Energy consumption prediction method and system coping with compressor driving and medium

Abstract

The invention relates to the technical field of energy consumption prediction, in particular to an energy consumption prediction method and system for coping with compressor driving, and the method comprises the following steps: collecting the operation data of a compressor through a plurality of sensors, and carrying out the data processing of the collected data through the combination of weighted Bayesian fusion and the dynamic adjustment of the weight of each sensor; constructing an energy consumption prediction model through the features captured by the multi-layer capture mechanism; a feedback adjustment mechanism is formulated to perform real-time monitoring on energy consumption prediction errors of an output result of the energy consumption prediction model, parameters of the model are automatically adjusted according to monitored error feedback information, and the adaptability of the energy consumption prediction model to real-time working conditions is improved; through incremental learning and a feedback regulation mechanism, the energy consumption prediction system can adapt to the change of the working condition of the compressor in real time; an energy consumption prediction model constructed through a multi-layer capture mechanism can capture time sequence dependence from the forward direction and the reverse direction of data at the same time, and capture the long and short time dependence relation of compressor energy consumption data.

Description

Technical Field

[0001] The present invention relates to the technical field of energy consumption prediction, and specifically to an energy consumption prediction method, system and medium for a compressor drive. Background Art

[0002] Compressors are widely used in fields such as refrigeration, air conditioning, and compressed gas in industrial production, and their energy consumption is the key to industrial energy conservation and cost control. Traditional compressor energy consumption prediction methods rely on simple linear regression or experience-based modeling, and cannot accurately capture complex non-linear relationships and time series dependencies. In addition, the operating conditions of compressors change frequently, and existing models lack real-time adaptability and incremental learning ability, resulting in low prediction accuracy and inability to effectively respond to environmental changes. Summary of the Invention

[0003] In view of the above problems, the present invention is proposed.

[0004] To solve the above technical problems, the present invention provides the following technical solutions: An energy consumption prediction method for a compressor drive, comprising the following steps, Collect compressor operation data using multiple sensors, and judge whether the data collected by the sensors is distorted through a sensor fault self-detection algorithm. At the same time, combine weighted Bayesian fusion to dynamically adjust the weights of each sensor to process the collected data; Capture the features of the processed data through a multi-layer capture mechanism, and construct an energy consumption prediction model based on the captured features. Specifically: After the weighted fusion data undergoes a linear transformation, it is input into the basic layer, and the self-attention mechanism is used to calculate the correlation between the linearly transformed data at different time steps. A feature vector is constructed based on the calculated correlation, and the processed feature vector is input into a feed-forward network to output the results of the basic layer. The long short-term memory network is used to predict the compressor energy consumption based on the output results of the basic layer; Formulate a feedback adjustment mechanism to monitor the energy consumption prediction error of the output results of the energy consumption prediction model in real time, and automatically adjust the parameters of the model according to the monitored error feedback information to improve the adaptability of the energy consumption prediction model to real-time operating conditions.

[0005] As a preferred solution of the energy consumption prediction method for a compressor drive according to the present invention, wherein: The judgment of whether the data collected by the sensors is distorted through the sensor fault self-detection algorithm is as follows: Construct a joint vector containing all sensor data , then there is, , where, represents the compressor operating state data, represents the drive system parameters, represents the environmental parameters, Denote the constructed combined vector; Use Mahalanobis distance to evaluate whether the data of the sensor deviates from the normal distribution, then there is Wherein, Denote the mean vector of historical normal data, Denote the transpose matrix, Denote the covariance matrix, which is the correlation between sensor variables, and is solved based on the exponentially weighted moving average strategy. The specific solution process is as follows: Solve the covariance matrix according to the weighted average of new data and historical data, then there is Wherein, Denote the covariance matrix at the previous moment, Denote the weighting factor, which controls the influence of historical data on the current matrix. The value of the weighting factor is proportional to the influence degree of historical data on the current matrix. Denote the calculated Mahalanobis distance, which is used to evaluate whether the data of the sensor deviates from the normal distribution. Specifically: Set the data normal distribution threshold , if the calculated Mahalanobis distance and the set distribution threshold satisfy the formula , it means that the current data point is abnormal, and the data collected by the corresponding sensor is distorted data, and the sensor needs to be adjusted to collect data again; If the calculated Mahalanobis distance and the set distribution threshold satisfy the formula , it means that the current data point is not abnormal, and the data collected by the corresponding sensor is normal data.

[0006] As a preferred solution of the energy consumption prediction method for a compressor drive described in the present invention, wherein: the data processing of the collected data is specifically as follows: Set As the reading of the th sensor, and combine the confidence level evaluated by the Mahalanobis distance to define the dynamic weight of each sensor, then there is Wherein, Denote the Mahalanobis distance of the th sensor at the current moment, Denote the dynamic weight corresponding to the th sensor, Denote the Mahalanobis distance of the th sensor at the current moment; Based on the calculated dynamic weights, perform data fusion, then there is Among them, represents the dynamic weight corresponding to the th sensor, represents the reading of the th sensor, represents the data after weighted fusion.

[0007] As a preferred solution of the energy consumption prediction method driven by a compressor according to the present invention, wherein: the calculation of the correlation between the data after linear transformation through the self-attention mechanism is specifically as follows: According to the data after linear transformation generate the corresponding query, key, and value, then there is Set the data after linear transformation to satisfy the formula , where represents the sequence length of the data, represents the feature dimension, and generate the corresponding query, key, and value, specifically: Among them, respectively represent the query weight matrix, key weight matrix, and value weight matrix, respectively represent the generated query matrix, key matrix, and value matrix; According to the dot product between the query matrix and the key matrix, calculate the similarity between the query and the key, then there is Among them, represents the th row in the query matrix, which is the query vector at the th time step, represents the th row in the key matrix, which is the key vector at the th time step, represents the transpose matrix, represents the similarity between the query and the key; Among them, represents the similarity between the query and the key, represents the dimension of the query and the key, represents the scaling result; For the query vector at the th time step The correlation between the query vector of a time step and the key vectors thereof is as follows: Wherein, represents the scaling result, represents the th time step's query vector and the correlation between it and all the th time step's key vectors.

[0008] As a preferred solution of the method for predicting energy consumption driven by a compressor according to the present invention, wherein: constructing a feature vector according to the calculated correlation is specifically as follows: Using the calculated correlation to perform weighted fusion on the value matrix, then there is Wherein, represents the th time step's query vector and the correlation between it and all the th time step's key vectors, represents the s th row in the value matrix, which is the th time step's value vector, represents the weighted result of the input query and the correlation with other time steps, and is the feature vector processed by the self-attention mechanism; Input the processed feature vector into the feed-forward network, and after being processed by the activation function, obtain the output result of the basic layer, specifically: Wherein, respectively represent the weight matrices of the first layer and the second layer of the feed-forward network, respectively represent the bias terms of the first layer and the second layer of the feed-forward network, represents the feature vector input into the feed-forward network, represents the output result of the feed-forward network, which is the output result of the basic layer.

[0009] As a preferred solution of the method for predicting energy consumption driven by a compressor according to the present invention, wherein: using a long short-term memory network to predict the compressor energy consumption based on the output result of the basic layer is specifically as follows: Input the output result of the basic layer into the application layer to achieve the capture of long-term dependencies of time series data, specifically: Wherein, represents the output result of the basic layer, represents a bidirectional long short-term memory network, including capturing from the initial time step to time The past dependencies of the input data and capturing from time to the last time step the future dependencies of the input data, specifically: Capturing the past dependencies, Capturing the future dependencies, Bidirectional output capture, wherein, represents the dependencies of the input data captured by the long short-term memory network from the initial time step to time the dependencies of the input data, represents the dependencies of the input data captured by the long short-term memory network from time to the last time step the dependencies of the input data, represents the output concatenation of the two dependencies; Based on the output result of the long short-term network for compressor energy consumption prediction, then there is, wherein, represents the weight matrix of the output layer, represents the bias term of the output layer, represents the output result of the long short-term memory network, represents the predicted energy consumption value.

[0010] As a preferred solution of the energy consumption prediction method for a compressor drive described in the present invention, wherein: the real-time monitoring of the energy consumption prediction error is specifically as follows: Based on the difference between the actual energy consumption of the compressor and the predicted value, specifically: wherein, represents the error between the predicted result of the energy consumption prediction model at the current moment and the actual energy consumption, represents the actual energy consumption of the compressor at the current moment, represents the output result of the prediction model, which is the energy consumption prediction result of the compressor at the current moment, Meanwhile, according to the calculated error, the error growth rate is further calculated, then there is, wherein, represents the error between the predicted result of the energy consumption prediction model at the current moment and the actual energy consumption, represents the error between the prediction result of the energy consumption prediction model at the previous moment and the actual energy consumption. represents the calculated error growth rate.

[0011] An energy consumption prediction system for a compressor drive, comprising a data acquisition and processing module, an energy consumption prediction model construction module, and a feedback adjustment module; the data acquisition and processing module is used to collect and process the compressor operation data; the energy consumption prediction model construction module constructs an energy consumption prediction model based on the processed data; the feedback adjustment module formulates a feedback adjustment mechanism to perform feedback optimization on the energy consumption prediction model.

[0012] A computer device includes a memory and a processor. The memory stores a computer program, and when the processor executes the computer program, it implements the steps of the energy consumption prediction method for a compressor drive.

[0013] A computer-readable storage medium stores a computer program, and when the computer program is executed by a processor, it implements the steps of the energy consumption prediction method for a compressor drive.

[0014] Advantages of the present invention: Through the incremental learning and feedback adjustment mechanism, the energy consumption prediction system can adapt to the changes in the compressor operating conditions in real time, which means that the system can continuously learn and optimize to more accurately predict the energy consumption of the compressor, thereby improving the prediction accuracy. At the same time, the improvement of real-time performance enables the system to respond to the changes in operating conditions more promptly, providing more effective support for industrial energy conservation and cost control; The energy consumption prediction model constructed through the multi-layer capture mechanism can capture the temporal dependencies from both the forward and backward directions of the data simultaneously, capturing the long-term and short-term dependencies of the compressor energy consumption data. This mechanism enables the model to understand the data more comprehensively and further improves the prediction accuracy; By using the sensor fault self-detection algorithm and the weighted Bayesian fusion technology, the weights of each sensor can be dynamically adjusted to ensure the reliability of the collected data. Even if some sensors fail or the data is distorted, the data can be supplemented and corrected by other normally operating sensors, thereby enhancing the robustness of the entire technical solution to sensor faults. Description of the Drawings

[0015] In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the following will briefly introduce the drawings required for the description of the embodiments. Obviously, the following drawings are only some embodiments of the present invention. For those of ordinary skill in the art, without creative efforts, other drawings can be obtained based on these drawings. Among them: Figure 1This is a schematic structural diagram of the overall method steps of an energy consumption prediction method for a compressor drive according to the present invention. Specific Embodiments

[0016] To make the above objects, features, and advantages of the present invention more apparent and understandable, the following detailed description of the specific embodiments of the present invention is provided in conjunction with the accompanying drawings of the specification. Obviously, the described embodiments are part of the embodiments of the present invention, rather than all embodiments. All other embodiments obtained by those of ordinary skill in the art based on the embodiments of the present invention without creative efforts shall fall within the scope of protection of the present invention.

[0017] In the following description, many specific details are set forth to facilitate a full understanding of the present invention. However, the present invention may also be implemented in other ways different from those described herein. Those skilled in the art can make similar generalizations without departing from the connotation of the present invention. Therefore, the present invention is not limited by the specific embodiments disclosed below.

[0018] Secondly, the so-called "one embodiment" or "embodiment" herein refers to a specific feature, structure, or characteristic that may be included in at least one implementation manner of the present invention. The phrase "in one embodiment" appearing in different places in this specification does not necessarily refer to the same embodiment, nor is it an independent or alternative embodiment that excludes other embodiments.

[0019] The present invention is described in detail in conjunction with the schematic diagrams. When detailing the embodiments of the present invention, for ease of explanation, the cross-sectional views showing the device structure will be enlarged locally in an unconventional scale, and the schematic diagrams are only examples and should not limit the scope of protection of the present invention herein. In addition, in actual production, three-dimensional spatial dimensions including length, width, and depth should be included.

[0020] At the same time, in the description of the present invention, it should be noted that the terms "first, second, or third" are only used for descriptive purposes and cannot be construed as indicating or implying relative importance.

[0021] Unless otherwise clearly defined and limited in the present invention, the terms "installed, connected, or coupled" should be understood in a broad sense. For example, it can be a fixed connection, a detachable connection, or an integral connection; it can also be a mechanical connection, an electrical connection, or a direct connection, or it can be indirectly connected through an intermediate medium, or it can be the internal communication of two components. For those of ordinary skill in the art, the specific meanings of the above terms in the present invention can be understood according to specific circumstances.

[0022] Example 1: Refer to Figure 1 , which is an embodiment of the present invention, providing an energy consumption prediction method for a compressor drive, including the following steps. S1: Collection and processing of compressor operation data.

[0023] Specifically, the acquisition and processing of the compressor operation data are carried out by means of dynamic sensor weight allocation. The dynamic sensor weight allocation judges whether the data collected by the sensor is distorted through a sensor fault self-detection algorithm, and combines weighted Bayesian fusion to dynamically adjust the weights of each sensor for data processing, improving the reliability of the collected data. The specific implementation is as follows: In the compressor operation system, sensors are respectively used to collect operation status data, drive system parameters and environmental parameters, specifically: Compressor operation status data , where represents the rotational speed, represents the torque, represents the inlet pressure, represents the outlet pressure, represents the cooling temperature.

[0024] Drive system parameters , where represents the frequency converter power, represents the current, represents the voltage, represents the switch state.

[0025] Environmental parameters , where represents the environmental temperature, represents the humidity, represents the atmospheric pressure.

[0026] In order to eliminate the influence of equipment aging, a sensor fault self-detection algorithm is used to judge whether the data collected by the sensor is distorted, specifically: Construct a joint vector containing all sensor data , then there is, , where represents the compressor operation status data, represents the drive system parameters, represents the environmental parameters, represents the constructed joint vector.

[0027] Use the Mahalanobis distance to evaluate whether the data of the sensor deviates from the normal distribution, then there is, where represents the mean vector of historical normal data, represents the transpose matrix, represents the covariance matrix, which is the correlation between sensor variables and is solved based on the exponentially weighted moving average strategy. The specific solution process is as follows: The covariance matrix is solved based on the weighted average of new data and historical data, so there is where represents the covariance matrix at the previous moment, represents the weighting factor, which controls the influence of historical data on the current matrix. The value of the weighting factor is proportional to the degree of influence of historical data on the current matrix. represents the calculated Mahalanobis distance, which is used to evaluate whether the data of the sensor deviates from the normal distribution. Specifically: Set the normal distribution threshold of the data , if the calculated Mahalanobis distance and the set distribution threshold satisfy the formula , it means that the current data point is abnormal, and the data collected by the corresponding sensor is distorted data, and the sensor needs to be adjusted to collect data again; if the calculated Mahalanobis distance and the set distribution threshold satisfy the formula , it means that the current data point is not abnormal, and the data collected by the corresponding sensor is normal data.

[0028] It should be noted that in order to enhance the reliability of the data, the weighted Bayesian fusion method is used to dynamically adjust the data sharing degree of each sensor. Specifically: Set as the reading of the th sensor, and combine the confidence level evaluated by the Mahalanobis distance to define the dynamic weight of each sensor. Then there is where represents the Mahalanobis distance of the th sensor at the current moment, represents the dynamic weight corresponding to the th sensor, represents the Mahalanobis distance of the th sensor at the current moment; Based on the calculated dynamic weight, data fusion is performed. Then there is where represents the dynamic weight corresponding to the th sensor, represents the reading of the th sensor, represents the data after weighted fusion.

[0029] It should be noted that by detecting whether the data collected by the Mahalanobis distance sensor is abnormal and dynamically adjusting the weights using weighted Bayesian fusion, the accuracy of the compressor operation data is ensured, providing accurate data for subsequent energy consumption prediction and enhancing the robustness of the entire technical solution to sensor failures.

[0030] S2: Construct an energy consumption prediction model based on machine learning algorithms.

[0031] Specifically, the energy consumption prediction model constructed based on machine learning algorithms captures the features of the input data layer by layer through a multi-layer capture mechanism and predicts the energy consumption of the compressor according to the captured features. The multi-layer capture mechanism includes a basic layer for capturing global dependencies and an application layer for capturing long-term dependencies, and the specific implementation is as follows: The data after weighted fusion After linear transformation, it is input into the basic layer, specifically: Among them, represents the linear transformation matrix, represents the bias term, represents the data after weighted fusion, represents the data after linear transformation; Calculate the correlation between the data after linear transformation at different time steps through the self-attention mechanism, specifically: According to the data after linear transformation Generate the corresponding query, key, and value, then there is, Set the data after linear transformation Satisfy the formula Among them, represents the sequence length of the data, represents the feature dimension, and generate the corresponding query, key, and value, specifically: Among them, respectively represent the query weight matrix, key weight matrix, and value weight matrix, respectively represent the generated query matrix, key matrix, and value matrix; Calculate the similarity between the query and the key according to the dot product between the query matrix and the key matrix, then there is, Among them, represents the th row in the query matrix, which is the query vector at the th time step, represents the th row in the key matrix, which is the The key vectors of each time step, Denotes the transpose matrix, Denotes the similarity between the query and the key; Scale the similarity between each query and key to prevent the dot product result from being too large in high-dimensional space. Then, Where, Denotes the similarity between the query and the key, Denotes the dimensions of the query and the key, Denotes the scaling result; For the query vector at the th time step, calculate the relevance between it and all the key vectors at the th time step. Then, Where, Denotes the scaling result, Denotes the relevance between the query vector at the th time step and all the key vectors at the th time step; Use the calculated relevance to perform weighted fusion on the value matrix. Then, Where, Denotes the relevance between the query vector at the th time step and all the key vectors at the th time step, Denotes the th row in the value matrix, which is the value vector at the th time step, Denotes the weighted result of the relevance between the input query and other time steps, which is the feature vector processed by the self-attention mechanism; Input the processed feature vector into the feed-forward network. After being processed by the activation function, obtain the output result of the basic layer, specifically: Where, Denote the weight matrices of the first and second layers of the feed-forward network respectively, Denote the bias terms of the first and second layers of the feed-forward network respectively, Denotes the feature vector input into the feed-forward network, Denotes the output result of the feed-forward network, which is the output result of the basic layer.

[0032] It should be noted that the first layer of the feedforward network performs a linear transformation on the input features to generate a new feature representation. Then, through the activation function, a non-linear transformation is performed on it to enhance the non-linear learning ability of the network. Next, through the linear transformation of the second layer, the output of the hidden layer is mapped to the final output space, and after adding the bias term of the second layer, the final output result is obtained, which is the output result of the basic layer.

[0033] The output result of the basic layer is input into the application layer to achieve the capture of long-term dependencies in time series data, specifically: Among them, represents the output result of the basic layer, represents a bidirectional long short-term memory network, including capturing the past dependencies of the input data from the initial time step to time and capturing the future dependencies of the input data from time to the last time step Specifically: Capture past dependencies, Capture future dependencies, Bidirectional output capture, Among them, represents the dependency of the input data captured from the initial time step to time through the long short-term memory network, represents the dependency of the input data captured from time to the last time step through the long short-term memory network, represents the output splicing of the two dependencies; Based on the output result of the long short-term network, compressor energy consumption prediction is performed, then Among them, represents the weight matrix of the output layer, represents the bias term of the output layer, represents the output result of the long short-term memory network, represents the predicted energy consumption value.

[0034] It should be noted that through the constructed multi-layer capture mechanism, the temporal dependencies can be captured simultaneously from the forward and reverse directions of the data, capturing the long-term and short-term dependencies of the compressor energy consumption data, and further improving the prediction accuracy.

[0035] S3: Perform feedback regulation based on the prediction results of the energy consumption prediction model.

[0036] Specifically, the feedback regulation based on the prediction results of the energy consumption prediction model is to formulate a feedback regulation mechanism for the output results of the energy consumption prediction model, monitor the error of the energy consumption prediction in real time, and automatically adjust the parameters of the model according to the error feedback information to improve the adaptability of the model to the real-time working conditions. The specific implementation is as follows: Based on the difference between the actual energy consumption of the compressor and the predicted value, specifically: Among them, represents the error between the prediction result of the energy consumption prediction model at the current moment and the actual energy consumption, represents the actual energy consumption of the compressor at the current moment, represents the output result of the prediction model, which is the energy consumption prediction result of the compressor at the current moment, At the same time, according to the calculated error, further calculate the error growth rate, then there is, Among them, represents the error between the prediction result of the energy consumption prediction model at the current moment and the actual energy consumption, represents the error between the prediction result of the energy consumption prediction model at the previous moment and the actual energy consumption, represents the calculated error growth rate; Set the error threshold and the error growth rate threshold , and trigger the feedback regulation mechanism according to the set thresholds, specifically: If the comparison result between the error between the prediction result of the energy consumption prediction model at the current moment and the actual energy consumption and the set error threshold satisfies the formula , it means that the prediction result of the current energy consumption prediction model exceeds the upper threshold, and the prediction result of the current energy consumption prediction model is inaccurate. Trigger the feedback regulation mechanism, and perform a secondary prediction of the compressor energy consumption by adjusting the weight matrix and bias term of the output layer of the energy consumption prediction model until the prediction result satisfies the formula ; If the comparison result between the error between the prediction result of the energy consumption prediction model at the current moment and the actual energy consumption and the set error threshold satisfies the formula , indicating that the prediction result of the current energy consumption prediction model does not exceed the upper threshold, the prediction result of the current energy consumption prediction model is accurate, and the relevant parameters of the current energy consumption prediction model are maintained to perform real-time prediction on the compressor energy consumption; If the comparison result between the error between the prediction result of the energy consumption prediction model at the current moment and the actual energy consumption and the set error threshold satisfies the formula , indicating that the prediction result of the current energy consumption prediction model does not exceed the upper threshold, a secondary judgment is made through the error growth rate of the current energy consumption prediction model, specifically: If the comparison of the error growth rate of the current energy consumption prediction model satisfies the formula , indicating that the prediction error of the current energy consumption prediction model shows an increasing trend, triggering a feedback adjustment mechanism, and adjusting the weight matrix and bias term of the output layer of the energy consumption prediction model to perform a secondary prediction on the compressor energy consumption until the prediction result satisfies the formula ; If the comparison of the error growth rate of the current energy consumption prediction model satisfies the formula , indicating that the prediction error of the current energy consumption prediction model shows a decreasing trend, the feedback adjustment mechanism is not triggered, and the relevant parameters of the current energy consumption prediction model are maintained to perform real-time prediction on the compressor energy consumption.

[0037] Embodiment 2: The second embodiment of the present invention provides an energy consumption prediction system for a compressor drive, including a data acquisition and processing module, an energy consumption prediction model construction module, and a feedback adjustment module; Specifically, the data acquisition and processing module is used to collect and process the compressor operation data; the energy consumption prediction model construction module constructs an energy consumption prediction model based on the processed data; the feedback adjustment module formulates a feedback adjustment mechanism to perform feedback optimization on the energy consumption prediction model.

[0038] Furthermore, the data acquisition and processing module collects various types of data in real time from the compressor operation system, including but not limited to sensor data such as rotational speed, torque, pressure, temperature, power, current, and voltage. The module is responsible for preprocessing the raw data to ensure that it can be used by the subsequent model, and directly provides the processed data for the energy consumption prediction model construction module to ensure the accuracy and real-time nature of the input data; The energy consumption prediction model construction module constructs a deep learning model based on the processed data to perform energy consumption prediction. A multi-layer capture mechanism is used to capture the long-term dependencies and short-term fluctuations in the time series data to ensure the accuracy of the model, and at the same time provide real-time energy consumption prediction results, and feedback the output of the model to the feedback adjustment module for optimization and adjustment; The feedback adjustment module is used to monitor the energy consumption prediction error in real time. When the prediction error exceeds the set threshold, it triggers a feedback mechanism to automatically optimize the learning process of the model. By adjusting model parameters, selecting training data, updating the learning rate, etc., it ensures that the prediction accuracy always remains at an ideal level. This module guarantees that the model always adapts to new working conditions and improves the prediction accuracy by adjusting the energy consumption prediction model in real time.

[0039] It should be noted that through the close cooperation and optimization of each module, the energy consumption prediction system of the present invention can not only monitor the operation data of the compressor in real time and accurately predict the energy consumption, but also continuously optimize the model through incremental learning and the feedback mechanism, improving the adaptability of the system to changing working conditions. Each module plays an indispensable role in the system, ensuring the optimization and improvement from data collection to prediction results.

[0040] Furthermore, if the above functions are implemented in the form of software function units and sold or used as independent products, they can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present invention, in essence, or the part that contributes to the prior art, or a part of this technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions for causing a computer device (which can be a personal computer, a server, or a network device, etc.) to execute all or part of the steps of the methods described in various embodiments of the present invention. The aforementioned storage medium includes: USB flash drives, mobile hard disks, read-only memories (ROM, Read-Only Memory), random access memories (RAM, Random Access Memory), magnetic disks, or optical discs, etc., which can store program codes.

[0041] The logic and / or steps represented in the flowchart or described in other ways herein, for example, can be considered as a definite sequence list of executable instructions for implementing logical functions, and can be specifically implemented in any computer-readable medium for use by an instruction execution system, apparatus, or device (such as a computer-based system, a system including a processor, or other systems that can fetch instructions from the instruction execution system, apparatus, or device and execute the instructions), or in combination with these instruction execution systems, apparatus, or devices. For the purposes of this specification, a "computer-readable medium" can be any device that can contain, store, communicate, propagate, or transmit a program for use by or in connection with an instruction execution system, apparatus, or device.

[0042] More specific examples (a non-exhaustive list) of computer-readable media include the following: electrical connections (electronic devices) having one or more wirings, portable computer diskettes (magnetic devices), random access memory (RAM), read-only memory (ROM), erasable programmable read-only memory (EPROM or flash memory), optical fiber devices, and portable compact disc read-only memory (CDROM). Additionally, the computer-readable media can even be paper or other suitable media on which the program can be printed, as the program can be obtained electronically, for example, by optically scanning the paper or other media, then editing, interpreting, or otherwise processing it as appropriate, and then storing it in a computer memory.

[0043] In addition, to provide a concise description of the exemplary embodiments, all features of the actual embodiments may not be described (i.e., those features that are not relevant to the currently contemplated best mode of carrying out the invention or those features that are not relevant to the implementation of the invention).

[0044] It should be understood that in the development of any actual implementation, as in any engineering or design project, numerous specific implementation decisions may be made. Such development efforts may be complex and time-consuming, but for those of ordinary skill in the art who benefit from this disclosure, without undue experimentation, such development efforts will be routine work in design, manufacturing, and production.

[0045] It should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and not to limit them. Although the present invention has been described in detail with reference to the preferred embodiments, those of ordinary skill in the art should understand that the technical solutions of the present invention can be modified or equivalently replaced without departing from the spirit and scope of the technical solutions of the present invention, and they should all be covered by the scope of the claims of the present invention.

Claims

1. A method for predicting the energy consumption driven by a compressor, characterized in that: It includes the following steps: Collect the operation data of the compressor using multiple sensors, and determine whether the data collected by the sensors is distorted through a sensor fault self-detection algorithm. At the same time, combined with weighted Bayesian fusion, dynamically adjust the weights of each sensor to process the collected data; Through a multi-layer capture mechanism, capture the characteristics of the processed data layer by layer, and construct an energy consumption prediction model based on the captured characteristics. Specifically: After the weighted fusion data undergoes a linear transformation, it is input to the basic layer, and the self-attention mechanism is used to calculate the correlation between the linearly transformed data at different time steps. According to the calculated correlation, a feature vector is constructed, and the processed feature vector is input to a feed-forward network to output the result of the basic layer. Use a long short-term memory network to predict the compressor energy consumption based on the output result of the basic layer; Formulate a feedback adjustment mechanism to monitor the energy consumption prediction error of the output result of the energy consumption prediction model in real time, and automatically adjust the parameters of the model according to the monitored error feedback information to improve the adaptability of the energy consumption prediction model to real-time working conditions.

2. The energy consumption prediction method for a compressor drive according to claim 1, characterized in that: The specific method for determining whether the data collected by the sensor is distorted through the sensor fault self-detection algorithm is as follows: Construct a combined vector containing all sensor data , then there is , where represents the compressor operating state data represents the drive system parameters represents the environmental parameters represents the constructed combined vector; Use the Mahalanobis distance to evaluate whether the data of the sensor deviates from the normal distribution. Then: Among them, represents the mean vector of historical normal data, represents the transpose matrix, represents the covariance matrix, which is the correlation between sensor variables and is solved based on the exponentially weighted moving average strategy. The specific solution process is as follows: Solve the covariance matrix according to the weighted average of the new data and historical data. Then: Among them, represents the covariance matrix at the previous moment, represents the weighting factor, which controls the influence of historical data on the current matrix. The value of the weighting factor is proportional to the degree of influence of historical data on the current matrix. represents the calculated Mahalanobis distance, which is used to evaluate whether the data of the sensor deviates from the normal distribution. Specifically: Set the normal distribution threshold of the data , if the calculated Mahalanobis distance compared with the set distribution threshold satisfies the formula , it indicates that the current data point is abnormal, and the data collected by the corresponding sensor is distorted data, and the sensor needs to be adjusted to re-collect the data; If the calculated Mahalanobis distance compared with the set distribution threshold satisfies the formula , it indicates that the current data point has no anomaly, and the data collected by the corresponding sensor is normal data.

3. The energy consumption prediction method for a compressor drive according to claim 2, characterized in that: The specific method for processing the collected data is as follows: Settings For the reading of the sensor, and combining the confidence evaluated by the Mahalanobis distance, define the dynamic weights of each sensor. Then, Among them, represents the Mahalanobis distance of the -th sensor at the current moment, represents the dynamic weight corresponding to the -th sensor, represents the Mahalanobis distance of the -th sensor at the current moment; Based on the calculated dynamic weights, perform data fusion. Then: Among them, represents the dynamic weight corresponding to the th sensor, represents the reading of the th sensor, represents the data after weighted fusion.

4. The energy consumption prediction method for a compressor drive according to claim 3, characterized in that: The specific method for calculating the correlation between the linearly transformed data at different time steps through the self-attention mechanism is as follows: Based on the data after linear transformation Generate the corresponding queries, keys, and values, then there are Data after setting linear transformation Satisfy the formula , where represents the sequence length of the data, represents the feature dimension, and generate corresponding queries, keys, and values, specifically: Among them, respectively represent the query weight matrix, the key weight matrix, and the value weight matrix, respectively represent the generated query matrix, key matrix, and value matrix; Calculate the similarity between the query and the key according to the dot product between the query matrix and the key matrix. Then: in, represents the first Row, is the query vector of the time step, represents the first Row, for The key vector for time steps, represents the transposed matrix, Indicates the similarity between the query and the key; Scale the similarity between each query and the key to prevent the dot product result from being too large in the high-dimensional space. Then: Among them, represents the similarity between the query and the key, represents the dimensions of the query and the key, represents the scaled result; For the query vector at the th time step , calculate the relevance between it and all key vectors at the th time step, then Among them, represents the scaling result, represents the association between the query vector at the -th time step and all key vectors at the -th time step.

5. The energy consumption prediction method for a compressor drive according to claim 4, characterized in that: The specific method for constructing a feature vector according to the calculated correlation is as follows: Use the calculated correlation to perform weighted fusion on the value matrix. Then: Among them, represents the correlation between the query vector at the -th time step and all the key vectors at the -th time step, represents the -th row in the value matrix, which is the value vector at the -th time step, represents the weighted result of the correlation between the input query and other time steps, which is the feature vector processed by the self-attention mechanism; Input the processed feature vector into a feed-forward network, and after being processed by an activation function, obtain the output result of the basic layer. Specifically: Among them, respectively represent the weight matrices of the first and second layers of the feedforward network, respectively represent the bias terms of the first and second layers of the feedforward network, represents the feature vector input to the feedforward network, represents the output result of the feedforward network, which is the output result of the basic layer.

6. The energy consumption prediction method for a compressor drive according to claim 5, wherein: The specific method for predicting the compressor energy consumption using the long short-term memory network based on the output result of the basic layer is as follows: Input the output result of the basic layer into the application layer to achieve the capture of long-term dependencies of time series data. Specifically: Among them, represents the output result of the base layer, represents a bidirectional long short-term memory network, including capturing the past dependencies of the input data from the initial time step to time and capturing the future dependencies of the input data from time to the last time step Specifically, Capture past dependencies, Capture future dependencies, Bidirectional output capture, Among them, represents the dependency of the input data captured by the long short-term memory network from the initial time step to time , represents the dependency of the input data captured by the long short-term memory network from time to the last time step , represents the output concatenation of the two dependencies; Predict the compressor energy consumption based on the output result of the long short-term network. Then: Among them, represents the weight matrix of the output layer, represents the bias term of the output layer, represents the output result of the long short-term memory network, represents the predicted energy consumption value.

7. The energy consumption prediction method for a compressor drive according to claim 6, wherein: The specific method for monitoring the energy consumption prediction error in real time is as follows: Based on the difference between the actual energy consumption and the predicted value of the compressor. Specifically: Among them, represents the error between the prediction result of the energy consumption prediction model at the current moment and the actual energy consumption, represents the actual energy consumption of the compressor at the current moment, represents the output result of the prediction model, which is the energy consumption prediction result of the compressor at the current moment, At the same time, according to the calculated error, further calculate the error growth rate. Then: Among them, represents the error between the prediction result of the energy consumption prediction model at the current moment and the actual energy consumption, represents the error between the prediction result of the energy consumption prediction model at the previous moment and the actual energy consumption, represents the calculated error growth rate.

8. An energy consumption prediction system for compressor drive, applied to an energy consumption prediction method for compressor drive described in claims 1 to 7, characterized in that, It includes a data acquisition and processing module, an energy consumption prediction model construction module, and a feedback adjustment module; The data acquisition and processing module is used to collect and process the operation data of the compressor; The energy consumption prediction model construction module constructs an energy consumption prediction model based on the processed data; The feedback adjustment module formulates a feedback adjustment mechanism to perform feedback optimization on the energy consumption prediction model.

9. A computer device, comprising a memory and a processor, the memory storing a computer program, characterized in that, When the processor executes the computer program, the steps of the method according to any one of claims 1 to 7 are implemented.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the computer program is executed by the processor, the steps of the method according to any one of claims 1 to 7 are implemented.

Images