Intelligent refrigeration system efficiency optimization method, system and device based on deep learning

Abstract

The invention provides an intelligent refrigeration system efficiency optimization method, system and device based on deep learning, and relates to the technical field of intelligent refrigeration system efficiency optimization. The method comprises the following steps: firstly, collecting historical operation parameters and environmental data of the refrigeration system, preprocessing, and integrating into a sample set according to a timestamp; and based on the hybrid model of the convolutional neural network and the long-short term memory network, constructing a state prediction model for predicting the system operation state. And in combination with the state prediction model, an optimization objective function is defined for data prediction at the next future moment, and an optimal adjustment strategy for system operation is formulated by using a deep Q network model. And finally, monitoring current operation parameters and environment data in real time, calculating a comprehensive safety score, comparing the comprehensive safety score with a safety threshold value, and deciding whether to continue to optimize the efficiency or preferentially carry out troubleshooting. According to the method, through combination of deep learning and physical modeling, efficient and safe refrigeration system operation optimization is realized.

Description

Technical Field

[0001] The present invention relates to the technical field of intelligent refrigeration system efficiency optimization, and specifically to an intelligent refrigeration system efficiency optimization method, system and device based on deep learning. Background Art

[0002] Refrigeration systems are widely used in industrial production, logistics and transportation, commercial and household air conditioning, and other fields. Their operating efficiency plays an important role in energy consumption and environmental impact. However, traditional refrigeration systems generally adopt a fixed parameter operation mode, which is difficult to dynamically adapt to changes in the external environment, which may lead to increased energy consumption and reduced efficiency. In addition, long-term operation of refrigeration systems may also cause performance degradation and even safety hazards due to equipment aging, parameter abnormalities and other problems. Therefore, higher requirements are placed on the efficiency optimization and safety assurance of refrigeration systems.

[0003] With the development of artificial intelligence, big data and Internet of Things technologies, intelligent technologies are gradually being introduced into refrigeration system management. By collecting system operating parameters and environmental data in real time, and combining deep learning algorithms to model and predict the operating status, dynamic adjustment and optimization of the refrigeration system can be achieved. It can not only improve system efficiency, but also identify potential faults at an early stage and reduce equipment operation risks. The application of this technology effectively makes up for the shortcomings of traditional refrigeration systems in terms of autonomous optimization and safety assurance.

[0004] The prior art has the following deficiencies: Although the existing intelligent refrigeration system efficiency optimization technology has achieved energy efficiency improvement and partial automation adjustment to a certain extent, it still has many shortcomings and is difficult to give full play to its advantages in a complex dynamic environment. First, the data-driven model also has bottlenecks in its ability to handle high-dimensional nonlinear problems, especially when it involves the synergy of multiple external environmental factors (such as temperature, humidity, load fluctuations) and internal system parameters (such as compressor operating status, coolant flow), it is difficult for traditional models to fully capture these complex relationships. Secondly, the existing technology still has shortcomings in real-time and dynamic optimization. Many optimization strategies for intelligent refrigeration systems rely on pre-set rules or periodic adjustments. This method cannot quickly respond to instantaneous changes in environmental conditions, resulting in unstable operating efficiency of the system under dynamic loads or complex working conditions. Finally, the existing technology also has shortcomings in the balance of optimization objectives. Some optimization schemes simply pursue efficiency improvement or energy consumption reduction, and easily ignore the safety and stability of the refrigeration system operation. Existing technologies usually lack systematic consideration of multi-objective optimization, and it is difficult to find the best balance between improving efficiency, reducing energy consumption and ensuring equipment safety.

[0005] The above information disclosed in this Background section is only for enhancement of understanding of the background of the present disclosure and therefore it may contain information that does not constitute the prior art that is already known to one of ordinary skill in the art. Summary of the invention

[0006] The purpose of the present invention is to provide a method, system and device for optimizing the efficiency of an intelligent refrigeration system based on deep learning to solve the problems raised in the above background technology.

[0007] To achieve the above object, the present invention provides the following technical solutions: A method for optimizing the efficiency of an intelligent refrigeration system based on deep learning, the specific steps comprising: Step 1: Collect the operating parameters and environmental data of the refrigeration system at historical moments and perform preprocessing operations on them. The preprocessed operating parameters and environmental data are integrated into a sample set according to the timestamp. The operating parameters include compressor power, refrigerant flow, condenser temperature, evaporator temperature, system pressure and refrigerant mass flow, and the environmental data include ambient temperature and ambient humidity; Step 2: Build a state prediction model based on the hybrid model architecture of convolutional neural network and long short-term memory network. The operating parameters and environmental data of multiple consecutive moments in the sample set are constructed into multi-dimensional time series data as input, and the operating parameters and environmental data of the next moment are used as labels to train the state prediction model. Step 3: Obtain the operating parameters and environmental data at multiple consecutive moments including the current moment, and input them into the trained state prediction model to output the operating parameters and environmental data; Step 4: Define the optimization objective function based on the output of the state prediction model, and train the depth according to the optimization objective function. Network models to determine adjustment strategies; Step 5: Collect the current operating parameters and environmental data, calculate the current comprehensive safety score, and compare the current comprehensive safety score with the preset safety threshold to determine whether to continue efficiency optimization or prioritize troubleshooting and repairing problems.

[0008] Furthermore, the specific logic for collecting the operating parameters and environmental data of the refrigeration system is as follows: A temperature sensor and a humidity sensor are installed at the condenser suction port to collect the ambient temperature and humidity; a power sensor is installed at the power input end of the compressor to collect the compressor power; a refrigerant flow meter is installed on the straight pipe section of the main loop of the refrigerant pipeline to collect the refrigerant flow data; a temperature sensor is installed on the refrigerant pipeline at the condenser outlet to collect the condenser temperature data; a temperature sensor is installed on the refrigerant pipeline at the evaporator inlet to collect the evaporator temperature; a mass flow meter is installed in the compressor outlet pipeline to collect the refrigerant mass flow; pressure sensors are installed at the condenser outlet and the evaporator inlet to collect the condensing pressure and the evaporating pressure; the average pressure of the two is calculated to represent the system pressure, and the formula is as follows: ; in, Indicates the system pressure, Indicates the condensing pressure, Indicates evaporation pressure; The preprocessing step includes removing outliers from the operating parameters and environmental data collected at historical moments, calculating the Z-score of each operating parameter and environmental data, and if the absolute value of the Z-score of the operating parameter and environmental data is , the parameter is regarded as an outlier and removed, and the missing value is filled by the mean filling method, and then the Min-Max normalization is used to normalize the operating parameters and environmental data to Within the range, the preprocessed data are integrated into a sample set.

[0009] Furthermore, the specific logic of the training state prediction model is as follows: The operating parameters and environmental data of multiple consecutive moments in the sample set are constructed into multidimensional time series data as the input of the state prediction model, which is expressed as: ; in, Represents multidimensional time series data, express The compressor power at the time, express The refrigerant flow rate at the moment, express The condenser temperature at time express The evaporator temperature at time express System pressure at all times, express The refrigerant mass flow rate at the moment, express The ambient temperature at the moment, express The ambient humidity at the moment. Similarly, the second row of parameters represents The operating parameters and environmental data at the moment. The last line of parameters indicates The operating parameters and environmental data at the time, t is the time variable, is the time interval between adjacent moments, Indicates the time window size; The operating parameters and environmental data labels of the next moment in the sample set are constructed into multidimensional time data, which can be expressed as: ; in, Represents multidimensional time data, express The compressor power at the time, express The refrigerant flow rate at the moment, express The condenser temperature at time express The evaporator temperature at time express System pressure at all times, express The refrigerant mass flow rate at the moment, express The ambient temperature at the moment, express Ambient humidity at all times; The sample set is divided into a training set and a test set with a ratio of 7:3. First, the multidimensional time series data in the training set is input into the convolutional neural network to extract local spatial features. Then, the extracted features are input into the LSTM module to learn the dynamic changes of the time series. Finally, the extracted high-order features are mapped to the prediction results of the target value through the fully connected layer to output the operating parameters and environmental data at time t+1. The data in the test set is then input for testing. The mean absolute error is used as the loss function. If the decrease is less than 0.001 within 10 consecutive rounds of training, the mean absolute error has converged and the training is completed.

[0010] Furthermore, the specific logic for determining the optimization goal and adjustment strategy is as follows: The optimization objective function is defined as: ; in, In order to optimize the objective function value, the optimization strategy is to minimize the total energy consumption of the compressor while maintaining the comfort of the indoor environment within the optimization time period. Indicates the current moment, that is, the moment when optimization starts. Indicates the time to stop optimization. represents the length of the optimization period, Indicates the current ambient temperature. Indicates the set indoor temperature. are the weight parameters of each item, ,and ; Define the current state as: ; in, Indicates the current state. They are the condenser temperature, evaporator temperature, compressor power, and refrigerant flow at the current moment; Defining actions Action 1: adjust the condenser target temperature; Action 2: adjust the compressor power output; Action 3: control the refrigerant flow; The reward function is defined as: ; in, represents the reward function; The goal is to train a policy that maximizes the cumulative long-term reward: ; ; in, represents the accumulated long-term reward, represents the discount factor, , Indicated in Instant rewards at all times, represents the index of the time interval within the optimization time, and ; Taking the current state as input, based on the depth The network model uses the Bellman equation to calculate the The value is taken as the output, and the prediction is minimized Values ​​and Goals The error between the values, update depth The parameters of the network are used to complete the model training. Each time the state prediction model outputs the predicted running parameters, the state is determined and input into the depth In the network model, select The action with the largest value is used as the adjustment strategy.

[0011] Furthermore, the specific logic for calculating the comprehensive safety score at the current moment is: The single parameter safety score formula for calculating the current operating parameters and environmental data is: ; ; in, Indicates The safety score of each parameter, Indicates The reference value of the parameter, Indicates The half width of the parameter safety range, Indicates The safety lower limit of the parameter, Indicates The safety upper limit of the parameter; The formula used to calculate the overall safety score is: ; in, Represents the comprehensive safety score, For the The weight coefficient of the parameter safety score, Indicates the total number of types of operating parameters and environmental data, is the parameter index, and ; Compare the current comprehensive safety score with the preset safety threshold. If the current comprehensive safety score is higher than the safety threshold, it indicates that the system is in normal operation, and efficiency optimization will continue. If the current comprehensive safety score is lower than the safety threshold, it indicates that there are safety hazards in the system, and the problem will be checked and repaired first before efficiency optimization.

[0012] The present invention further provides an intelligent refrigeration system efficiency optimization system based on deep learning, wherein the intelligent refrigeration system efficiency optimization system based on deep learning is used to implement the above-mentioned intelligent refrigeration system efficiency optimization method based on deep learning, including: The data acquisition and preprocessing module is used to collect the operating parameters and environmental data of the refrigeration system at historical moments and perform preprocessing operations on them. The preprocessed operating parameters and environmental data are integrated into a sample set according to the timestamp. The operating parameters include compressor power, refrigerant flow, condenser temperature, evaporator temperature, system pressure and refrigerant mass flow, and the environmental data include ambient temperature and ambient humidity; The state prediction model building module is used to build a state prediction model based on a hybrid model architecture of a convolutional neural network and a long short-term memory network. The operating parameters and environmental data of multiple consecutive moments in the sample set are constructed into multi-dimensional time series data as input, and the operating parameters and environmental data of the next moment are used as labels to train the state prediction model. A real-time state prediction module is used to obtain the operating parameters and environmental data at multiple consecutive moments including the current moment, and input them into the trained state prediction model to output the operating parameters and environmental data; The optimization strategy generation module is used to define the optimization objective function based on the output of the state prediction model and train the depth according to the optimization objective function. Network models to determine adjustment strategies; The safety assessment and decision-making module is used to collect the current operating parameters and environmental data, calculate the current comprehensive safety score, and compare the current comprehensive safety score with the preset safety threshold to determine whether to continue efficiency optimization or prioritize troubleshooting and repairing problems.

[0013] The present invention further provides an intelligent refrigeration system efficiency optimization device based on deep learning, the device comprising a memory and a processor, the memory storing a computer program that can be run on the processor, and the processor implementing the steps of the above-mentioned method for optimizing the efficiency of an intelligent refrigeration system based on deep learning when executing the computer program.

[0014] In the above technical solution, the technical effects and advantages provided by the present invention are: This intelligent refrigeration system efficiency optimization method integrates deep learning and reinforcement learning technologies to achieve dynamic improvement of system operating efficiency and real-time guarantee of operating safety. Convolutional neural networks and long short-term memory network models are used to accurately predict complex data, enhancing the robustness and reliability of the algorithm; at the same time, the optimal adjustment strategy is generated through a deep Q network to meet the energy-saving and consumption-reduction goals while taking into account the safety and stability of equipment operation. In addition, the system equipment status is comprehensively monitored, and an early warning mechanism for potential faults is established to identify risks in advance and avoid faults. This method has the characteristics of strong real-time performance, high intelligence, and wide adaptability. It can effectively reduce energy consumption and reduce environmental burden, providing innovative solutions for the intelligent upgrade and sustainable development of refrigeration systems. BRIEF DESCRIPTION OF THE DRAWINGS

[0015] Figure 1 It is a schematic diagram of the overall method flow of the present invention; Figure 2 Schematic diagram of the system structure of the present invention. DETAILED DESCRIPTION

[0016] In order to make the objectives, technical solutions and advantages of the present invention more clearly understood, the present invention is further described in detail below in conjunction with specific embodiments.

[0017] It should be noted that, unless otherwise defined, the technical terms or scientific terms used in the present invention should be understood by people with ordinary skills in the field to which the present invention belongs. The words "first", "second" and similar words used in the present invention do not indicate any order, quantity or importance, but are only used to distinguish different components. "Include" or "comprise" and similar words mean that the elements or objects appearing before the word include the elements or objects listed after the word and their equivalents, without excluding other elements or objects. "Connect" or "connected" and similar words are not limited to physical or mechanical connections, but may include electrical connections, whether direct or indirect. "Up", "down", "left", "right" and the like are only used to indicate relative positional relationships. When the absolute position of the described object changes, the relative positional relationship may also change accordingly.

[0018] Example: See also Figure 1 , the present invention provides a technical solution: A method for optimizing the efficiency of an intelligent refrigeration system based on deep learning, the specific steps comprising: Step 1: Collect the operating parameters and environmental data of the refrigeration system at historical moments and perform preprocessing operations on them. The preprocessed operating parameters and environmental data are integrated into a sample set according to the timestamp. The operating parameters include compressor power, refrigerant flow, condenser temperature, evaporator temperature, system pressure and refrigerant mass flow, and the environmental data include ambient temperature and ambient humidity; In this embodiment, the specific logic for collecting the operating parameters and environmental data of the refrigeration system is as follows: A temperature sensor and a humidity sensor are installed at the condenser suction port to collect the ambient temperature and humidity; a power sensor is installed at the power input end of the compressor to collect the compressor power; a refrigerant flow meter is installed on the straight pipe section of the main loop of the refrigerant pipeline to collect the refrigerant flow data; a temperature sensor is installed on the refrigerant pipeline at the condenser outlet to collect the condenser temperature data; a temperature sensor is installed on the refrigerant pipeline at the evaporator inlet to collect the evaporator temperature; a mass flow meter is installed in the compressor outlet pipeline to collect the refrigerant mass flow; pressure sensors are installed at the condenser outlet and the evaporator inlet to collect the condensing pressure and the evaporating pressure; the average pressure of the two is calculated to represent the system pressure, and the formula is as follows: ; in, Indicates the system pressure, Indicates the condensing pressure, Indicates evaporation pressure; The preprocessing step includes removing outliers from the operating parameters and environmental data collected at historical moments, calculating the Z-score of each operating parameter and environmental data, and if the absolute value of the Z-score of the operating parameter and environmental data is , the parameter is regarded as an outlier and removed, and the missing value is filled by the mean filling method, and then the Min-Max normalization is used to normalize the operating parameters and environmental data to Within the range, the preprocessed data are integrated into a sample set.

[0019] The designed sensor layout logic ensures that necessary data is collected at key nodes. For example, the power sensor of the compressor can directly reflect the energy consumption. The refrigerant flow meter, temperature sensor and pressure sensor monitor the refrigerant circulation status, heat exchange performance and pressure changes in the system respectively. These parameters are important indicators affecting the refrigeration efficiency. This sensor layout method is scientific and feasible, and can meet the high-dimensional input data requirements required by the deep learning model. Outlier removal, missing value filling and normalization are common methods for data cleaning and standardization, which can effectively improve data quality. The Z-score method is a classic statistical tool for outlier detection. It judges based on the degree of deviation of each data in the overall distribution and has high accuracy. The mean filling method can avoid the negative impact of missing data on the model while maintaining the data distribution characteristics. Min-Max normalization can map data to a unified range while avoiding information loss, which is suitable for the characteristics of the deep learning model. The operating state of the refrigeration system is continuous and dynamic, and there is a time dependency between its historical operating parameters and environmental data. By integrating the sample set according to the timestamp, it can provide structured data input for the deep learning model, which is convenient for the model to capture the laws in the time series.

[0020] Step 2: Build a state prediction model based on the hybrid model architecture of convolutional neural network and long short-term memory network. The operating parameters and environmental data of multiple consecutive moments in the sample set are constructed into multi-dimensional time series data as input, and the operating parameters and environmental data of the next moment are used as labels to train the state prediction model. In this embodiment, the specific logic of the training state prediction model is: The operating parameters and environmental data of multiple consecutive moments in the sample set are constructed into multidimensional time series data as the input of the state prediction model, which is expressed as: ; in, Represents multidimensional time series data, express The compressor power at the time, express The refrigerant flow rate at the moment, express The condenser temperature at time express The evaporator temperature at time express System pressure at all times, express The refrigerant mass flow rate at the moment, express The ambient temperature at the moment, express The ambient humidity at the moment. Similarly, the second row of parameters represents The operating parameters and environmental data at the moment. The last line of parameters indicates The operating parameters and environmental data at the time, t is the time variable, is the time interval between adjacent moments, Indicates the time window size; The operating parameters and environmental data labels of the next moment in the sample set are constructed into multidimensional time data, which can be expressed as: ; in, Represents multidimensional time data, express The compressor power at the time, express The refrigerant flow rate at the moment, express The condenser temperature at time express The evaporator temperature at time express System pressure at all times, express The refrigerant mass flow rate at the moment, express The ambient temperature at the moment, express Ambient humidity at all times; The sample set is divided into a training set and a test set with a ratio of 7:3. First, the multidimensional time series data in the training set is input into the convolutional neural network to extract local spatial features. Then, the extracted features are input into the LSTM module to learn the dynamic changes of the time series. Finally, the extracted high-order features are mapped to the prediction results of the target value through the fully connected layer to output the operating parameters and environmental data at time t+1. The data in the test set is then input for testing. The mean absolute error is used as the loss function. If the decrease is less than 0.001 within 10 consecutive rounds of training, the mean absolute error has converged and the training is completed.

[0021] Convolutional neural networks (CNNs) have excellent local feature extraction capabilities when processing high-dimensional data. Multidimensional time series data (such as compressor power, refrigerant flow, temperature, pressure, etc.) have complex nonlinear relationships between variables, and CNN can automatically extract local spatial features between these variables through convolution operations. By scanning the input data with a convolution kernel, the local correlation between different operating parameters can be captured, such as the relationship between condenser temperature and system pressure, thereby effectively characterizing the working state of the refrigeration system. Through weight sharing and dimensionality reduction operations, key features are retained while reducing computational complexity, avoiding data redundancy and greatly improving model processing speed. CNN can filter out irrelevant or noisy information in high-dimensional data, providing a purer feature representation for subsequent long short-term memory network (LSTM) modules.

[0022] LSTM is good at processing time series data, and its built-in memory gating mechanism enables it to capture short-term dynamic changes and long-term dependencies in time series. For the operating status of the refrigeration system, the parameters at the current moment are usually significantly affected by the status in the past period of time, and LSTM can model these dynamic relationships. It can quickly capture short-term changes in operating parameters, such as the immediate impact of fluctuations in ambient temperature and humidity on compressor power. Changes in the efficiency of the refrigeration system often have a certain cumulative effect. For example, the gradual reduction in refrigerant flow may affect the trend of changes in condenser temperature in the future. The memory mechanism of LSTM can effectively store these long-term dependent information. Combining the two models can give full play to the advantages of the two network models. After CNN extracts the local spatial features of high-dimensional data, LSTM further learns the dynamic relationship of features changing over time to achieve deep fusion of information. The hybrid model can simultaneously model the short-term correlation and long-term dependency between different operating parameters of the refrigeration system, thereby more comprehensively reflecting the operating laws of the system. The hybrid model can more accurately predict the operating parameters and environmental data at the next moment, thereby providing high-quality state prediction for refrigeration system optimization.

[0023] Step 3: Obtain the operating parameters and environmental data at multiple consecutive moments including the current moment, and input them into the trained state prediction model to output the operating parameters and environmental data; Collect the operating parameters and environmental data at multiple times in the current time period, pre-process the collected data, input the pre-processed data into the trained state prediction model, and obtain the operating parameters and environmental data predicted at the next moment output by the state prediction model; Step 4: Define the optimization objective function based on the output of the state prediction model, and train the depth according to the optimization objective function. Network models to determine adjustment strategies; In this embodiment, the specific logic for determining the optimization target and the adjustment strategy is as follows: The optimization objective function is defined as: ; in, In order to optimize the objective function value, the optimization strategy is to minimize the total energy consumption of the compressor while maintaining the comfort of the indoor environment within the optimization time period. Indicates the current moment, that is, the moment when optimization starts. Indicates the time to stop optimization. represents the length of the optimization period, Indicates the current ambient temperature. Indicates the set indoor temperature. are the weight parameters of each item, ,and ; Define the current state as: ; in, Indicates the current state. They are the condenser temperature, evaporator temperature, compressor power, and refrigerant flow at the current moment; Defining actions Action 1: adjust the condenser target temperature; Action 2: adjust the compressor power output; Action 3: control the refrigerant flow; The reward function is defined as: ; in, represents the reward function; The goal is to train a policy that maximizes the cumulative long-term reward: ; in, express The value is rounded up; ; in, represents the accumulated long-term reward, represents the discount factor, , Indicated in Instant rewards at all times, represents the index of the time interval within the optimization time, and ; Taking the current state as input, based on the depth The network model uses the Bellman equation to calculate the The value is taken as the output, and the prediction is minimized Values ​​and Goals The error between the values, update depth The parameters of the network are used to complete the model training. Each time the state prediction model outputs the predicted running parameters, the state is determined and input into the depth In the network model, select The action with the largest value is used as the adjustment strategy.

[0024] The first item quantifies the total energy consumption during system operation by integrating the compressor energy consumption over time, which directly reflects the goal of energy saving optimization. It is the core source of energy consumption in the refrigeration system. The larger the objective function Increases, indicating that the energy consumption ratio is too high, and the optimization effect is worse. Reduce the objective function The smaller the indoor temperature, the better the optimization effect. The second item quantifies the system's performance in maintaining comfort (how close the indoor temperature is to the set value) by integrating the indoor ambient temperature deviation over time. The larger the objective function Increase, indicating that the indoor temperature control is poor and the optimization effect is poor. Reduce, the objective function The optimization effect is better when the value is reduced. The deep Q network is based on the reinforcement learning framework and has significant adaptive capabilities through dynamic learning of states, actions, and rewards. The weight parameter indicates the importance of controlling the total energy consumption in the optimization objective. If the value is larger, the optimization process will tend to reduce the energy consumption of the compressor and may reduce the priority of comfort. The weight parameter indicates the importance of the control in the optimization objective. If If the optimization process is large, the indoor comfort will be improved, which may increase energy consumption. Here, the optimization goal focuses on energy saving first, so the weight relationship is , indicating that the optimization goal is more inclined to reduce energy consumption. and The sum of is limited to 1, which ensures that the two weight coefficients are within a standardized range. , comprehensively characterizes the real-time operating status of the refrigeration system, including key parameters such as condenser temperature, evaporator temperature, compressor power, and refrigerant flow. The three defined actions (adjusting the condenser target temperature, adjusting the compressor power, and controlling the refrigerant flow) cover the main controllable parameters of the refrigeration system, providing sufficient operating space for dynamic regulation. Through the reward function , the objective function value will be optimized The minimization task is converted into the reward maximization goal under the reinforcement learning framework to ensure the consistency of the optimization process with the actual operation requirements. This optimization strategy based on reinforcement learning can adapt to complex dynamic environments (such as ambient temperature fluctuations, load changes, etc.), make dynamic decisions under different working conditions, and realize the intelligent operation of the refrigeration system.

[0025] Step 5: Collect the current operating parameters and environmental data, calculate the current comprehensive safety score, and compare the current comprehensive safety score with the preset safety threshold to determine whether to continue efficiency optimization or prioritize troubleshooting and repairing problems; In this embodiment, the specific logic for calculating the comprehensive safety score at the current moment is: The single parameter safety score formula for calculating the current operating parameters and environmental data is: ; ; in, Indicates The safety score of each parameter, Indicates The reference value of the parameter, Indicates The half width of the parameter safety range, Indicates The safety lower limit of the parameter, Indicates The safety upper limit of the parameter; The formula used to calculate the overall safety score is: ; in, Represents the comprehensive safety score, For the The weight coefficient of the parameter safety score, Indicates the total number of types of operating parameters and environmental data, is the parameter index, and ; The specific logic for obtaining the safety score of each single parameter is as follows: The formula for calculating the compressor power safety score at the current moment is: ; in, The safety score representing the parameter compressor power, The unit is ; The formula for calculating the refrigerant flow safety score at the current moment is: ; in, Indicates the safety score of the parameter refrigerant flow rate, The unit is ; The formula for calculating the condenser temperature safety score at the current moment is: ; in, represents the safety score of the parameter condenser temperature, The unit is ℃; The formula for calculating the evaporator temperature safety score at the current moment is: ; in, The safety score for the parameter evaporator temperature, The unit is ℃; The formula for calculating the system pressure safety score at the current moment is: ; in, represents the safety score of the parameter system pressure, The unit is ; The formula for calculating the refrigerant mass flow safety score at the current moment is: ; in, Indicates the safety score of the parameter refrigerant mass flow rate, The unit is ; The formula for calculating the current ambient temperature safety score is: ; in, represents the safety score of the parameter ambient temperature, The unit is ℃; The formula for calculating the current ambient humidity safety score is: ; in, Indicates the safety score of the parameter ambient humidity; The formula used to calculate the overall safety score is: ; in, Represents the comprehensive safety score, is the weight coefficient of the safety score of each corresponding parameter, ,and ; Compare the current comprehensive safety score with the preset safety threshold. If the current comprehensive safety score is higher than the safety threshold, it indicates that the system is in normal operation, and efficiency optimization will continue. If the current comprehensive safety score is lower than the safety threshold, it indicates that there are safety hazards in the system, and the problem will be checked and repaired first before efficiency optimization.

[0026] System pressure is the core safety parameter of the refrigeration system, which directly affects the mechanical safety of the system. Pressure exceeding the safety range may cause equipment damage (such as condenser, pipeline rupture) or refrigerant leakage, so it has the highest weight; refrigerant flow is the "blood" of the refrigeration system. Too small a flow will lead to a decrease in refrigeration efficiency, and too large a flow may cause liquid hammer or compressor overload. The flow has a significant impact on the heat exchange performance of the evaporator and condenser, so it has a higher weight; ambient temperature directly affects the heat dissipation performance of the condenser. High temperature may lead to reduced condensation effect and increased system pressure; low temperature may lead to too low system operating efficiency, and changes in ambient temperature may be a trigger for safety issues, so the weight is moderate; compressor power determines the core energy consumption and operating status of the system. Excessive power may cause equipment overheating, and too small power may lead to insufficient load, so its weight is moderate; the impact of ambient humidity is mainly on heat exchange performance, especially in the condenser part. Although it will affect the efficiency of the system, its impact is lower than that of pressure and temperature, so its weight is lower; condenser temperature reflects heat dissipation capacity. High temperature may lead to a decrease in system efficiency, but its impact is more focused on efficiency rather than safety, so its weight is lower; evaporator temperature mainly affects the refrigeration effect rather than system operation safety. It can be indirectly controlled by other parameters (such as refrigerant flow), so it has the lowest weight.

[0027] Single parameter scoring The value of is limited to [0,1]. Equal to reference value When the score reaches the maximum value 1, it indicates that the parameter is in the best state; when the parameter Beyond the safety range (i.e. or ), the score is 0, indicating that the parameter is completely unsafe; when the parameter When the parameters are within the safety range, the score decreases as the parameters deviate from the reference value. The closer the parameters are to the ideal value, the higher the safety and the higher the score. The higher the value, the less likely it is that the parameter will contribute to the safety of the system.

[0028] See also Figure 2 The present invention further provides an intelligent refrigeration system efficiency optimization system based on deep learning, wherein the intelligent refrigeration system efficiency optimization system based on deep learning is used to implement the above-mentioned intelligent refrigeration system efficiency optimization method based on deep learning, including: The data acquisition and preprocessing module is used to collect the operating parameters and environmental data of the refrigeration system at historical moments and perform preprocessing operations on them. The preprocessed operating parameters and environmental data are integrated into a sample set according to the timestamp. The operating parameters include compressor power, refrigerant flow, condenser temperature, evaporator temperature, system pressure and refrigerant mass flow, and the environmental data include ambient temperature and ambient humidity; The state prediction model building module is used to build a state prediction model based on a hybrid model architecture of a convolutional neural network and a long short-term memory network. The operating parameters and environmental data of multiple consecutive moments in the sample set are constructed into multi-dimensional time series data as input, and the operating parameters and environmental data of the next moment are used as labels to train the state prediction model. A real-time state prediction module is used to obtain the operating parameters and environmental data at multiple consecutive moments including the current moment, and input them into the trained state prediction model to output the operating parameters and environmental data; The optimization strategy generation module is used to define the optimization objective function based on the output of the state prediction model and train the depth according to the optimization objective function. Network models to determine adjustment strategies; The safety assessment and decision-making module is used to collect the current operating parameters and environmental data, calculate the current comprehensive safety score, and compare the current comprehensive safety score with the preset safety threshold to determine whether to continue efficiency optimization or prioritize troubleshooting and repairing problems.

[0029] The present invention further provides an intelligent refrigeration system efficiency optimization device based on deep learning, the device comprising a memory and a processor, the memory storing a computer program that can be run on the processor, and the processor implementing the steps of the above-mentioned method for optimizing the efficiency of an intelligent refrigeration system based on deep learning when executing the computer program.

[0030] The above formulas are all dimensionless and numerical calculations. The formula is a formula for the most recent real situation obtained by collecting a large amount of data and performing software simulation. The preset parameters in the formula are set by technicians in this field according to actual conditions.

[0031] The above embodiments may be implemented in whole or in part by software, hardware, firmware or any other combination thereof. When implemented by software, the above embodiments may be implemented in whole or in part in the form of a computer program product. Those skilled in the art may appreciate that the units and algorithm steps of each example described in conjunction with the embodiments disclosed herein may be implemented by electronic hardware, or a combination of computer software and electronic hardware. Whether these functions are performed by hardware or software methods depends on the specific application and design constraints of the technical solution.

[0032] The units described as separate components may or may not be physically separated, and the components shown as units may or may not be physical units, and may be located in one place or distributed on multiple network units. Some or all of the units may be selected according to actual needs to achieve the purpose of the solution of this embodiment.

[0033] The above description is only a specific implementation manner of the present application, but the protection scope of the present application is not limited thereto. Any technician familiar with the technical field can easily think of changes or substitutions within the technical scope disclosed in the present application, which should be included in the protection scope of the present application.

Claims

1. A method for optimizing the efficiency of an intelligent refrigeration system based on deep learning, characterized in that: The specific steps include: Step 1: Collect the operating parameters and environmental data of the refrigeration system at historical moments and perform preprocessing operations on them. The preprocessed operating parameters and environmental data are integrated into a sample set according to the timestamp. The operating parameters include compressor power, refrigerant flow, condenser temperature, evaporator temperature, system pressure and refrigerant mass flow, and the environmental data include ambient temperature and ambient humidity; Step 2: Build a state prediction model based on the hybrid model architecture of convolutional neural network and long short-term memory network. The operating parameters and environmental data of multiple consecutive moments in the sample set are constructed into multi-dimensional time series data as input, and the operating parameters and environmental data of the next moment are used as labels to train the state prediction model. Step 3: Obtain the operating parameters and environmental data at multiple consecutive moments including the current moment, and input them into the trained state prediction model to output the operating parameters and environmental data; Step 4: Define the optimization objective function based on the output of the state prediction model, and train the depth according to the optimization objective function. Network models to determine adjustment strategies; Step 5: Collect the current operating parameters and environmental data, calculate the current comprehensive safety score, and compare the current comprehensive safety score with the preset safety threshold to determine whether to continue efficiency optimization or prioritize troubleshooting and repairing problems.

2. The method for optimizing the efficiency of an intelligent refrigeration system based on deep learning according to claim 1, characterized in that: The specific logic for collecting the operating parameters and environmental data of the refrigeration system is as follows: A temperature sensor and a humidity sensor are installed at the condenser suction port to collect the ambient temperature and humidity; a power sensor is installed at the power input end of the compressor to collect the compressor power; a refrigerant flow meter is installed on the straight pipe section of the main loop of the refrigerant pipeline to collect the refrigerant flow data; a temperature sensor is installed on the refrigerant pipeline at the condenser outlet to collect the condenser temperature data; a temperature sensor is installed on the refrigerant pipeline at the evaporator inlet to collect the evaporator temperature; a mass flow meter is installed in the compressor outlet pipeline to collect the refrigerant mass flow; pressure sensors are installed at the condenser outlet and the evaporator inlet to collect the condensing pressure and the evaporating pressure; the average pressure of the two is calculated to represent the system pressure, and the formula is as follows: ； in, Indicates the system pressure, Indicates the condensing pressure, Indicates evaporation pressure; The preprocessing step includes removing outliers from the operating parameters and environmental data collected at historical moments, calculating the Z-score of each operating parameter and environmental data, and if the absolute value of the Z-score of the operating parameter and environmental data is , the parameter is regarded as an outlier and removed, and the missing value is filled by the mean filling method, and then the Min-Max normalization is used to normalize the operating parameters and environmental data to Within the range, the preprocessed data are integrated into a sample set.

3. The method for optimizing the efficiency of an intelligent refrigeration system based on deep learning according to claim 2, characterized in that: The specific logic behind the training state prediction model is: The operating parameters and environmental data of multiple consecutive moments in the sample set are constructed into multidimensional time series data as the input of the state prediction model, which is expressed as: ； in, Represents multidimensional time series data, express The compressor power at the time, express The refrigerant flow rate at the moment, express The condenser temperature at time express The evaporator temperature at time express System pressure at all times, express The refrigerant mass flow rate at the moment, express The ambient temperature at the moment, express The ambient humidity at the moment. Similarly, the second row of parameters represents The operating parameters and environmental data at the moment. The last line of parameters indicates The operating parameters and environmental data at the time, t is the time variable, is the time interval between adjacent moments, Indicates the time window size; The operating parameters and environmental data labels of the next moment in the sample set are constructed into multidimensional time data, which can be expressed as: ； in, Represents multidimensional time data, express The compressor power at the time, express The refrigerant flow rate at the moment, express The condenser temperature at time express The evaporator temperature at time express System pressure at all times, express The refrigerant mass flow rate at the moment, express The ambient temperature at the moment, express Ambient humidity at all times; The sample set is divided into a training set and a test set with a ratio of 7:

3. First, the multidimensional time series data in the training set is input into the convolutional neural network to extract local spatial features. Then, the extracted features are input into the LSTM module to learn the dynamic changes of the time series. Finally, the extracted high-order features are mapped to the prediction results of the target value through the fully connected layer to output t+ The operating parameters and environmental data at each moment are input into the data in the test set for testing. The mean absolute error is used as the loss function. If the decrease is less than 0.001 within 10 consecutive rounds of training, the mean absolute error has converged and the training is completed.

4. The method for optimizing the efficiency of an intelligent refrigeration system based on deep learning according to claim 3, characterized in that: The specific logic for determining optimization goals and adjusting strategies is as follows: The optimization objective function is defined as: ； in, In order to optimize the objective function value, the optimization strategy is to minimize the total energy consumption of the compressor while maintaining the comfort of the indoor environment within the optimization time period. Indicates the current moment, that is, the moment when optimization starts. Indicates the time to stop optimization. represents the length of the optimization period, Indicates the current ambient temperature. Indicates the set indoor temperature. are the weight parameters of each item, ,and ; Define the current state as: ； in, Indicates the current state. They are the condenser temperature, evaporator temperature, compressor power, and refrigerant flow at the current moment; Defining actions Action 1: adjust the target temperature of the condenser; Action 2: adjust the output of the compressor power; Action 3: control the refrigerant flow; The reward function is defined as: ； in, represents the reward function; The goal is to train a policy that maximizes the accumulated long-term reward: ； ； in, represents the accumulated long-term reward, represents the discount factor, , Indicated in Instant rewards at all times, represents the index of the time interval within the optimization time, and ; Taking the current state as input, based on the depth The network model uses the Bellman equation to calculate the The value is taken as the output, and the prediction is minimized Values ​​and Goals The error between the values, update depth The parameters of the network are used to complete the model training. Each time the state prediction model outputs the predicted running parameters, the state is determined and input into the depth In the network model, select The action with the largest value is used as the adjustment strategy.

5. The method for optimizing the efficiency of an intelligent refrigeration system based on deep learning according to claim 4, characterized in that: The specific logic for calculating the comprehensive safety score at the current moment is: The single parameter safety score formula for calculating the current operating parameters and environmental data is: ； ； in, Indicates The safety score of each parameter, Indicates The reference value of the parameter, Indicates The half width of the parameter safety range, Indicates The safety lower limit of the parameter, Indicates The safety upper limit of the parameter; The formula used to calculate the overall safety score is: ； in, Represents the comprehensive safety score, For the The weight coefficient of the parameter safety score, Indicates the total number of types of operating parameters and environmental data, is the parameter index, and ; Compare the current comprehensive safety score with the preset safety threshold. If the current comprehensive safety score is higher than the safety threshold, it indicates that the system is in normal operation, and efficiency optimization will continue. If the current comprehensive safety score is lower than the safety threshold, it indicates that there are safety hazards in the system, and the problem will be checked and repaired first before efficiency optimization.

6. An intelligent refrigeration system efficiency optimization system based on deep learning, characterized in that: The intelligent refrigeration system efficiency optimization system based on deep learning is used to implement the intelligent refrigeration system efficiency optimization method based on deep learning according to any one of claims 1 to 5, comprising: The data acquisition and preprocessing module is used to collect the operating parameters and environmental data of the refrigeration system at historical moments and perform preprocessing operations on them. The preprocessed operating parameters and environmental data are integrated into a sample set according to the timestamp. The operating parameters include compressor power, refrigerant flow, condenser temperature, evaporator temperature, system pressure and refrigerant mass flow, and the environmental data include ambient temperature and ambient humidity; The state prediction model building module is used to build a state prediction model based on a hybrid model architecture of a convolutional neural network and a long short-term memory network. The operating parameters and environmental data of multiple consecutive moments in the sample set are constructed into multi-dimensional time series data as input, and the operating parameters and environmental data of the next moment are used as labels to train the state prediction model. A real-time state prediction module is used to obtain the operating parameters and environmental data at multiple consecutive moments including the current moment, and input them into the trained state prediction model to output the operating parameters and environmental data; The optimization strategy generation module is used to define the optimization objective function based on the output of the state prediction model and train the depth according to the optimization objective function. Network models to determine adjustment strategies; The safety assessment and decision-making module is used to collect the current operating parameters and environmental data, calculate the current comprehensive safety score, and compare the current comprehensive safety score with the preset safety threshold to determine whether to continue efficiency optimization or prioritize troubleshooting and repairing problems.

7. An intelligent refrigeration system efficiency optimization device based on deep learning, characterized in that: The device includes a memory and a processor, wherein the memory stores a computer program that can be run on the processor, and the processor implements the steps of the optimization method according to any one of claims 1 to 5 when executing the computer program.

Images