference threshold): Parameter type = filter pressure difference monitoring threshold; Adjustment type = adaptation; Adjustment target value = current threshold + 5Pa; Execution start time = immediately; Execution duration = 10 minutes; Safety constraint clause = pressure difference threshold does not exceed the unit's rated tolerance value of 100Pa.

[0058] Breakdown of Command Items for Regular Control Periods: In the next 10-30 minutes, the core objective is to maintain stable temperature and humidity, avoiding secondary fluctuations. The breakdown logic involves fine-tuning core parameters and optimizing auxiliary parameters. The breakdown operations are as follows: Core Parameter Command Breakdown: Based on the predicted effect of control during emergency periods and the medium-term trend of dynamic forecast values, the core parameters are slightly modified to ensure that temperature and humidity gradually approach the standard thresholds: Command Item 5 (Supply Air Temperature): Parameter Type = Supply Air Temperature; Adjustment Type = Fine-tuning Increase; Adjustment Target Value = Emergency Period Target Value + 0.5℃; Execution Start Time = After the Emergency Period Ends (i.e., 10 minutes); Execution Duration = 20 minutes; Safety Constraint Clause = Adjustment Amount ≤ 1℃ / time, Unit Load Rate Maintained at 40%-60%; Command Item 6 (Supply Air Volume): Parameter Type = Supply Air Volume; Adjustment Type = Fine-tuning Decrease; Adjustment Target Value = Emergency Period Target Value - 5%; Execution Start Time =10th minute; Execution duration = 20 minutes; Safety constraint clause = Air supply volume not lower than the standard air volume corresponding to the minimum air exchange rate in the operating room; Auxiliary parameter instruction breakdown: For the persistence of environmental disturbances (such as dense personnel, continuous equipment operation), optimize auxiliary parameters to improve control stability: Instruction item 7 (Air supply path zoning): Parameter type = Zoning air supply ratio; Adjustment type = Optimization; Adjustment target value = 70% air supply ratio in the core area of ​​the operating room and 30% in the surrounding area; Execution start time = 10th minute; Execution duration = 20 minutes; Safety constraint clause = Zoning air volume distribution deviation ≤ ±5%; Instruction item 8 (Dehumidification module operation intensity): Parameter type = Dehumidification module operating power; Adjustment type = Maintenance; Adjustment target value = 60% of rated power; Execution start time = 10th minute; Execution duration = 20 minutes; Safety constraint clause = Dehumidification capacity not exceeding 120% of the current environmental humidity requirement; The breakdown of instructions for the pre-control period is as follows: In the next 30-60 minutes, the core objective is to prevent potential deviations and gradually restore the basic operating state. The breakdown logic is core parameter regression + auxiliary parameter reset. The breakdown operation is as follows: Core parameter instruction breakdown: Based on the long-term stable trend of dynamic prediction values ​​(or the prediction of weakening disturbance factors), the core parameters are gradually returned to the system's basic setpoints: Instruction item 9 (supply air temperature): Parameter type = supply air temperature; Adjustment type = regression; Adjustment target value = standard supply air temperature for the operating room; Execution start time = after the end of the regular period (i.e., the 30th minute); Execution duration = 30 minutes; Safety constraint clause = adjustment range ≤ 0.3℃ every 5 minutes to avoid sudden temperature changes; Instruction item 10 (supply air volume): Parameter type = supply air volume; Adjustment type = regression; Adjustment target value = system basic supply air volume; Execution start time = the 30th minute; Execution duration = 30 minutes; Duration of execution = 30 minutes; Safety constraint clause = Maintain the minimum air exchange rate standard during the return process, and gradually reduce the unit load rate to below 40%; Auxiliary parameter instruction breakdown: Simultaneously reset the auxiliary parameters to the normal operating state to reserve redundancy for the next round of control: Instruction item 11 (air supply angle): Parameter type = top air supply outlet angle; Adjustment type = reset; Adjustment target value = normal operating angle; Execution start time = 30 minutes; Execution duration = 30 minutes; Safety constraint clause = the angle reset process is smooth and no airflow turbulence is generated; Instruction item 12 (filter pressure difference threshold): Parameter type = filter pressure difference monitoring threshold; Adjustment type = reset; Adjustment target value = system standard threshold (e.g., 80Pa); Execution start time = 30 minutes; Execution duration = 30 minutes; Safety constraint clause = the deviation between the threshold reset and the current actual filter pressure difference ≤ ±3Pa.

[0059] In the above disassembly process, the adjustment target values ​​of all instruction items are determined quantitatively based on the adjustment range of the optimal strategy, the current operating parameters of the unit, and the degree of deviation of the dynamic prediction value. Moreover, the execution start time and duration of the core parameter instruction and the auxiliary parameter instruction are fully coordinated within the same time period to ensure the consistency of multi-parameter linkage control.

[0060] In this embodiment, load safety verification and control effect verification are performed on the operating parameters of the air conditioning system and the temperature and humidity change trajectory of the core area of ​​the operating room during the execution of the instructions. The load safety verification includes whether the unit load rate, continuous running time, and parameter fluctuation amplitude meet the requirements of the constraint clauses. The control effect verification includes whether the deviation value between the temperature and humidity trajectory and the purification environment threshold meets the requirements. When the verification results meet the requirements, the encoded initial time-segmented instruction entries are used as forward-looking control instructions. When any verification fails, a rollback mechanism is executed: if the load safety verification fails, the instruction amplitude is adjusted according to the unit constraint conditions, and the simulation submodule is re-entered; if the control effect verification fails and there is no feasible adjustment scheme, it is downgraded to executing only the first-level priority instruction and triggering a manual confirmation request to the remote central monitoring platform; when three consecutive verifications fail, the default conservative strategy is activated: the current operating parameters are maintained and only a warning signal is issued.

[0061] The working principle and beneficial effects of the above technical solution are as follows: The preprocessing submodule divides the control period according to the time urgency, calculates the temperature and humidity deviation difference and rate, marks the real-time trend type, and forms real-time scenario data by combining the unit status and disturbance scenario. This can accurately grasp the current status and provide a detailed and dynamic evaluation basis for subsequent control. The construction submodule builds a dynamic strategy pool based on historical data, and the filtering submodule can select matching strategy subsets from it. It can also generate temporary control strategies when there are no matches, ensuring the diversity and adaptability of control strategies and being able to cope with various complex scenarios. The decomposition submodule decomposes the strategy into specific instruction items for different time periods, and the encoding submodule encodes them according to the urgency of the control period, making the control instructions more targeted and orderly, and ensuring that emergency situations are handled first. The simulation submodule constructs a virtual operation scenario, simulates the parameter and temperature and humidity change trajectory during instruction execution, and the verification submodule verifies the load safety and control effect, ensuring the safety and effectiveness of forward-looking control instructions in actual execution and reducing control risks.

[0062] It should be noted that, in this document, relational terms such as "first" and "second" are used only to distinguish one entity or operation from another, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Furthermore, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such process, method, article, or apparatus.

[0063] Although embodiments of the invention have been shown and described, it will be understood by those skilled in the art that various changes, modifications, substitutions and alterations can be made to these embodiments without departing from the principles and spirit of the invention.

Claims

1. An operating room purification status prediction and early warning system based on multi-source data fusion, characterized in that, include: The data acquisition module is configured to collect operating room environmental data and environmental disturbance data in real time. The environmental data includes at least the time-series pressure and temperature and humidity data obtained by pressure and temperature and humidity sensors deployed in multiple locations inside the operating room, such as the air conditioning unit, inlet and outlet chilled water pipes, air supply ducts, return air ducts, and the operating room itself. The environmental disturbance data includes at least surgical scheduling information, personnel entry and exit logs, and the start and stop status of medical equipment. The data processing module is configured to perform timestamp alignment, missing value processing, and feature extraction on the environmental data and environmental disturbance data to generate a fused feature dataset. The hybrid prediction module includes a coupled time series analysis module and a machine learning module, configured to receive the fused feature dataset and output dynamic prediction values ​​of pressure and temperature and humidity in the core area of ​​the operating room within a future set time period. The time series analysis module is used to capture the periodicity and trend of pressure and temperature and humidity data, and the machine learning module is used to learn the nonlinear mapping relationship between environmental disturbance data and dynamic changes in pressure and temperature and humidity. The early warning control module is configured to compare the dynamic predicted value with a preset purification environment threshold, generate a graded early warning signal before the predicted value deviates from the threshold, and simultaneously feed the dynamic predicted value back to the operating room air conditioning automatic control system to drive the operating room air conditioning automatic control system to perform adaptive regulation.

2. The operating room purification status prediction and early warning system based on multi-source data fusion according to claim 1, characterized in that, The data acquisition module includes: The environmental sensing module is configured to collect operating room environmental data in real time through distributed intelligent sensors. The environmental data includes time-series pressure and temperature and humidity data obtained by pressure and temperature and humidity sensors at multiple locations inside the operating room, such as air conditioning units, inlet and outlet chilled water pipes, air supply ducts, return air ducts, and pressure and temperature and humidity sensors. The disturbance sensing module is configured to collect environmental disturbance data in real time. The environmental disturbance data includes at least surgical scheduling information, personnel entry and exit logs, and the start and stop status of medical equipment. The quality verification module is configured to receive raw data from the environmental sensing module and the disturbance sensing module, perform physical rationality verification and cross-sensor consistency verification on the sensor data, and automatically generate equipment calibration and fault check prompts when the verification finds abnormal data, and interpolate and repair short-term abnormal data based on historical normal data.

3. The operating room purification status prediction and early warning system based on multi-source data fusion according to claim 2, characterized in that, The data acquisition module employs an adaptive acquisition method, specifically including: Receive surgical scheduling information, which includes at least the operating room number, planned start time, estimated duration, and surgical type; Based on the surgical scheduling information and the current time, the current operational phase of the target operating room is determined, including the preoperative preparation phase, the surgical procedure phase, or the postoperative cleaning phase. Based on the determined current operational stage and the type of surgery, data acquisition strategy instructions are dynamically generated and issued, including at least the following: The sampling frequency of at least some of the pressure and temperature / humidity sensors deployed in the core area of ​​the operating room in the environmental sensing module is adjusted. The monitoring parameters of the components used to monitor personnel entry and exit and the components used to monitor equipment start and stop in the disturbance sensing module are adjusted respectively. The adjustment of the personnel entry and exit component includes monitoring sensitivity and reporting threshold, and the adjustment of the equipment start and stop component includes status judgment conditions and response delay. The adjustment strategy follows this principle: when the surgery is in progress, data is collected from the pressure and temperature / humidity sensors in the core area using a first sampling frequency; when the postoperative cleaning period is in progress, data is collected from the same batch of sensors using a second sampling frequency, with the first sampling frequency being higher than the second sampling frequency.

4. The operating room purification status prediction and early warning system based on multi-source data fusion according to claim 1, characterized in that, The data processing module includes: The synchronization alignment module is configured to receive asynchronous data streams from the data acquisition module, perform clock correction and sampling point alignment on environmental data and environmental disturbance data based on a unified time reference, and generate a multi-source data sequence with a unified time axis. The missing data processing module is configured to identify and differentiate missing values ​​in multi-source data sequences. For random missing values ​​caused by intermittent failures of regular sensors, it uses linear interpolation based on adjacent valid data to fill in the missing data. For long-term continuous missing values ​​caused by system maintenance or equipment replacement, it uses the most relevant historical data of the same type from the same period to fill in the missing data and adds a missing data label. The feature construction module is configured to parse environmental disturbance data and, based on a predefined rule base, mark the start and end timestamps and event types of disturbance events on a unified time axis. The time axis marked with disturbance events is integrated with the processed multi-source data sequence, and key features are extracted to construct the fused feature dataset. The key features include the statistical features of pressure and temperature and humidity within the current and historical windows, the encoding features of the current disturbance event type, and the time interval since the start of the last similar disturbance event.

5. The operating room purification status prediction and early warning system based on multi-source data fusion according to claim 1, characterized in that, The hybrid prediction module also includes: The feature allocation module is configured to divide the fused feature dataset into a first data subset and a second data subset and perform intelligent distribution based on the type and source of the data features; The time series analysis module receives a first data subset with continuous time series characteristics allocated by the feature allocation module. The first data subset includes at least historical and current pressure and temperature and humidity time series data. The machine learning module receives a second data subset representing discrete events assigned by the feature allocation module. The second data subset includes at least surgical scheduling, personnel entry and exit events, and equipment status change data. The weight allocation module is configured to dynamically determine the fusion weight of the first prediction sub-result and the second prediction sub-result based on the activity analysis of current and historical environmental disturbance data. Among them, the fusion weight is adjusted in reverse correlation based on the level of business disturbance in which the operating room is located.

6. The operating room purification status prediction and early warning system based on multi-source data fusion according to claim 5, characterized in that, The weight allocation module adjusts the fusion weights according to the level of service disturbance, specifically as follows: When environmental disturbance data analysis indicates that the operating room is at a high disturbance level, the weight allocation module increases the fusion weight of the second prediction sub-result. The high disturbance level is triggered by the following conditions: In the future, the system will be designed to detect intensive surgical schedules within a set time period, monitor personnel entry and exit frequencies exceeding thresholds in real time, and trigger planned or unplanned start-ups and shutdowns of high-power medical equipment. When the environmental disturbance data analysis indicates that the operating room is in a low-disturbance steady state, the weight allocation module increases the fusion weight of the first prediction sub-result. The low-disturbance steady state corresponds to a period in which there are no surgical appointments, little personnel activity, and no high-power equipment operating in the operating room.

7. The operating room purification status prediction and early warning system based on multi-source data fusion according to claim 1, characterized in that, The early warning control module includes: The graded early warning module is configured to pre-store early warning thresholds and alarm thresholds associated with the purification environment threshold. It compares the received dynamic prediction value with the early warning threshold and alarm threshold in real time. When the dynamic prediction value does not exceed the early warning threshold, the status is determined to be normal. When the dynamic prediction value exceeds the early warning threshold but does not reach the alarm threshold, a first-level early warning signal is generated and issued. When the dynamic prediction value reaches the alarm threshold, a second-level alarm signal is generated and issued. Among them, the first-level warning signal and the second-level alarm signal are converted into mutually distinguishable audio-visual alarm forms and structured status messages, and sent to the local monitoring terminal in the operating room and the remote central monitoring platform for display; The forward-looking control module is configured to generate forward-looking control instructions based on dynamic predicted values ​​and the current operating parameters of the air conditioning system. The forward-looking control instructions are used to instruct the air conditioning automatic control system to adjust the chilled water side parameters, supply air temperature, supply air volume and unit operating mode in advance before the dynamic predicted values ​​reach the warning threshold, in order to offset the predicted environmental fluctuations.

8. The operating room purification status prediction and early warning system based on multi-source data fusion according to claim 7, characterized in that, The forward-looking control instructions include intervention measures for the air conditioning automatic control system, including: If the dynamic forecast values ​​indicate that the pressure, temperature and humidity in the core area of ​​the operating room will change in a direction that deviates from the preset comfort range, the forward-looking control command specifically includes a coordinated adjustment strategy for the supply air temperature, supply air volume and unit operating frequency, which is used to proactively input a reverse environmental parameter to the operating room environment to compensate before the predicted deviation occurs. If the dynamic forecast value combined with environmental disturbance data analysis indicates that a known type of disturbance event will occur in the future, the forward-looking control instructions will further include a standard response plan pre-bound to the type of disturbance event. The standard response plan specifies the target operating mode and parameter benchmarks of the air conditioning unit before the event is expected to begin.

9. The operating room purification status prediction and early warning system based on multi-source data fusion according to claim 4, characterized in that, The feature construction module includes: Build submodules for: The environmental disturbance data is parsed based on a predefined rule base, and the event types are labeled and the event levels are divided. The event intensity value is calculated and normalized based on the event duration, scope of impact and event weight coefficient; the co-occurrence frequency of the two types of events within 30 minutes in historical data is statistically analyzed, normalized to obtain the event correlation degree, and an event correlation table is constructed. Using a unified timestamp as an index, construct a structured timeline of timestamp-event ID-event type-event level-intensity value-related event ID-related degree of association; Integration submodule, used for: Based on the structured timeline, different stages of the event are determined; processed multi-source data sequences corresponding to different stages of the event are obtained, and the change characteristics of each type of environmental data in the processed multi-source data sequences corresponding to different stages of the event are calculated to obtain the environmental data responsiveness to the event and normalize it; based on the absolute value of the responsiveness, the event hierarchy coefficient, and the event intensity value, event association weights are assigned to the processed multi-source data for each timestamp and normalized. ; in, The responsiveness of environmental data to events; The rate of change of data during the event; The rate of change of the baseline data before the event; This represents the event intensity value. An integrated data sequence of timestamp-environmental data value-event information-response rate-association weight is generated based on timestamp, environmental data value, event information, responsiveness, and correlation weight. The feature extraction submodule is used for: Based on multi-scale historical windows, the mean, variance, maximum, minimum, median, range, first-order difference mean, second-order difference mean, and number of consecutive increases / decreases of environmental data in the integrated data sequence are calculated. The event types in the integrated data sequence are one-hot encoded, the event levels are binary encoded, the instantaneous values ​​and 30-minute moving averages of the event intensity values ​​are extracted, and the time intervals from the previous similar events and first-level events, the related event type codes, the correlation values, and the average intensity of the related events are extracted. Calculate the instantaneous value of responsivity, 30-minute moving average, peak responsivity and its occurrence time in the integrated data sequence; the coupling value between responsivity and event intensity and the total coupling during the event; the difference between the current data value and the baseline value before the event; the percentage of offset; and the duration of offset; calculate the impact of preceding events and residual offset values. The fusion submodule is used for: Calculate the correlation coefficient between each feature and the event type and the trend of environmental data change, and filter features with correlation coefficients greater than or equal to a preset correlation coefficient threshold to obtain the first feature set; measure the similarity between feature vectors in the first feature set based on cosine similarity, and filter features with similarity coefficients less than or equal to a preset similarity coefficient threshold to obtain the second feature set; divide the features in the second feature set into core features and auxiliary features, add source labeling, validity labeling, and applicable scenario labeling to each feature, and organize the data based on the format of core feature-auxiliary feature-event metadata-data quality label using timestamp as index to generate a fused feature dataset.

10. The operating room purification status prediction and early warning system based on multi-source data fusion as described in claim 7, characterized in that, The forward-looking regulation module includes: The preprocessing submodule is used for: The dynamic forecast values ​​are divided into three levels according to the urgency of time: emergency control period, regular control period, and preparatory control period. The deviation difference and deviation rate between the temperature and humidity forecast values ​​and the purification environment threshold are calculated for each period, and the corresponding real-time trend type is marked. Collect real-time operating parameters of the air conditioning system and determine the real-time status of the unit by combining the rated parameter range of the unit. Acquire real-time environmental disturbance data to determine the real-time disturbance scenario; Real-time trend types, real-time unit status, and real-time disturbance scenarios are used as real-time scenario data. A submodule is constructed to acquire historical data on the control of the purification status of the operating room. Based on the historical data, a dynamic strategy pool is constructed for trend type, unit status, and disturbance scenario. Each strategy subset contains core adjustment parameters, auxiliary adjustment parameters, adjustment range, and execution constraints. The filtering submodule is used for: Select a subset of policies from the dynamic policy pool that match the real-time data of the scenario. If there are multiple matching subsets, determine the optimal target policy through the first dual-objective optimization mechanism. When no policy subset is matched, the policy adaptive supplementation mechanism is triggered, which generates a temporary control policy based on historical successful control cases. The decomposition submodule is used to break down the optimal target strategy or temporary control strategy into specific instruction entries for each time period based on three control periods: emergency, normal, and preparation. These initial time-segment instruction entries are divided into core parameter instructions and auxiliary parameter instructions. Each initial time-segment instruction entry includes the parameter adjustment type, the target adjustment value, the execution start time, and the execution duration. The encoding submodule is used to encode the initial time-segmented instruction entries with three levels of priority based on the urgency of the control period. The first priority is to encode all parameter adjustment instructions during the emergency period, the second priority is to encode all instruction entries during the regular control period, and the third priority is to encode all parameter adjustment instructions during the preparation period, thus obtaining the encoded initial time-segmented instruction entries. The simulation submodule is used to construct a virtual operation scenario consistent with the actual scenario based on the current air conditioning system operating parameters and environmental disturbance information. The encoded initial time-segmented instruction entries are input into the virtual operation scenario to simulate the change trajectory of air conditioning system operating parameters and temperature and humidity in the core area of ​​the operating room during the execution of the encoded initial time-segmented instruction entries. The verification submodule is used to perform load safety verification and control effect verification on the operating parameters of the air conditioning system and the temperature and humidity change trajectory of the core area of ​​the operating room during the execution of the instruction. When it is determined that the load safety verification result and the control effect verification result both meet the requirements, the encoded initial time-segmented instruction entry is used as a forward-looking control instruction.

Images