Electromechanical equipment control method and system for building intelligent system integration

Abstract

The invention relates to the technical field of monitoring systems, in particular to an electromechanical equipment control method and system for building intelligent system integration. Comprising the following steps: collecting operation parameters of electromechanical equipment and building environment parameters in real time; the operation parameters and the building environment parameters are transmitted to a central control unit through a communication network; the operation parameters of the central control unit and the building environment parameters are input into a preset evaluation model, and the operation state and abnormality diagnosis information of the equipment are output; generating a corresponding control strategy through an intelligent algorithm; the control command is converted into a control command through the central control unit and is sent to the execution mechanism through the communication network, and the execution mechanism regulates and controls the electromechanical equipment and sends a feedback signal to the central control unit. The multi-source sensor is used for collecting operation parameters and environment parameters in real time, and the data precision is guaranteed in combination with a self-calibration mechanism; meanwhile, data interference and leakage risks are reduced through encryption processing, and a high-quality data basis is provided for subsequent analysis.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of monitoring systems, and in particular to a mechanical and electrical equipment control method and system for building intelligent system integration. BACKGROUND

[0002] Mechanical and electrical equipment, commonly referred to as M&E equipment, refers to devices that provide various basic services and functions in a building environment, including heating, ventilation, and air conditioning (HVAC) systems, electrical systems, lighting devices, water supply and drainage systems, elevators, and escalators. Such equipment plays a crucial role in modern buildings, not only providing comfort and functionality, but also involving energy use and safety management.

[0003] In existing building intelligent systems, there are key technical problems in the operation state monitoring and control of mechanical and electrical equipment (such as HVAC, electrical systems, lighting devices, etc.). Due to the lack of accurate processing and intelligent evaluation mechanisms for multi-source sensor data, it is difficult to quickly and accurately identify the abnormal state of the equipment and generate targeted control strategies. Specifically, the accuracy is insufficient due to interference during data transmission, the diagnosis of equipment abnormalities relies on human experience or a single model, and the generation and execution of control strategies lack dynamic optimization and adjustment, resulting in delayed response to equipment failures, low operating efficiency, and serious impact on the stability and reliability of building intelligent systems.

[0004] Therefore, it is necessary to design a mechanical and electrical equipment control method and system for building intelligent system integration. SUMMARY

[0005] To address the technical deficiencies in the background art, the present application proposes a mechanical and electrical equipment control method and system for building intelligent system integration, which solves the above technical problems and meets actual needs. The specific technical solution is as follows: The mechanical and electrical equipment control method for building intelligent system integration includes the following steps: Based on multi-source sensors, real-time collection of operating parameters and building environment parameters of mechanical and electrical equipment; Encrypt the operating parameters and building environment parameters, and transmit them to the central control unit through a mixed wired and wireless communication network; The central control unit inputs the decrypted operating parameters and building environment parameters into the preset evaluation model, and outputs the operating state and abnormal diagnosis information of the equipment; Based on the operating state of the equipment, generate corresponding control strategies through intelligent algorithms; Based on the control strategy, the control instructions are converted by the central control unit and sent to the actuator through the communication network. Based on the control instructions, the actuator regulates the electromechanical equipment and sends feedback signals to the central control unit.

[0006] Further, the operating state of the equipment includes a normal state and a plurality of abnormal states, and the evaluation model sequentially performs the steps of data preprocessing, feature engineering processing, model evaluation calculation and result output after inputting the operating parameters and building environment parameters.

[0007] Further, the data preprocessing includes adaptive standardization and outlier correction, and the adaptive standardization specifically includes: The operating parameters and building environment parameters are adaptively standardized by calculating the mean and standard deviation through a sliding time window; The outlier correction specifically includes: The standardized operating parameters and building environment parameters are detected for outliers by an isolation forest model, and when an outlier data point is detected, a cubic spline interpolation method is used for correction.

[0008] Further, the feature engineering processing includes feature extraction and feature selection and dimensionality reduction, and the feature extraction specifically includes: The time domain statistical features and frequency domain harmonic features are extracted from the operating parameters and building environment parameters after data preprocessing, the time domain statistical features include mean, variance and peak value, and the frequency domain harmonic features include amplitude, phase and energy distribution characteristics; The feature selection and dimensionality reduction specifically include: Based on the extracted time domain statistical features and frequency domain harmonic features, a combination strategy of mutual information and multi-class feature weight selection algorithm is used to screen features, and a principal component analysis algorithm is used for dimensionality reduction to output the feature vector after dimensionality reduction.

[0009] Further, the model evaluation calculation specifically includes: The feature vector after feature engineering processing is input into a pre-set evaluation model to output an equipment state probability distribution, a multi-level confidence judgment mechanism is used to determine the equipment operating state, and when the equipment state probability distribution is less than a pre-set value, a standby model is started for voting.

[0010] Further, the result output specifically includes: The operating state of the equipment is determined by the central control unit, and when the operating state of the equipment is in an abnormal state, an abnormal diagnosis process is started; The key parameters with the highest correlation degree to the abnormal state are extracted from the operation parameters and the building environment parameters, the extracted key parameters are compared with the typical feature mode of the corresponding abnormal state type in the preset abnormal feature database in multiple dimensions, and the deviation amplitude of the actual value of the parameter from the normal threshold value is calculated; The contribution degree of each key parameter to the current abnormal state is determined by using the preset feature weight matrix. The abnormal diagnosis information is generated in combination with the type of the abnormal state, the deviation amplitude and the contribution degree of the key parameters.

[0011] Further, the multi-source sensor includes sensors distributed in each electromechanical equipment and environmental sensors, and the multi-source sensor adopts a self-calibration sensor that performs self-calibration once every preset time period. During the calibration process, the multi-source sensor collects standard signals of a standard signal source, compares the collected signals with the standard signals, and automatically adjusts the internal parameters of the sensor if the deviation exceeds the allowable range.

[0012] Further, the generation of the corresponding control strategy by the intelligent algorithm specifically includes: The running state of the equipment is determined, and when the equipment is in an abnormal state, the fault knowledge base is queried based on the pre-established fault knowledge base. If the control strategy corresponding to the type of the abnormal state exists in the fault knowledge base, the control strategy is directly called. If the control strategy corresponding to the type of the abnormal state does not exist in the fault knowledge base, the intelligent algorithm is started to generate a strategy. After the control strategy is called or a new control strategy is generated, the strategy is optimized and adjusted according to the current building environment parameters, and the optimized control strategy is output.

[0013] Further, the central control unit evaluates the effect of the regulation and control of the actuator after receiving the feedback signal, and the effect evaluation specifically includes: The running parameters and the building environment parameters before and after the regulation and control are input into a preset effect model, and a regulation and control effect evaluation value is output. When the regulation and control effect evaluation value is less than a preset value, it is determined that the regulation and control effect is good. When the regulation and control effect evaluation value is greater than or equal to the preset value, a control strategy is regenerated for regulation and control until the regulation and control effect evaluation value is less than the preset value. When the number of regulation and control is greater than a preset number, an alarm information is sent to the system administrator.

[0014] The electromechanical equipment control system integrated by the building intelligent system includes a processor and a memory for storing executable instructions of the processor; The processor is configured to execute the instructions to implement the steps of the electromechanical equipment control method integrated by the building intelligent system.

[0015] Compared with the prior art, the electromechanical equipment control method and system integrated by the building intelligent system provided by the present application have the following beneficial effects: The application adopts multi-source sensors (including equipment sensors and environment sensors) to collect running parameters and environment parameters in real time, combines a self-calibration mechanism (periodically compares standard signals and automatically adjusts parameters) to ensure data accuracy, and simultaneously reduces data interference and leakage risks through encryption processing, wired and wireless hybrid communication network transmission, thereby providing a high-quality data basis for subsequent analysis.

[0016] The application dynamically optimizes by calling a fault knowledge base or generating a new strategy through intelligent algorithms based on the equipment running state, and combining real-time environment parameters; feedback signals after the execution mechanism control can trigger effect evaluation, and if the control is not good, a new strategy is generated, forming a closed-loop control to ensure that the equipment quickly recovers to stable operation in an abnormal state, and to improve system response speed and control accuracy.

[0017] The application improves data quality by self-adaptive standardization and isolated forest model correction of abnormal values, and then optimizes input through feature engineering, uses an evaluation model to output state probability distribution, and cooperates with multi-level confidence judgment and backup model voting mechanism to greatly improve abnormal state recognition accuracy and solve the one-sidedness problem of traditional diagnosis methods. BRIEF DESCRIPTION OF DRAWINGS

[0018] Figure 1 It is a flowchart of the control method of the mechanical and electrical equipment integrated in the building intelligent system in the application. DETAILED DESCRIPTION

[0019] In the description of the application, it should be understood that the terms "upper", "lower", "front", "back", "left", "right", "vertical", "horizontal", "middle", "inner" and the like indicate the orientation or positional relationship based on the orientation or positional relationship shown in the drawings, and are only for the convenience of describing the application and simplifying the description, and therefore cannot be understood as indicating or implying that the devices or elements indicated must have a particular orientation, be constructed and operated in a particular orientation, and therefore cannot be understood as limiting the application. In addition, the terms "first", "second" and the like are only for the purpose of description, and cannot be understood as indicating or implying relative importance or implicitly indicating the number of the technical features indicated. Therefore, the features limited by "first", "second" and the like can explicitly or implicitly include one or more of the features. In the description of the application, it should be noted that unless otherwise explicitly specified and limited, the terms "mounting", "connection", "connection" should be understood in a broad sense, for example, it can be fixedly connected, or it can be detachably connected, or integrally connected; it can be directly connected, or indirectly connected through an intermediate medium; it can be the communication between the two elements inside. For those skilled in the art, the specific meaning of the above terms in the application can be understood through specific circumstances.

[0020] The following describes the implementation of the present invention in conjunction with the accompanying drawings and relevant embodiments. The implementation of the present invention is not limited to the following embodiments, and the present invention relates to the relevant necessary components in this technical field, which should be regarded as the common knowledge in this technical field and can be known and mastered by technical personnel in this technical field.

[0021] See Figure 1 The present invention provides a method for controlling electromechanical equipment integrated with a building intelligent system, comprising the following steps: Step S100 : Based on multi-source sensors, operating parameters of electromechanical equipment and building environment parameters are collected in real time.

[0022] Multi-source sensors are physical sensing devices deployed within electromechanical equipment and the building environment, enabling data collection through a distributed network architecture. Exemplarily, these sensors include equipment sensors (e.g., current sensors, speed sensors, and pressure sensors) and environmental sensors (e.g., temperature and humidity sensors, light sensors, and air quality sensors). Operating parameters are physical quantities that directly reflect the operating status of the equipment, including voltage, current, speed, power, and temperature. For example, compressor current and fan speed for central air conditioners are examples of equipment operating parameters. Building environmental parameters are environmental indicators related to equipment function, including temperature, humidity, light intensity, and CO2 concentration. For example, indoor temperature is directly related to air conditioning system operation. Real-time data collection is achieved through the sensor network's continuous monitoring and data refresh mechanism. The data synchronization mechanism coordinates the sampling frequencies of each sensor, for example, high-frequency sampling (e.g., 10 times per second) for key equipment parameters and medium-frequency sampling (e.g., once per minute) for environmental parameters. This multi-dimensional data collaborative collection model overcomes the limitations of traditional control methods that focus solely on a single parameter. For example, in lighting system control, simultaneous collection of lamp current, voltage, and ambient light intensity can be used to comprehensively determine artificial lighting needs.

[0023] In step S200 , the operating parameters and building environment parameters are encrypted and transmitted to the central control unit via a hybrid wired and wireless communication network.

[0024] The encryption processing is a data protection operation based on a symmetric key algorithm. For example, the AES encryption algorithm is used to divide the plaintext data into fixed-length blocks, and the ciphertext is generated through multiple rounds of permutation, substitution and shuffle operations. The wired communication network is an industrial Ethernet or a fiber network, which is suitable for transmitting high-frequency and high-precision data. The wireless communication network is a 5G technology, which is suitable for flexible deployment of distributed sensor nodes. The hybrid communication network realizes data interconnection through a protocol converter. For example, the data of the wireless sensor network is converted into an Ethernet protocol through a gateway and then connected to the central control unit. Through the combination of encryption algorithms and hybrid network architecture, the stability of the transmission of key data is ensured, and the conditions of wide distribution of equipment and difficult wiring in the building environment are adapted. At the same time, the risk of data tampering and theft is prevented, for example, to avoid misleading the air conditioning system operation by malicious attackers through tampering with temperature data.

[0025] In step S300, the central control unit inputs the decrypted operating parameters and building environment parameters into a preset evaluation model, and outputs the operating state of the equipment and abnormal diagnosis information.

[0026] The evaluation model adopts a hybrid neural network architecture combining deep belief networks and convolutional neural networks. The deep belief network is stacked by multiple restricted Boltzmann machines (RBM) and is responsible for layer-by-layer feature learning and abstraction of the input multidimensional parameters. The convolutional neural network includes convolutional layers, pooling layers and fully connected layers, and focuses on capturing local correlation features and dynamic change rules in parameter sequences. This hybrid architecture not only retains the feature dimension reduction capability of deep belief networks for high-dimensional data, but also takes advantage of the extraction of convolutional neural networks for time series data and spatial distribution features, and can effectively process complex high-dimensional data sets composed of mechanical and electrical equipment operating parameters and building environment parameters. The input layer dimension of the model is dynamically set according to the actual number of parameters collected. Each input neuron corresponds to a standardized parameter value, including equipment voltage, current, speed, etc. operating parameters and temperature, humidity, light intensity, etc. environmental parameters. The input layer is connected to the visible layer of the first restricted Boltzmann machine, and the non-linear mapping of the features is realized through layer-by-layer training. The convolutional layer uses a 3x3 convolutional kernel to perform sliding window calculation on the parameter matrix after feature mapping, extracting local correlation features between different parameters, such as the coupling relationship between voltage fluctuation and current change, and the correlation features between environmental temperature and equipment heat dissipation efficiency. The pooling layer uses the maximum pooling strategy to downsample the convolution results, retaining key features while reducing data dimension and reducing model computation. The fully connected layer fuses the feature vectors after pooling with the abstract features output by the deep belief network, realizes high-level combination of features through a non-linear activation function, and the final output layer uses a softmax activation function to output the classification probability distribution of the equipment operating state, including normal operating state and multiple abnormal state categories.

[0027] Step S400, based on the running state of the equipment, the corresponding control strategy is generated through an intelligent algorithm.

[0028] The intelligent algorithm is an optimization tool of genetic algorithm and particle swarm optimization. Illustratively, the genetic algorithm can search for fan speed combinations to balance energy consumption and cooling demand, and the particle swarm optimization can adjust the water pump frequency to maintain stable water pressure. The control strategy is a solution set of a multi-objective optimization function, for example, to restore the environmental temperature to 22 DEG C while constraining the equipment current to be no more than 110% of the rated value. The strategy generation process is an iterative calculation of candidate strategies and evaluation of their energy consumption, recovery time and equipment loss indicators. For example, when the air conditioning system is insufficiently refrigerated, a scheme of increasing the compressor frequency to 90% or starting the standby fan is generated. By dynamic optimization instead of traditional fixed rules, such as avoiding energy waste caused by running equipment at full speed due to high temperature, precise control under partial load demand is achieved.

[0029] Step S500, based on the control strategy, the control instruction is converted through the central control unit and sent to the actuator through the communication network. Based on the control instruction, the actuator regulates and controls the electromechanical equipment and sends a feedback signal to the central control unit.

[0030] The control instruction is a protocol signal recognizable by the equipment, for example, the optimized fan speed parameter is sent to the frequency converter through the Modbus protocol. The actuator is a frequency converter, relay or valve driver that adjusts the equipment parameters in real time according to the instructions. The feedback signal is the measured parameter value and state flag of the equipment, which is used to verify the control effect. The central control unit compares the feedback value with the target value, and if the deviation exceeds the allowed range, the strategy is regenerated. Through continuous feedback and iterative adjustment, a complete cycle of "evaluating strategy, executing feedback" is formed. For example, when the water pump pressure is insufficient, multiple strategy iterations are performed until the pressure is restored, while the equipment state after troubleshooting is continuously monitored.

[0031] The present application can achieve the technical effects of improving the comprehensiveness of equipment running state perception, fault response speed, energy efficiency level and system self-adaptability, while reducing the need for manual intervention and operation and maintenance costs.

[0032] In an embodiment of the present application, the running state of the equipment includes a normal state and a plurality of abnormal states, and the evaluation model sequentially performs the steps of data preprocessing, feature engineering processing, model evaluation calculation and result output after inputting the running parameters and building environment parameters.

[0033] It should be noted that the data preprocessing includes adaptive standardization and outlier correction, the adaptive standardization specifically includes: the running parameters and building environment parameters are adaptively standardized by calculating the mean and standard deviation through a sliding time window; the outlier correction specifically includes: the standardized running parameters and building environment parameters are detected for outliers by an isolation forest model, when detecting the data points of outliers, the correction is performed by using a cubic spline interpolation method.

[0034] The sliding time window is a time interval division method for dynamically adjusting the data reference range, the statistical characteristic values of the parameters in a specific time range can be obtained by real-time rolling calculation, and the sliding time window includes time periods with different lengths of 5 seconds or 60 seconds. The running parameters are high-frequency change data generated during the operation of the electromechanical equipment. The building environment parameters are slowly changing data reflecting the state of the internal and external environment of the building. The mean and standard deviation are statistical quantities describing the data distribution characteristics, which are calculated by the parameter values in the sliding time window. The mean and standard deviation can be used to represent the central tendency and dispersion of the parameters in a specific time range. The adaptive standardization is a dynamic processing process of converting the original data into a standard normal distribution.

[0035] The isolation forest model is an outlier detection algorithm based on tree structure, which identifies low-density sample regions in the data space by random partitioning. The cubic spline interpolation method is a numerical analysis method for constructing a smooth curve using a polynomial function, which determines the cubic polynomial coefficients by four adjacent data points to generate a continuous and differentiable interpolation curve. The outlier correction is a numerical compensation process for the detected abnormal data points, which constructs a cubic spline function by extracting two normal data points before and after the abnormal point, and calculates the corrected parameter value. In one specific embodiment, when the sensor generates a transient current spike due to electromagnetic interference, the isolation forest model can identify the abnormal point, the system extracts four normal data points before and after the abnormal point, generates a smooth curve by cubic spline interpolation, and replaces the abnormal value with the corrected value that conforms to the data trend.

[0036] By dynamically adjusting the length of the sliding time window, the mean and standard deviation of the parameters in the last one hour are calculated in real time, and the data is converted by the standardization formula, so that the reference value is dynamically updated with the equipment running state and environmental changes, for example, the reference value of the temperature parameter can reflect the environmental changes in time when the seasons change. This process avoids the misjudgment that may be caused by the static threshold, while maintaining the normality characteristics of the data distribution. For outlier correction, the combination of the isolation forest model and the cubic spline interpolation is as follows: after detecting the abnormal data point, two normal data points before and after the abnormal point are extracted, a smooth curve is constructed by a cubic spline function, and the corrected parameter value is calculated. This method not only eliminates the noise caused by interference, but also retains the overall trend characteristics of the data, for example, the current spike is corrected to a smooth value that conforms to the time series rule.

[0037] The present application realizes adaptive standardization by calculating the mean and standard deviation of dynamic sliding time window parameters, detects abnormal values by combining the Isolation Forest model, and corrects data by using the cubic spline interpolation method, so as to achieve the technical effects of improving the environmental adaptability of data preprocessing and data quality. Specifically, the dynamic benchmark updating mechanism effectively solves the misjudgment problem of static threshold when the environment changes, and the abnormal value correction method based on polynomial interpolation eliminates noise while retaining data trend characteristics, thereby providing a more reliable data basis for subsequent control algorithms and enhancing the stability and accuracy of the system in complex environments.

[0038] It should be noted that the feature engineering processing includes feature extraction and feature selection and dimension reduction, and the feature extraction specifically includes: Time domain statistical features and frequency domain harmonic features are extracted from the running parameters and building environment parameters after data preprocessing, the time domain statistical features include mean, variance and peak value, and the frequency domain harmonic features include amplitude, phase and energy distribution characteristics; Among them, the time domain statistical feature is a basic statistical quantity describing the instantaneous fluctuation characteristics of the parameter, which can be obtained by mathematical operation on the preprocessed time series data. The mean is the arithmetic mean of the parameter in a certain time window, for example, the mean of the current signal represents the average load of the equipment; the variance is the degree of dispersion of the parameter from the mean, reflecting the degree of load fluctuation; the peak value is the maximum absolute value of the parameter in the time domain, for example, the current peak value indicates the instantaneous maximum current value. The frequency domain harmonic feature is the energy distribution and phase relationship of the signal obtained by frequency spectrum analysis, which can be realized by fast Fourier transform. For example, the amplitude can be the amplitude intensity of a certain frequency component, such as the difference between the fundamental wave and harmonic amplitude of motor current, which can reflect the winding fault; the phase is the phase difference between different parameters or different frequency components of the same parameter, for example, the abnormal phase difference between voltage and current indicates the decrease of power factor; the energy distribution characteristic can be the energy ratio of the signal in the frequency band, for example, the sudden increase of certain harmonic energy may correspond to bearing wear.

[0039] The feature selection and dimension reduction specifically includes: Based on the extracted time domain statistical features and frequency domain harmonic features, a combination strategy of mutual information and multi-class feature weight selection algorithm is used to screen features, and then principal component analysis algorithm is used for dimension reduction to output the dimension-reduced feature vector.

[0040] Mutual information is a quantitative indicator of non-linear correlation between features and equipment states, which realizes feature selection by calculating conditional independence of information entropy. Mutual information can identify features strongly related to abnormal equipment states, such as sensitive parameters of peak current and temperature variance to overload failure. Multi-class feature weight selection algorithm is a model for evaluating the ability of features to distinguish multiple categories of states, including but not limited to linear discriminant analysis or feature importance evaluation based on random forest. This algorithm can calculate the distinguishing weight of features for multiple faults such as "bearing wear" and "blade jamming", for example, retaining the peak speed instead of the vibration frequency amplitude to optimize the distinguishing degree. Principal component analysis algorithm is a dimension reduction method that realizes orthogonal transformation of features by covariance matrix decomposition, which can retain principal components with cumulative variance contribution rate exceeding 95%, for example, compressing 20 original features to 5 principal components to improve the calculation efficiency of hybrid neural network.

[0041] The present application quantifies the instantaneous fluctuations and spectral characteristics of equipment operating states by extracting time domain statistical features and frequency domain harmonic features from preprocessed data, optimizes feature distinguishing ability by combining mutual information screening and multi-class weight algorithm, and reduces feature redundancy, improves calculation efficiency and enhances model generalization ability through principal component analysis compression. Specifically, the combination of basic statistical features and core harmonic features reduces the risk of dimension explosion while ensuring information integrity, and the multi-algorithm combination strategy is more accurate in identifying key features of multiple categories of states than single method. The final reduced feature vector not only retains key information but also reduces model training complexity, for example, in an air conditioning system, irrelevant environmental parameters can be eliminated and core features such as temperature amplitude and current phase difference can be retained, which can improve the calculation efficiency of the state evaluation model by more than 30%.

[0042] It should be noted that the model evaluation calculation specifically includes: inputting the feature vector processed by feature engineering into the preset evaluation model, outputting the probability distribution of equipment state, determining the equipment operating state through a multi-level confidence judgment mechanism, and starting the standby model voting when the probability distribution of equipment state is less than the preset value.

[0043] Among them, the feature vector is a set of equipment operating parameters after data cleaning, standardization or dimension conversion, obtained through feature selection and dimension reduction in feature engineering, and the feature vector includes standardized values or statistical derivative indicators of device temperature, vibration frequency, current intensity sensor data.

[0044] The probability distribution of equipment state is a multi-dimensional probability vector output by the model, each dimension of which corresponds to the probability value of the equipment being in a normal, abnormal or fault state. For example, the probability distribution can be represented as [0.7, 0.2, 0.1], corresponding to the probabilities of normal, slight abnormal and severe failure, respectively.

[0045] The multi-level confidence judgment mechanism is a hierarchical decision logic based on threshold demarcation points, and the judgment conditions include but are not limited to comparison of the maximum probability value or the confidence mean value in the probability distribution, when the maximum probability value is lower than 0.85, the first level judgment is triggered, and when the maximum probability value is lower than 0.8, the second level backup process is triggered.

[0046] By inputting the feature vector into the evaluation model and outputting the probability distribution, the preliminary classification of the device state can be realized; by determining the running state through the multi-level confidence judgment mechanism, the size relationship between the maximum value of the probability distribution and the preset threshold can be compared, and then it is determined whether to start the backup model voting process, so that the technical effect of introducing a redundant judgment mechanism to improve the reliability of decision-making when the model output is highly uncertain can be achieved.

[0047] The application can achieve the technical effect of reducing the risk of misjudgment through parallel classification process when there is uncertainty in single model output, while keeping the original model architecture and algorithm logic unchanged, by inputting the feature vector processed by feature engineering into the mixed neural network evaluation model to output the device state probability distribution, and combining the multi-level confidence judgment mechanism to compare the probability distribution with the hierarchical threshold value, and starting the backup model voting process when the judgment result does not reach the preset confidence.

[0048] It should be noted that the result output specifically includes: The running state of the device is determined by the central control unit, and when the running state of the device is in an abnormal state, an abnormal diagnosis process is started.

[0049] The central control unit is a module responsible for monitoring and analyzing the running state of the device, and its determination function can be realized by comparing the real-time collected device parameters with the preset threshold value, and the determination of the running state of the device includes but is not limited to comprehensive evaluation through sensor data, device self-checking signals or model prediction results. When the determination result meets the preset abnormal condition, for example, the parameter exceeds the tolerance range or the logic contradiction, the system will trigger the subsequent diagnosis process.

[0050] The key parameters with the highest correlation degree to the abnormal state are extracted from the running parameters and the building environment parameters, the extracted key parameters are compared with the typical feature patterns of the corresponding abnormal state type in the preset abnormal feature database in multiple dimensions, and the deviation amplitude of the actual value of the parameter and the normal threshold value is calculated.

[0051] The abnormal feature database is a structured knowledge base for storing typical parameter combinations of various abnormal states, and its construction methods include but are not limited to training based on historical fault data or case library provided by equipment manufacturers. The typical feature mode is a parameter feature set corresponding to a specific abnormal state, for example, the combination mode of the current peak value exceeding the limit and the temperature rise rate abnormality corresponding to "motor overload". Multi-dimensional comparison is to match the actual parameter value with the multiple dimensions of the feature mode in the database, including calculating the absolute deviation and relative deviation of the parameter deviation from the normal threshold.

[0052] The contribution of each key parameter to the current abnormal state is determined by using a preset feature weight matrix.

[0053] The feature weight matrix is a numerical table quantifying the importance of parameters, and its construction methods include but are not limited to analytic hierarchy process or principal component analysis. The determination of weight values can be based on the sensitivity of parameters to abnormal states or the statistical rules of historical fault cases. The contribution degree is the influence degree of the parameter deviation from the normal value on the abnormal state, for example, in the "overheating protection" abnormality, the weight of the environmental temperature may be 0.7, and the weight of the cooling water flow is 0.3, indicating that the environmental temperature has a more significant influence on the abnormal state.

[0054] Abnormal diagnosis information is generated in combination with the type of abnormal state, the deviation amplitude of key parameters and the contribution degree.

[0055] The abnormal diagnosis information is a structured fault description. If it is determined that "the compressor is overheating", the diagnosis information can clearly indicate that the main reasons are the excessive environmental temperature and the insufficient cooling fan speed. This process can convert the fuzzy abnormality classification into traceable fault positioning basis through weighted calculation of parameter deviation and weight.

[0056] The present application determines the abnormal state of the equipment in real time through the central control unit and triggers the diagnosis process, realizes multi-dimensional comparison of parameters and typical fault modes in combination with the abnormal feature database, quantifies the contribution of parameters by using the feature weight matrix, and finally integrates the abnormal type, deviation amplitude and weight to generate structured diagnosis information. It can achieve the technical effect of converting traditional abnormal classification into quantitative diagnosis of specific parameter deviation, improving fault positioning accuracy and interpretability, reducing manual troubleshooting time, and providing data support for subsequent targeted optimization of control strategies.

[0057] In an embodiment of the present application, the multi-source sensor includes sensors distributed in each electromechanical equipment and environmental sensors, and the multi-source sensor uses a self-calibration sensor that performs self-calibration every preset time period. During the calibration process, the multi-source sensor collects standard signals of a standard signal source, compares the collected signals with the standard signals, and automatically adjusts the internal parameters of the sensor if the deviation exceeds the allowed range.

[0058] The multi-source sensor is a physical quantity sensing device distributed in the electromechanical equipment and the environment, and is realized by the electromechanical equipment body or external environment deployment. The sensors distributed in the electromechanical equipment include temperature, vibration, and current equipment state monitoring sensors. The environmental sensors can include environmental parameter monitoring sensors such as temperature and humidity and illumination intensity. The self-calibration sensor is an intelligent sensing unit with built-in calibration reference and parameter correction function, which can be realized by combining hardware circuit and software algorithm. The functional modules include a standard signal generation circuit, a deviation calculation unit, and a parameter adjustment executor. The standard signal source is a high-precision reference signal generation device for providing a calibration reference, which is realized by a physical module accurately matched with the measurement range of the sensor. The standard signal source of the temperature sensor is a temperature control module with preset temperature points of 0℃, 25℃, and 50℃. The standard signal source of the voltage sensor can be a reference electrical signal generator outputting 50%, 100%, and 120% of the rated voltage value.

[0059] The deviation calculation is a process of determining the calibration error by the difference between the measured value and the standard value, which can be realized by the sensor signal acquisition circuit and the built-in calculation module. When the temperature sensor measures 25.1℃ under the standard signal of 25℃, the deviation Δ is +0.1℃. The allowable range is the error threshold interval corresponding to the accuracy level of the sensor, which can be set in the form of range percentage or absolute value. The allowable deviation of the 0100℃ temperature sensor is ±0.5℃. The internal parameter adjustment is an operation of eliminating errors by modifying the sensor signal processing circuit or compensating algorithm parameters, which can be realized by hardware circuit gain adjustment or software coefficient correction. The voltage sensor can reduce the reading deviation by increasing the gain of the amplification circuit, and the temperature sensor can offset the influence of environmental temperature fluctuations by adjusting the thermistor compensation coefficient.

[0060] The periodic calibration execution is an automatic calibration process started according to a preset time interval, which is realized by combining the embedded clock module with the control logic. The calibration period can be set to be executed every 24 hours in the low load period in the early morning. The fault tolerance mechanism is a redundant safeguard measure started after detecting calibration failure, which can be realized by continuous multiple calibration failure judgment and standby channel switching logic. For example, when the calibration deviation exceeds the limit for three consecutive times, a fault alarm is triggered and the redundant channel of the adjacent sensor is enabled. The closed-loop calibration loop can be an automatic control loop composed of standard signal acquisition, deviation calculation, and parameter correction. It can be realized by the physical connection of the sensor measurement channel and the calibration reference and the algorithm closed loop. The standard signal is directly input into the calibration closed loop through the measurement circuit of the sensor.

[0061] The application realizes comprehensive perception of equipment and environmental parameters through distributed deployment of multi-source sensors, constructs a closed-loop calibration benchmark through a self-calibration sensor built-in standard signal source, realizes dynamic error correction through periodic calibration and deviation comparison, guarantees data acquisition continuity through parameter adjustment and redundancy switching, and can achieve the technical effects of significantly reducing long-term operation error, improving data acquisition stability, and reducing the frequency of manual maintenance. The scheme controls long-term data error within ±0.3% through high-frequency automatic calibration, improves data reliability compared with the traditional manual calibration method, and ensures system reliability through fault tolerance design, thereby providing a high-precision data basis for mechanical and electrical equipment state evaluation and intelligent control.

[0062] In an embodiment of the application, the generation of the corresponding control strategy through the intelligent algorithm specifically includes: determining the running state of the equipment, querying the pre-established fault knowledge base when the equipment is in an abnormal state, directly calling the control strategy if the fault knowledge base contains a control strategy corresponding to the type of the abnormal state, starting the intelligent algorithm to generate a strategy if the fault knowledge base does not contain a control strategy corresponding to the type of the abnormal state, and optimizing and adjusting the strategy according to the current building environmental parameters after calling the control strategy or generating a new control strategy, and outputting the optimized control strategy.

[0063] The fault knowledge base is a structured database that stores abnormal type labels, control parameter adjustment thresholds, and environmental parameter adaptation ranges, and is constructed through classification and feature vectorization of historical fault cases. For example, the fault knowledge base contains a voltage limit threshold corresponding to "motor overload" and a specific parameter of a fan speed regulation ratio corresponding to "filter clogging". The intelligent algorithm generation is a combined optimization process of genetic algorithm and particle swarm optimization algorithm. The genetic algorithm uses real number coding, the initial population size is set to 100 individuals, the selection operation uses the tournament selection method to retain the top 30% of individuals in terms of fitness, the crossover operation uses an arithmetic crossover probability of 0.7, and the mutation operation uses a Gaussian mutation probability that linearly decreases from 0.05 to 0.01. The particle swarm optimization algorithm updates particle speed and position through parameter configuration of inertia weight and learning factor. For example, for the "poor heat dissipation" abnormality, the particle swarm optimization algorithm can iteratively optimize the parameter combination of the heat dissipation fan speed and the cooling medium flow rate. The environmental parameter optimization is the joint application of a dynamic adjustment model and a multiple linear regression method.

[0064] The fault knowledge base query realizes abnormal type matching through cosine similarity-based feature vector comparison, for example, the current abnormality current peak value, temperature change rate characteristic parameter and the feature template of the abnormal type label in the knowledge base are projected in the vector space, when the similarity exceeds the 0.85 threshold, a preset strategy is directly called, for example, for the "filter blockage" abnormality, a control scheme of increasing the fan rotating speed by 20% and prolonging the cleaning reminder cycle is called. By reusing historical experience, the control response time of common abnormalities is shortened to 1 / 5 of that generated by the traditional algorithm. The parameterization configuration of the intelligent algorithm generation link balances the global search and local convergence ability through the collaborative optimization of the genetic algorithm and the PSO algorithm, for example, for the unknown abnormality "poor heat dissipation", the PSO algorithm can optimize the strategy parameter combination that can both reduce the equipment temperature to the safe range and consume the least power in 10 seconds. The environmental parameter optimization constructs a parameter correction equation through a dynamic equation and a multivariate linear regression model, automatically increases the power regulation coefficient of the refrigeration equipment in summer high temperature, solves the problem that the traditional fixed strategy cannot adapt to environmental changes. The feasibility verification ensures the safety of the strategy through the constraint condition library and the parameter correction mechanism, the constraint condition library includes hard limits such as the maximum current of the motor being 120% of the rated value, if the generated voltage adjustment value exceeds the safety threshold, the parameters are re-optimized through the gradient descent method or the constraint penalty function, for example, the compressor frequency is adjusted to a suboptimal value and the running time is prolonged, avoiding overloading of the equipment.

[0065] The present application triggers the strategy generation mechanism by the equipment running state determination, realizes the instant response of high-frequency abnormalities by concentrating the computing resources and quickly calling the fault knowledge base, ensures the global optimization ability of unknown abnormalities through the parameterization configuration of the intelligent algorithm, enhances the scene adaptability of the strategy through the environmental parameter optimization model, and filters unsafe control actions through the feasibility verification link. The technical effect of constructing a dynamic control system of "knowledge-driven + intelligent optimization + environment adaptation" can be achieved. The system significantly improves the equipment abnormality recovery efficiency, for example, when the fan bearing is worn, the speed reduction strategy is called through the knowledge base and the speed reduction amplitude is dynamically adjusted according to the real-time vibration frequency, which suppresses the abnormality while avoiding excessive speed reduction to affect the performance, and at the same time, the safety of the control action is ensured through the constraint condition library, which overall balances the rapid response of high-frequency abnormalities and the flexible handling ability of complex working conditions.

[0066] In an embodiment of the present application, the central control unit evaluates the effect of the regulation and control of the actuator after receiving the feedback signal, and the effect evaluation specifically includes: inputting the running parameters and building environment parameters before and after the regulation and control into a preset effect model, and outputting a regulation and control effect evaluation value; when the regulation and control effect evaluation value is less than a preset value, it is determined that the regulation and control effect is good; when the regulation and control effect evaluation value is greater than or equal to the preset value, a control strategy is regenerated for regulation and control until the regulation and control effect evaluation value is less than the preset value; when the regulation and control times are greater than a preset number of times, an alarm information is sent to the system administrator.

[0067] The preset effect model is a mathematical calculation framework for converting multi-dimensional parameter change amount into a single evaluation index, and realizes the comprehensive evaluation of parameter difference through a preset quantization formula. In one specific embodiment, the model calculates the regulation effect evaluation value by using the root mean square error formula. The model integrates the multi-parameter change amount into a comparable numerical index through the root mean square operation of the square difference.

[0068] The preset value is a threshold parameter set according to the device type and the abnormal feature, and the value is determined by weighted summation. The weight of each parameter is set according to the device characteristics, for example, the weight of the voltage parameter is 0.3, the weight of the temperature parameter is 0.5, and the weight of the rotating speed parameter is 0.2. The comprehensive threshold value is obtained by multiplying and adding the threshold values of each parameter and the weight. For example, when the temperature parameter threshold value is set to 2.25℃², its contribution to the total threshold value is 0.5*2.25=1.125. This threshold setting mechanism can dynamically adjust the parameter selection range according to the type of abnormality. For example, the fan abnormality focuses on the rotating speed and vibration parameters, and the pump abnormality focuses on the flow and pressure parameters.

[0069] The process of regenerating the control strategy is an iterative optimization process based on failure case analysis, and the maximum regulation number limit is used to prevent infinite loop. In one specific embodiment, when the regulation effect evaluation value exceeds the threshold value for three consecutive times, the system triggers the manual intervention mechanism. For example, due to the encoder failure of an elevator, the system sends a three-level alarm containing the device positioning, the type of abnormality and the recommended solution after three unsuccessful strategy adjustments. This mechanism improves the probability of solving complex abnormalities through multiple strategy iterations, and compared with the traditional single regulation method, the effectiveness of the strategy in the successful case library can be improved by more than 40%.

[0070] The application also provides a building intelligent system integrated electromechanical equipment control system, comprising a processor and a memory for storing executable instructions of the processor; wherein the processor is configured to execute the instructions to implement the steps of the building intelligent system integrated electromechanical equipment control method as described above.

[0071] The processor is a hardware unit for executing control logic operations, and realizes the functions of parameter acquisition, evaluation calculation and strategy generation through a preset instruction set. The processor uses an industrial-grade embedded microcontroller, and its core functions include real-time data processing and control strategy iteration. The memory is a non-volatile storage device for storing program code and historical data, and ensures information traceability through a structured data storage method. In one specific embodiment, the memory uses a solid state disk and is configured with a RAID1 redundant array, and the storage content is indexed according to the device identifier and the time stamp to form a traceable regulation file.

[0072] The above merely describes the preferred embodiments of the present application, and it should be pointed out that, for those skilled in the art, several improvements and refinements can be made without departing from the principles of the present application, and these improvements and refinements should also be considered as falling within the protection scope of the present application.

Claims

1. A method for controlling mechanical and electrical equipment integrated in a building intelligent system, characterized in that, The method comprises the following steps: Real-time collection of operation parameters and building environment parameters of the electromechanical equipment based on multi-source sensors; Encryption of the operation parameters and the building environment parameters, and transmission of the encrypted operation parameters and the building environment parameters to a central control unit through a mixed wired and wireless communication network; Input of the decrypted operation parameters and the building environment parameters into a preset evaluation model by the central control unit, and sequential data preprocessing, feature engineering processing, model evaluation calculation and result output, to output the operation state of the equipment and abnormal diagnosis information; Generation of a corresponding control strategy based on the operation state of the equipment through an intelligent algorithm; Conversion of the control strategy into control instructions by the central control unit, and transmission of the control instructions to an execution mechanism through the communication network, based on the control instructions, the execution mechanism regulates and controls the electromechanical equipment and sends feedback signals to the central control unit; The data preprocessing includes adaptive standardization and outlier correction, and the adaptive standardization specifically includes: Adaptive standardization of the operation parameters and the building environment parameters through sliding time window calculation of mean value and standard deviation.

2. The method of claim 1, wherein the method further comprises: The operation state of the equipment includes a normal state and multiple abnormal states.

3. The method of claim 2, wherein the method further comprises: The outlier correction specifically includes: Outlier detection on the standardized operation parameters and the building environment parameters through an isolation forest model, and correction of the data points of the detected outliers by using a cubic spline interpolation method.

4. The method of claim 3, wherein the method further comprises: The feature engineering processing includes feature extraction, feature selection and dimension reduction, and the feature extraction specifically includes: Extraction of time domain statistical features and frequency domain harmonic features from the data preprocessed operation parameters and the building environment parameters, the time domain statistical features include mean value, variance and peak value, and the frequency domain harmonic features include amplitude, phase and energy distribution characteristics; The feature selection and dimension reduction specifically include: Based on the extracted time domain statistical features and the frequency domain harmonic features, a combination strategy of mutual information and multi-class feature weight selection algorithm is used to screen the features, and a principal component analysis algorithm is used for dimension reduction to output a dimension-reduced feature vector.

5. The electromechanical equipment control method for building intelligent system integration according to claim 4, characterized in that: The model evaluation calculation specifically includes: Input of the feature vector after the feature engineering processing into a preset evaluation model, output of the equipment state probability distribution, determination of the equipment operation state through a multi-level confidence judgment mechanism, and start of a backup model voting when the equipment state probability distribution is less than a preset value.

6. The method of claim 5, wherein the method further comprises: The result output specifically includes: Determination of the operation state of the equipment by the central control unit, and start of an abnormal diagnosis process when the operation state of the equipment is in an abnormal state; Extraction of key parameters with the highest correlation degree to the abnormal state from the operation parameters and the building environment parameters, multi-dimensional comparison of the extracted key parameters with typical feature patterns of corresponding abnormal state types in a preset abnormal feature database, and calculation of the deviation amplitude of the actual values of the parameters from normal threshold values; Determination of the contribution degree of each key parameter to the current abnormal state by using a preset feature weight matrix; Generation of abnormal diagnosis information in combination with the type of the abnormal state, the deviation amplitude of the key parameters and the contribution degree.

7. The method of claim 1, wherein the method further comprises: The multi-source sensor includes sensors distributed in each electromechanical device and environmental sensors, and the multi-source sensor adopts a self-calibration sensor, which is self-calibrated every preset time period, and in the calibration process, the multi-source sensor collects standard signals of a standard signal source, compares the collected signals with the standard signals, and if the deviation exceeds the allowed range, automatically adjusts the internal parameters of the sensor.

8. The method of claim 2, wherein the method further comprises: The corresponding control strategy generated by the intelligent algorithm specifically includes: The running state of the device is determined, and when the device is in an abnormal state, the pre-established fault knowledge base is queried, and if the fault knowledge base has a control strategy corresponding to the type of the abnormal state, the control strategy is directly called; if the fault knowledge base does not have a control strategy corresponding to the type of the abnormal state, the intelligent algorithm is started to generate a strategy, and after the control strategy is called or a new control strategy is generated, the strategy is optimized and adjusted according to the current building environment parameters, and the optimized control strategy is output.

9. The method of claim 1, wherein the method further comprises: After receiving the feedback signal, the central control unit evaluates the effect of the control of the actuator, and the effect evaluation specifically includes: The running parameters and building environment parameters before and after the control are input into a preset effect model, and a control effect evaluation value is output, when the control effect evaluation value is less than a preset value, it is determined that the control effect is good; when the control effect evaluation value is greater than or equal to the preset value, a new control strategy is generated for control until the control effect evaluation value is less than the preset value, and when the control times are greater than a preset number, an alarm information is sent to the system administrator.

10. A building intelligent system integrated electromechanical equipment control system, characterized in that, It includes a processor and a memory for storing instructions executable by the processor; The processor is configured to execute the instructions to implement the steps of the building intelligent system integrated electromechanical device control method according to any one of claims 1 to 9.

Images