Refrigeration equipment operation energy-saving control method and equipment based on artificial intelligence

Abstract

The invention relates to the technical field of refrigeration equipment control, and discloses an artificial intelligence-based refrigeration equipment operation energy-saving control method and equipment, and the method comprises the following steps: collecting multi-dimensional data in a refrigeration environment in real time through a multi-sensor system, the data including environment variable data and equipment operation data; preprocessing the collected multi-sensor data, and generating a comprehensive environment variable through credibility evaluation and data fusion; carrying out load prediction on the current environment variable and the equipment operation state based on a deep learning model, and predicting a future refrigeration load demand; and according to the load prediction result, an optimization objective function is constructed, and an optimal control strategy of the refrigeration equipment is generated. According to the invention, through a deep learning load dynamic prediction technical scheme, a technical effect of high-precision prediction of a future refrigeration load demand is achieved, so that the change trend of an environment variable and an equipment operation state can be accurately captured, and the reliability of load prediction is improved.

Description

Technical Field

[0001] The present invention relates to the technical field of refrigeration equipment control, and in particular to an artificial intelligence-based refrigeration equipment operation energy-saving control method and equipment. Background Art

[0002] With the continuous increase in global demand for energy conservation and emission reduction and the rapid development of artificial intelligence technology, energy-saving operation control of refrigeration equipment has gradually become the focus of industry research and application. The operation mode of traditional refrigeration equipment often relies on fixed parameter settings or simple empirical formulas for regulation. This method has obvious shortcomings in dynamic and complex environments, such as slow response when environmental variables fluctuate rapidly, excessive energy consumption, and inability to meet user comfort requirements. Therefore, energy-saving control methods for refrigeration equipment operation are needed to improve efficiency; Existing refrigeration equipment usually operates with a combination of multi-sensor monitoring, fixed logic-based control strategies, and offline optimized load forecasting models. These systems collect data such as ambient temperature and humidity, light intensity, and combine them with preset parameters to make a simple forecast of the refrigeration load, and then generate control strategies based on the optimization model to operate the equipment; However, most of the existing operation control methods for refrigeration equipment use linear models or simple empirical formulas, which are difficult to accurately capture the nonlinear dynamic relationship between environmental variables and load demand, especially in complex dynamic scenarios, which show high prediction errors. In addition, multi-sensor data processing lacks credibility assessment and fusion mechanisms. When the sensor data quality is inconsistent or there is a single point failure, the system is prone to misperception, resulting in the failure of the control strategy. The existing optimization objectives are usually single, failing to effectively balance energy saving and user comfort requirements, and are difficult to adapt to diverse application scenarios.

[0003] Therefore, the present invention proposes an energy-saving control method and device for refrigeration equipment operation based on artificial intelligence to solve the deficiencies of the prior art. Summary of the invention

[0004] In view of the shortcomings of the prior art, the present invention provides an energy-saving control method and equipment for the operation of refrigeration equipment based on artificial intelligence, which solves the problem that most of the existing operation control methods for refrigeration equipment adopt linear models or simple empirical formulas, which are difficult to accurately capture the nonlinear dynamic relationship between environmental variables and load demand, especially in complex dynamic scenarios, and exhibit high prediction errors. In addition, multi-sensor data processing lacks credibility assessment and fusion mechanism. When the sensor data quality is inconsistent or there is a single point failure, the system is prone to misperception, resulting in failure of the control strategy.

[0005] To achieve the above objectives, the present invention is implemented through the following technical solutions: an artificial intelligence-based refrigeration equipment operation energy-saving control method, comprising the following steps: Collect multi-dimensional data in the refrigeration environment in real time through a multi-sensor system, the data including environmental variable data and equipment operation data; Preprocessing the collected multi-sensor data to generate comprehensive environmental variables through credibility assessment and data fusion; Based on the deep learning model, the current environmental variables and equipment operating status are used to predict the future cooling load demand; According to the load forecast result, an optimization objective function is constructed to generate an optimal control strategy for the refrigeration equipment, wherein the optimization objective function aims to minimize the environmental comfort deviation and the equipment energy consumption; The optimal control strategy is executed, and the parameters of the deep learning model and the optimization objective function are updated through the feedback data collected in real time to form a dynamic adaptive closed loop.

[0006] Preferably, the multi-sensor system is used to collect multi-dimensional data of the operating environment of the refrigeration equipment, specifically including: The indoor and outdoor temperatures are collected through temperature sensors, the indoor and outdoor humidity is collected through humidity sensors, and the ambient light intensity is collected through light sensors; The real-time power consumption of the equipment is monitored through energy consumption sensors, the refrigerant flow is collected through flow sensors, and the start / stop status and operating frequency of the compressor are monitored through equipment status sensors; The data are collected at fixed time intervals and sent to the control system through the data transmission module for subsequent processing.

[0007] Preferably, the data fusion comprises the following steps: The credibility modeling is performed on the observation values ​​of each sensor, assuming that the data of each sensor follows a normal distribution:

[0008] in, For sensor The output value of is the mean, is the noise variance, Indicates that the measured values ​​follow the mean , the variance is Normal distribution of Optimize the posterior distribution through variational inference and calculate the trust weight of each sensor data; The Dempster-Shafer theory is used to perform credibility fusion on multi-source data and generate comprehensive environmental variables.

[0009] Preferably, the load forecasting is achieved based on a combination of nonlinear dynamic system modeling and deep learning, and the dynamic system model is:

[0010] in, Indicates cooling load Over time The rate of change, is the cooling load demand, For environment variables, Control parameters for device operation. For time, is a random disturbance; The function The cooling load demand in future time periods is predicted through long short-term memory network modeling.

[0011] Preferably, the optimization objective function includes the following objectives: The first target is the environmental comfort deviation, which is the square of the difference between the current environmental state and the target environmental state; The second goal is the real-time energy consumption of equipment operation; The optimization objective function is defined as:

[0012] in, is the objective function value, To optimize the time interval, is the current environment status, is the target environment state, is the control input, is the energy consumption weight.

[0013] Preferably, the optimal control strategy is generated by dynamic programming, which recursively optimizes the optimal value function of the current state based on the Bellman equation:

[0014] in, is the optimal value of the current state, Indicates the current status Take control The immediate cost, Indicates the next state The value function of Indicates that the control variable Take the minimum value.

[0015] Preferably, the deep learning model is dynamically updated and optimized in the following ways: The feedback data collected in real time is used to adjust the model weights online and optimize the accuracy of load forecasting; Dynamically reconstruct the priority weights of input features based on long-term trend changes in environmental variables and equipment operating status; Error analysis is used to adaptively adjust the model structure and improve its ability to model nonlinear and multivariate relationships.

[0016] Preferably, the feedback data includes: Based on the real-time collected equipment operation data and environmental status, dynamically analyze the deviation between the current environmental status and the target environmental status; According to the error value, the environmental comfort weight and energy consumption weight in the optimization objective function are adjusted to improve the adaptability of the control strategy to the energy-saving target; Feedback data into the deep learning model to reallocate the priority of input features and strengthen the model’s response to key variables; Implement multi-level feedback updates and dynamically adjust control parameters, including equipment compressor power, operating hours, and start and stop frequency.

[0017] The present invention also provides a refrigeration device based on artificial intelligence, including a cold storage plate heat exchanger, a cold release plate heat exchanger, a cold storage refrigeration pump, a cold storage submersible pump and a cold release circulation pump; The cold storage plate heat exchanger and the cooling plate heat exchanger are connected to the cold storage refrigeration pump, the cold storage submersible pump and the cooling circulation pump through a pipeline system, and the pipeline system includes: The refrigerant storage pipeline is used to transfer the refrigerant between the cold storage refrigeration pump and the cold storage plate heat exchanger to complete the cold energy storage; The refrigerant release pipeline is used to transfer the refrigerant between the cold storage submersible pump and the cold release plate heat exchanger to complete the cold energy release; The circulation pipeline is used to connect the cooling plate heat exchanger with the target environment under the drive of the cooling circulation pump to achieve cold energy transfer; Pipeline control valves are used to adjust the flow and direction of each pipeline to achieve efficient switching of cold energy between cold storage and cold release equipment.

[0018] The present invention also provides an artificial intelligence-based refrigeration equipment operation energy-saving control system, comprising: Multi-sensor module, used to collect environmental variable data and equipment operation data; Data fusion module, used to conduct credibility assessment and fusion of multi-sensor data to generate comprehensive environmental variables; A deep learning prediction module for predicting future cooling load demand based on comprehensive environmental variables and equipment operating status; An optimization control module, used to generate an optimal control strategy for equipment operating parameters based on an optimization objective function; The feedback module is used to dynamically update the deep learning model and optimize the control parameters based on real-time monitoring data to achieve closed-loop control.

[0019] The present invention provides an artificial intelligence-based refrigeration equipment operation energy-saving control method and device. It has the following beneficial effects: 1. The present invention achieves a high-precision prediction technology effect for future refrigeration load demand through a deep learning load dynamic prediction technology solution. Compared with the technical solutions in the prior art that rely on simple linear models or fixed parameter models, the problem of insufficient modeling capabilities for complex nonlinear dynamic relationships is solved, thereby accurately capturing the changing trends of environmental variables and equipment operating status, and improving the reliability of load prediction.

[0020] 2. The present invention achieves the technical effect of high-reliability processing and fusion of multi-dimensional environmental variable data and equipment operation status data by adopting the multi-sensor data fusion technical solution of Dempster-Shafer theory. Compared with the technical solution of single sensor or simple weighted average method in the prior art, it solves the shortcomings of low data reliability and susceptibility to single-point sensor failure, thereby enhancing the system's adaptability to complex dynamic environments.

[0021] 3. The present invention achieves the technical effect of adjusting the control strategy in real time and adapting to the rapid changes in the environment by adopting the dynamic adaptive optimization technical solution of the feedback mechanism. Compared with the technical solutions of static control or fixed parameter control in the prior art, it solves the problems of delayed response to sudden environmental changes and insufficient control accuracy, thereby realizing efficient and energy-saving operation of refrigeration equipment in a dynamic environment.

[0022] 4. The present invention achieves the comprehensive optimization technical effect of taking into account both environmental comfort and equipment energy consumption by adopting a control strategy generation technical solution based on a weighted optimization objective function. Compared with the technical solution in the prior art that only focuses on a single goal (such as energy saving or comfort), it solves the problem of insufficient ability to balance multiple goals, thereby better meeting user needs and improving the overall operating performance of the system. BRIEF DESCRIPTION OF THE DRAWINGS

[0023] Figure 1 is a flow chart of the method steps of the present invention; Figure 2 It is a system architecture diagram of the present invention. DETAILED DESCRIPTION

[0024] The following will be combined with the drawings in the specification of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0025] The present invention provides an artificial intelligence-based energy-saving control method and equipment for refrigeration equipment operation. Through technical means such as multi-sensor data acquisition and fusion, deep learning load prediction, optimization objective function construction and dynamic feedback control, dynamic optimization of energy consumption and precise control of environmental comfort during the operation of refrigeration equipment are achieved, thereby improving the energy-saving efficiency and intelligent operation level of the refrigeration equipment.

[0026] Please see attached Figure 1 The artificial intelligence refrigeration equipment operation energy-saving control method of the present invention may include the following steps.

[0027] S1. Multi-sensor data acquisition and preprocessing; S2. Interactive verification and fusion of multi-sensor data; S3. Dynamic prediction of cooling load based on deep learning; S4. Optimize the objective function to generate the optimal control strategy for the refrigeration equipment; S5. Execute control strategies and perform dynamic adaptive optimization based on real-time feedback.

[0028] For step S1, real-time collection of environmental variables and equipment operation data lays the foundation for subsequent data fusion, load forecasting and optimization control. Multi-dimensional data with high temporal and spatial resolution is obtained through multi-sensor deployment and systematic collection strategy to accurately describe the operating environment and operating status of the refrigeration equipment.

[0029] Generally, the data collection needs to ensure stability and accuracy under multivariable and complex conditions. The collected data should include environmental variable data and equipment operation data, and the data content should cover the key parameters of refrigeration equipment operation. In this embodiment, the data collection frequency and transmission mechanism design should match the subsequent data fusion and processing modules to ensure the real-time and integrity of data transmission.

[0030] In this embodiment, multi-dimensional data in the operating environment of the refrigeration equipment is collected by deploying multiple types of sensor modules. The sensor modules include but are not limited to the following types: temperature sensors for collecting indoor and outdoor temperature data; humidity sensors for monitoring indoor and outdoor humidity; light sensors for detecting ambient light intensity; flow sensors for monitoring the real-time flow of refrigerant; energy consumption sensors for recording the real-time power consumption of the refrigeration equipment; equipment status sensors for monitoring the start and stop status, operating frequency, etc. of the refrigeration equipment (such as a compressor).

[0031] Generally speaking, environmental variable data include indoor and outdoor temperature and humidity, ambient light intensity, etc., which are used to reflect the external environmental conditions for the operation of refrigeration equipment; equipment operation data include compressor power, refrigerant flow, etc., which are used to reflect the current operating status of the equipment.

[0032] As an option, the deployment location of the sensors can be optimized according to the application scenario, for example: In a data center environment, temperature sensors can be placed at the air inlets and outlets of server cabinets to monitor air temperature distribution; In building environments, humidity sensors and light sensors can be placed on different floors and near windows to capture the diversity of environmental changes.

[0033] The frequency of data collection can be adjusted according to needs. In some embodiments, the collection frequency can be set to 1 second to 10 seconds to achieve high-resolution data collection. In another possible implementation, for parameters with slower environmental changes (such as light intensity), the collection frequency can be reduced to once per minute to save storage and transmission resources.

[0034] Specifically, in order to ensure the quality of the collected data, the raw data needs to be preprocessed in the present invention. Generally, the data preprocessing steps mainly include: Outlier processing: Eliminate outliers or observations with large data deviations. For example, data with indoor temperatures exceeding 50°C or below 0°C will be considered outliers and corrected by elimination or interpolation.

[0035] Data interpolation: When some sensors are missing data due to failure or network fluctuations, linear interpolation or historical mean is used to fill in the missing data to ensure data integrity.

[0036] Data normalization: For data of multiple dimensions (such as temperature, humidity, and light intensity), they are uniformly converted to the [0,1] interval. The normalization formula is:

[0037] in, is the normalized data, is the original data, and are the minimum and maximum values ​​of the variable, respectively.

[0038] As an implementation, sensor data is delivered via the following transport mechanisms: The data collected by the sensors are transmitted to the central control system through the IoT module (such as wireless communication module or industrial Ethernet).

[0039] In some embodiments, data can be transmitted over long distances via 5G communications or LoRa networks to accommodate the environmental requirements of large-scale distributed refrigeration equipment.

[0040] Specifically, the data transmission protocol can choose a lightweight protocol based on MQTT or HTTP to reduce data transmission latency and bandwidth usage.

[0041] Appropriate redundancy mechanisms need to be adopted during transmission to ensure data integrity, such as: In one possible implementation, a data packet verification mechanism may be designed to detect the accuracy of the transmitted data; In another implementation, when data is lost, the loss can be compensated by resending the data packet from the local cache.

[0042] For step S2, the multi-sensor data is interactively verified and fused. Through credibility assessment, multi-source fusion algorithm and mathematical derivation, this step can generate comprehensive environmental variables with higher confidence, providing high-quality input for subsequent steps. In this step, the collected multi-sensor data is first preprocessed. In general, there may be a certain proportion of abnormal values ​​in the sensor collected data, such as extreme data caused by sensor failure or missing data caused by network delay. In view of this situation, the following steps are used in this embodiment to preprocess the data: The historical mean and standard deviation of each type of sensor data collected are calculated, and a normal data range is established based on this; when the collected value exceeds this range, it is marked as an outlier and removed.

[0043] As an option for missing data that occurs during the collection process, a missing data filling strategy based on interpolation can be adopted, such as estimating missing values ​​using the linear interpolation method of the previous and next values ​​of the time series.

[0044] After preprocessing, this embodiment introduces a credibility evaluation mechanism to model and verify sensor data. Specifically, in a possible implementation, it is assumed that the data of each sensor follows a normal distribution:

[0045] in, For sensor The output value of is the mean, is the noise variance, Indicates that the measured values ​​follow the mean , the variance is The normal distribution of .

[0046] By statistically analyzing the historical data of the sensor, the mean and noise variance of each sensor are dynamically estimated to evaluate the credibility of its data. As an option, the credibility can be calculated by the following formula:

[0047] in, For sensor The credibility weight of is the noise variance. Generally speaking, the smaller the noise variance, the higher the reliability of the sensor.

[0048] After completing the credibility assessment, the present embodiment further utilizes the Dempster-Shafer theory (DS theory) to fuse the multi-sensor data. Specifically, for different data sources provided by multiple sensors, the DS theory generates the final comprehensive environmental variables by integrating the confidence of multiple sensors.

[0049] Specifically, in one possible implementation, assuming Indicates sensor About the event Confidence (i.e., supporting events The confidence after multi-sensor fusion can be calculated by the following formula:

[0050] in is the comprehensive confidence after fusion, and represent subsets of the event space, and Sensor 1 and sensor 2 are the events and The confidence level, Represents two propositions and The intersection of .

[0051] As a possible implementation, the DS theory fusion process can be extended to the data integration of more than two sensors. In this case, the confidence of each sensor can be gradually fused in a recursive manner to finally obtain a comprehensive high-confidence environmental variable.

[0052] In some embodiments, for sensor data with time series characteristics, such as temperature and humidity data and device power consumption data, the data fusion results can be dynamically smoothed in combination with the time series analysis method to further reduce the impact of noise. For example, the Kalman filter method is used to dynamically optimize the fused environmental variables, and the update formula is as follows:

[0053]

[0054] in, For the moment The final estimate of is the predicted value, is the Kalman gain, is the current observation value, is the observation matrix, is the prediction covariance, is the observation noise covariance.

[0055] For step S3, a deep learning method based on nonlinear dynamic system modeling and long short-term memory network (LSTM) is used to fit the complex dynamic relationship of refrigeration load changes, thereby achieving accurate load demand prediction.

[0056] In this embodiment, the cooling load requirement is described as the following nonlinear dynamic system:

[0057] in, Indicates cooling load Over time The rate of change, is the cooling load demand, For environment variables, Control parameters for device operation. For time, is a random disturbance, It is a dynamic relationship function.

[0058] In one possible implementation, the function The specific form of is fitted by the long short-term memory network (LSTM).

[0059] In this embodiment, the structure and implementation of the LSTM network are as follows: As an option, the design of the LSTM network combines the time series characteristics of load demand and the dynamic interaction relationship of multi-dimensional input variables. Specifically, the input data of the network includes:

[0060] Historical environment variable sequence ; Historical equipment operating status sequence ; The output of the network is the load demand forecast for a period of time in the future. .

[0061] Generally, the input variables have high dimensions, so the input layer is required to perform feature extraction and dimensionality reduction on the raw data. As an implementation method, an embedding layer can be used to vectorize each input variable.

[0062] In the specific implementation, the core part of the LSTM network is composed of multiple layers of LSTM units. The LSTM unit uses the following recursive relationship to update the state:

[0063] in, is the hidden state at the current moment, is the input data at the current moment, is the weight matrix of the LSTM unit, , , is the bias of the LSTM unit, is the activation function (usually the Sigmoid function), is the hyperbolic tangent function.

[0064] In some embodiments, in order to improve the learning ability of the network, a multi-layer LSTM structure can be added, and a fully connected layer can be connected at the end to output the final prediction result.

[0065] In this embodiment, in order to ensure the prediction accuracy, the LSTM network needs to be trained and verified. In general, the training goal is to minimize the prediction error, and the loss function is defined as:

[0066] in, represents the loss function value, is the total number of training samples, is the load value predicted by the model, Indicates the actual load value, Represents the square of the error between the predicted value and the true value.

[0067] In one possible implementation, the network parameters can be iteratively optimized using stochastic gradient descent (SGD) or an adaptive optimization algorithm such as Adam.

[0068] As an option, in order to avoid the overfitting problem, a Dropout layer can be added to the network to randomly discard the output of a portion of the hidden units. In addition, the best hyperparameters (such as learning rate, number of hidden layer units, etc.) can be selected through cross-validation.

[0069] The output of the LSTM network is the load demand forecast for a period of time in the future. In general, the prediction duration It can be flexibly adjusted according to specific application scenarios. For example: In the short-term forecast scenario, select minutes, used to optimize equipment operating parameters in real time; In the medium- and long-term forecast scenario, select hours or longer for developing equipment start and stop plans and energy allocation strategies.

[0070] The prediction results will be directly used to optimize the construction of the objective function and provide support for the generation of equipment operation control strategies.

[0071] For step S4, the comprehensive environmental variables generated by multi-sensor data fusion are combined with the operating status of the refrigeration equipment to dynamically predict the future refrigeration load demand using a deep learning model. This step follows the previous step and realizes forward-looking estimation of load demand by modeling the time series of environmental variables and equipment status, thereby providing a data basis for optimizing the control strategy. The accuracy of load prediction directly affects the energy-saving effect and operating stability of refrigeration equipment. Therefore, in the present invention, a technical solution combining nonlinear dynamic system modeling with long short-term memory network (LSTM) is adopted to perform high-precision modeling and prediction of dynamic loads.

[0072] In this embodiment, the optimization objective function is defined as:

[0073] in, is the objective function value, To optimize the time interval, is the current environment status, is the target environment state, is the control input, is the energy consumption weight.

[0074] In general, the target environment state It can be determined by user preset values, such as the target temperature is set to 22°C and the target humidity is set to 50%. As an option, the energy consumption weight It can be adjusted dynamically according to the operation scenario, such as increasing the power consumption appropriately during peak hours when energy is scarce. value to reduce the energy consumption of the device.

[0075] Specifically, the first term in the optimization objective function Represents the square of the deviation between the current environmental state and the target state, which is used to quantify the loss of environmental comfort; the second term It represents the control cost of equipment operation and is positively correlated with the real-time energy consumption of the equipment.

[0076] In this embodiment, the optimization objective function is solved by a dynamic programming method. The core idea of ​​dynamic programming is to decompose the global optimization problem into a series of local optimal problems and solve them step by step in a recursive manner.

[0077] In one possible implementation, dynamic programming is implemented based on the Bellman equation, and its recursive relationship is:

[0078] in, is the optimal value of the current state, Indicates the current status Take control The immediate cost, Indicates the next state The value function of Indicates that the control variable Take the minimum value.

[0079] As an implementation approach, the state transition equation can be modeled as:

[0080] in, Indicates that the system is at time The state vector at time, and is the state transfer matrix of the system; is the random noise term, System in time The state vector at time, represents the control input vector.

[0081] Specifically, the recursive solution process starts from the end point and works backwards. For example, assuming the end point cost , and then gradually calculate the optimal value function of the previous moment , Until the starting moment.

[0082] As an alternative, the dynamic programming solution can be solved by discretizing the time step. For example, Divided into discrete moments , optimize the control input at each discrete moment .

[0083] In this embodiment, the generated control strategy includes the following contents: The control strategy generated through optimization accurately adjusts the real-time power output of the compressor to adapt to the predicted load demand.

[0084] Determine the optimal start and stop times for the equipment to avoid the impact of frequent starts and stops on equipment life and energy consumption.

[0085] By adjusting the refrigerant flow, accurate matching of refrigeration capacity can be achieved to avoid over-cooling or under-cooling.

[0086] As an implementation method, the optimal control strategy can be sent to the refrigeration equipment execution unit, including the compressor, condenser, evaporator and other modules, through the controller in the form of a command signal.

[0087] For step S5, by real-time monitoring of the operating status and environmental changes of the refrigeration equipment, executing control instructions and combining the feedback mechanism to achieve dynamic adaptive optimization. Generally, during the operation of the refrigeration equipment, the actual operating effect may deviate from the expected due to rapid fluctuations in the external environment or changes in the equipment status. Therefore, the control strategy is dynamically adjusted through the feedback mechanism to ensure that the energy-saving effect and environmental comfort of the equipment operation can be maintained within the target range for a long time.

[0088] In this embodiment, the real-time feedback data comes from the multi-sensor system and the equipment operation monitoring module, including the following two categories: Environmental status feedback data: comprehensive environmental variables generated by the current ambient temperature and humidity collected by the temperature and humidity sensors and the changes in ambient light intensity monitored by the light sensors.

[0089] Equipment operation status feedback data: monitor the current real-time power consumption of the equipment through energy consumption sensors, collect refrigerant flow through flow sensors, compressor operation frequency, start and stop status, etc.

[0090] As an option, feedback data is collected at fixed time intervals (e.g., 5 seconds) and transmitted to the central control unit for processing via the IoT module. In another possible implementation, the data collection frequency can be dynamically adjusted according to the equipment operation requirements, such as increasing the collection frequency when the environment fluctuates rapidly.

[0091] Specifically, the feedback data is first preprocessed, including noise filtering, data smoothing, and outlier removal. For example, when the temperature and humidity sensor has extreme data in a short period of time, the data curve can be smoothed by the moving average method.

[0092] In this embodiment, the error analysis and control strategy adjustment are implemented as follows: Real-time feedback data is used to calculate the error between the current environment state and the target environment state:

[0093] in, is the error at the current moment, is the current environment state, generated by feedback data, The target environment state.

[0094] As a possible implementation, error analysis can be further divided into short-term error and long-term trend error. For example, short-term error is used to quickly respond to current environmental changes, while long-term trend error reflects the global adaptability of the control strategy.

[0095] Based on the error calculation, the controller gain parameter Dynamic adjustments are made to ensure that the controller has an appropriate response strength to different types of errors. Specifically, the output of the controller can be described as:

[0096] in, Indicates the control quantity or control output, Proportional gain, used to quickly respond to the current error, Represents the error at the current moment, Integral gain, used to eliminate long-term errors, Derivative gain, used to predict the error trend, Indicates the rate of change of error.

[0097] As an implementation method, the above control algorithm can be further optimized by adaptively adjusting the gain based on feedback, for example, appropriately increasing the gain when the external interference is large. To improve the system's anti-interference ability.

[0098] In this embodiment, during long-term operation, the weight parameters in the optimization objective function can be dynamically modified according to feedback data. Specifically: energy consumption weight Adjustment, appropriately increase the power consumption during peak hours , give priority to reducing equipment operating energy consumption; reduce , appropriately increase the priority of environmental comfort.

[0099] As an extension method, the energy consumption weight can be dynamically adjusted according to the real-time electricity price data to achieve further cost optimization.

[0100] In this embodiment, the core of executing the control strategy is to convert the optimized control instructions into the actual operating parameters of the equipment, mainly including: adjusting the compressor power to match the real-time load demand; optimizing the refrigerant flow distribution to maintain the refrigeration effect; controlling the start and stop timing of the equipment to avoid the impact of frequent start and stop on the equipment life.

[0101] As an implementation method, control instructions are transmitted to equipment execution units such as compressors, condensers, evaporators, etc. through PLC (Programmable Logic Controller). In order to ensure the execution effect of control instructions, feedback data can be used to verify whether the equipment status meets expectations; As the core execution unit of the control strategy, PLC (Programmable Logic Controller) is responsible for converting the control instructions generated by the optimization control module into operational signals that can be executed by the equipment. It provides flexible logic programming capabilities and can respond to dynamically adjusted control instructions in real time; the equipment execution unit includes key components of refrigeration equipment such as compressors, condensers, and evaporators, receives control signals from PLC (Programmable Logic Controller), and specifically performs tasks such as power adjustment, refrigerant flow optimization, and start-stop control; PLC (Programmable Logic Controller) is connected to the equipment execution unit through a wired or wireless signal transmission network to ensure the rapid transmission and accurate execution of instructions. The equipment execution unit usually contains drive circuits and actuators (such as motors, valves, etc.), which can convert control signals into actual physical operations.

[0102] PLC (Programmable Logic Controller) is the core coordination unit, connecting the optimization control module, equipment execution unit, and sensor feedback module. Each module realizes efficient data transmission through the Internet of Things to form a closed-loop control. In this way, the control strategy is converted into specific operating instructions of the equipment through PLC (Programmable Logic Controller), and the state verification and dynamic adjustment are realized in combination with sensor feedback. The whole process builds a closed-loop control system to ensure that the operation effect of the refrigeration equipment can meet the energy-saving and environmental comfort goals in real time, while effectively responding to the impact of changes in the external environment or fluctuations in the equipment state.

[0103] The refrigeration equipment operation energy-saving control system based on artificial intelligence described below and the refrigeration equipment operation energy-saving control method based on artificial intelligence described above can be referenced to each other.

[0104] Please see attached Figure 2 The present invention also provides an energy-saving control system for refrigeration equipment operation based on artificial intelligence, which combines multi-sensor data acquisition and fusion technology, deep learning prediction technology, optimization objective function construction technology and dynamic feedback control technology. Through the coordinated operation of the multi-sensor module, data fusion module, deep learning prediction module, optimization control module and feedback module in the system module, the system can collect and process environmental variables and equipment operation data in real time, dynamically predict refrigeration load requirements, generate optimal control strategies, and make adaptive adjustments based on real-time feedback to achieve precise control and efficient energy saving of refrigeration equipment operation. The system effectively balances the relationship between environmental comfort and equipment energy consumption, and provides reliable technical support for energy-saving operation of refrigeration equipment in complex dynamic environments.

[0105] The multi-sensor module is used to collect multi-dimensional environmental data and equipment status data of refrigeration equipment in real time. The module supports the collaborative work of multiple types of sensors to achieve comprehensive perception of environmental variables (such as temperature and humidity, light intensity) and equipment operating status (such as compressor start and stop status, refrigerant flow, real-time energy consumption).

[0106] In this embodiment, the multi-sensor module includes a temperature sensor, a humidity sensor, a light sensor, an energy consumption sensor, a flow sensor, etc. The sensor is connected to the central control unit via an industrial communication protocol (such as Modbus or CAN bus), and the acquisition frequency of each sensor can be dynamically adjusted according to demand. For example, when environmental variables change rapidly, the sensor acquisition frequency is increased to ensure the timeliness of the data.

[0107] Generally speaking, the sensor acquisition frequency can be related to the rate of change of the variable. The basic form of the formula is as follows:

[0108] in: is the current sensor acquisition frequency, is the minimum acquisition frequency of the sensor, is the adjustment coefficient, is the change in the environment variable, The time interval between variable changes (unit: seconds).

[0109] The deployment of multi-sensor modules enables comprehensive monitoring of the environment and equipment operating status, providing high-quality raw data input for data fusion and deep learning prediction.

[0110] The data fusion module performs credibility assessment, noise filtering and information fusion on the raw data collected by multiple sensors to generate comprehensive environmental variable data, providing input for subsequent load forecasting and optimization control.

[0111] In this embodiment, the data fusion module first models the credibility of the sensor data, assuming that the observations of each sensor obey a certain statistical distribution, and combines the observations of multiple sensors through a weighted average method. Subsequently, a fusion algorithm based on evidence theory is used to further improve the accuracy of the data. For example, the Dempster-Shafer theory can be used to process conflicting information output by different sensors and generate a highly credible comprehensive variable.

[0112] The data fusion module effectively improves the reliability and consistency of multi-source data, solves the observation error problem that may exist in a single sensor, and provides basic support for the stable operation of the system.

[0113] The deep learning prediction module is used to predict the cooling load demand in the future based on the integrated environmental variables and equipment operation status data, providing guidance for the optimization control module.

[0114] In this embodiment, the deep learning prediction module uses a long short-term memory network (LSTM) to dynamically model the cooling load. The input features include the fused historical environmental variable sequence and the equipment operation status sequence, and the network output is the future cooling load forecast value. Model training is optimized offline through historical operation data, and online parameter adjustment is performed according to real-time feedback data in actual operation.

[0115] The deep learning prediction module significantly improves the accuracy of load forecasting, reduces the system's response delay to external changes, and provides precise load input for optimized control.

[0116] The optimization control module constructs the optimization objective function based on the predicted load demand and generates the optimal control strategy for the refrigeration equipment to ensure a balance between environmental comfort and equipment energy saving.

[0117] In this embodiment, the optimization objective function is constructed with the environmental comfort deviation and the equipment operation energy consumption as weights, and the optimal control strategy is generated by combining the dynamic programming algorithm. For example, by recursively solving the optimization problem, the dynamic programming method gradually determines the optimal equipment operation parameters (such as compressor power, refrigerant flow, and equipment start and stop timing) for each time step. During the optimization process, the energy consumption weight of the objective function can be dynamically adjusted according to real-time feedback data.

[0118] The objective function is defined as At that moment), minimize the weighted sum of environmental comfort deviation and equipment operation energy consumption:

[0119] in: is the total objective function value, is the time step, is the environmental comfort deviation, The energy consumption of the equipment operation is , is the weight of environmental comfort deviation and energy consumption.

[0120] The optimization control module generates precise control strategies to avoid over- or under-operation of equipment, while significantly reducing operating energy consumption.

[0121] The feedback module is used to collect the environmental status and control effect of the equipment in real time, analyze the deviation between the actual operating status and the target status, and dynamically optimize the control strategy through an adaptive mechanism.

[0122] In this embodiment, the feedback module calculates the error value between the current state and the target state by monitoring the sensor data and the operating status of the equipment, and adjusts the parameters of the control strategy. For example, the controller gain parameters (such as proportional gain, integral gain and differential gain) are dynamically adjusted according to the deviation value to optimize the control response speed and stability. In addition, the feedback module can also retrain the deep learning model based on long-term feedback data to improve the prediction accuracy.

[0123] The setting of the feedback module enables the system to have adaptive capabilities, and can quickly adjust the operating strategy when environmental conditions change, ensuring stable and efficient operation of the equipment in a dynamic environment.

[0124] The system of this embodiment can be used to execute the above method embodiments, and its principles and technical effects are similar, which will not be repeated here.

[0125] The present invention also provides a refrigeration device based on artificial intelligence, including a cold storage plate heat exchanger, a cold release plate heat exchanger, a cold storage refrigeration pump, a cold storage submersible pump and a cold release circulation pump; The cold storage plate heat exchanger and the cooling plate heat exchanger are connected to the cold storage refrigeration pump, the cold storage submersible pump and the cooling circulation pump through a pipeline system. The pipeline system includes: The refrigerant storage pipeline connects the cold storage refrigeration pump and the cold storage plate heat exchanger, and is used to transfer the refrigerant between the cold storage refrigeration pump and the cold storage plate heat exchanger to complete the cold energy storage; when the refrigeration equipment is in a low-load operation state, the excess cold energy is transferred to the cold storage device through the refrigerant for storage; it is equipped with temperature monitoring points and pressure sensors to monitor the temperature and pressure state of the refrigerant flow in real time to ensure the safety and efficiency of the energy storage process; The refrigerant release pipeline connects the cold storage submersible pump and the cooling plate heat exchanger, and is used to transmit refrigerant between the cold storage submersible pump and the cooling plate heat exchanger to complete the cold energy release; it is used to release the cold energy stored in the cold storage device to the target environment as needed to achieve dynamic adjustment of the ambient temperature; it is equipped with flow sensors and energy consumption monitoring equipment to monitor the refrigerant release rate and equipment energy consumption status to optimize the cold energy release efficiency.

[0126] The circulation pipeline connects the cooling plate heat exchanger and the target environment. It is used to connect the cooling plate heat exchanger and the target environment under the drive of the cooling circulation pump to realize the transfer of cold energy. The cooling circulation pump drives the refrigerant to circulate between the equipment and the environment, continuously providing cold energy to maintain the ambient temperature. The pipeline is equipped with a smooth coating to reduce flow resistance, and a reflux valve is set to improve the utilization efficiency of the refrigerant and reduce energy consumption.

[0127] The pipeline control valve is equipped with multiple electronic control valves to adjust the flow and direction of each pipeline to achieve efficient switching of cold energy between cold storage and cold release equipment; it is used to dynamically adjust the refrigerant flow and direction of each pipeline according to the control strategy generated by the optimization control module; the control valve is linked with the multi-sensor module to adjust the priority path of the refrigerant flow through real-time data feedback to improve the response speed of the pipeline system.

[0128] Although embodiments of the present invention have been shown and described, it will be appreciated by those skilled in the art that various changes, modifications, substitutions and variations may be made to the embodiments without departing from the principles and spirit of the present invention, and that the scope of the present invention is defined by the appended claims and their equivalents.

Claims

1. A refrigeration equipment operation energy-saving control method based on artificial intelligence, characterized in that: The following steps are involved: Collect multi-dimensional data in the refrigeration environment in real time through a multi-sensor system, the data including environmental variable data and equipment operation data; Preprocessing the collected multi-sensor data to generate comprehensive environmental variables through credibility assessment and data fusion; Based on the deep learning model, the current environmental variables and equipment operating status are used to predict the future cooling load demand; According to the load forecast result, an optimization objective function is constructed to generate an optimal control strategy for the refrigeration equipment, wherein the optimization objective function aims to minimize the environmental comfort deviation and the equipment energy consumption; The optimal control strategy is executed, and the parameters of the deep learning model and the optimization objective function are updated through the feedback data collected in real time to form a dynamic adaptive closed loop.

2. The method for energy-saving control of refrigeration equipment operation based on artificial intelligence according to claim 1 is characterized in that: The multi-sensor system is used to collect multi-dimensional data of the operating environment of the refrigeration equipment, specifically including: The indoor and outdoor temperatures are collected through temperature sensors, the indoor and outdoor humidity is collected through humidity sensors, and the ambient light intensity is collected through light sensors; The real-time power consumption of the equipment is monitored through energy consumption sensors, the refrigerant flow is collected through flow sensors, and the start / stop status and operating frequency of the compressor are monitored through equipment status sensors; The data are collected at fixed time intervals and sent to the control system through the data transmission module for subsequent processing.

3. The refrigeration equipment operation energy-saving control method based on artificial intelligence according to claim 1 is characterized in that: The data fusion comprises the following steps: The credibility modeling is performed on the observation values ​​of each sensor, assuming that the data of each sensor follows a normal distribution: ； in, For sensor The output value of is the mean, is the noise variance, Indicates that the measured values ​​follow the mean , the variance is Normal distribution of Optimize the posterior distribution through variational inference and calculate the trust weight of each sensor data; The Dempster-Shafer theory is used to perform credibility fusion on multi-source data and generate comprehensive environmental variables.

4. The method for energy-saving control of refrigeration equipment operation based on artificial intelligence according to claim 1 is characterized in that: The load forecasting is achieved based on the combination of nonlinear dynamic system modeling and deep learning. The dynamic system model is: ； in, Indicates cooling load Over time The rate of change, is the cooling load demand, For environment variables, Control parameters for device operation. For time, is a random disturbance; The function The cooling load demand in future time periods is predicted through long short-term memory network modeling.

5. The method for energy-saving control of refrigeration equipment operation based on artificial intelligence according to claim 1 is characterized in that: The optimization objective function includes the following objectives: The first target is the environmental comfort deviation, which is the square of the difference between the current environmental state and the target environmental state; The second goal is the real-time energy consumption of equipment operation; The optimization objective function is defined as: ； in, is the objective function value, To optimize the time interval, is the current environment status, is the target environment state, is the control input, is the energy consumption weight.

6. The method for energy-saving control of refrigeration equipment operation based on artificial intelligence according to claim 1 is characterized in that: The optimal control strategy is generated by dynamic programming, which recursively optimizes the optimal value function of the current state based on the Bellman equation: ； in, is the optimal value of the current state, Indicates the current status Take control The immediate cost, Indicates the next state The value function of Indicates that the control variable Take the minimum value.

7. The method for energy-saving control of refrigeration equipment operation based on artificial intelligence according to claim 1 is characterized in that: The deep learning model is dynamically updated and optimized in the following ways: The feedback data collected in real time is used to adjust the model weights online and optimize the accuracy of load forecasting; Dynamically reconstruct the priority weights of input features based on long-term trend changes in environmental variables and equipment operating status; Error analysis is used to adaptively adjust the model structure and improve its ability to model nonlinear and multivariate relationships.

8. The method for energy-saving control of refrigeration equipment operation based on artificial intelligence according to claim 1 is characterized in that: The feedback data includes: Based on the real-time collected equipment operation data and environmental status, dynamically analyze the deviation between the current environmental status and the target environmental status; According to the error value, the environmental comfort weight and energy consumption weight in the optimization objective function are adjusted to improve the adaptability of the control strategy to the energy-saving target; Feedback data into the deep learning model to reallocate the priority of input features and strengthen the model’s response to key variables; Implement multi-level feedback updates and dynamically adjust control parameters, including equipment compressor power, operating hours, and start and stop frequency.

9. An artificial intelligence-based refrigeration device, applied to the artificial intelligence-based refrigeration device operation energy-saving control method according to any one of claims 1 to 8, characterized in that: Including cold storage plate heat exchanger, cooling plate heat exchanger, cold storage refrigeration pump, cold storage submersible pump and cooling circulation pump; The cold storage plate heat exchanger and the cooling plate heat exchanger are connected to the cold storage refrigeration pump, the cold storage submersible pump and the cooling circulation pump through a pipeline system, and the pipeline system includes: The refrigerant storage pipeline is used to transfer the refrigerant between the cold storage refrigeration pump and the cold storage plate heat exchanger to complete the cold energy storage; The refrigerant release pipeline is used to transfer the refrigerant between the cold storage submersible pump and the cold release plate heat exchanger to complete the cold energy release; The circulation pipeline is used to connect the cooling plate heat exchanger with the target environment under the drive of the cooling circulation pump to achieve cold energy transfer; Pipeline control valves are used to adjust the flow and direction of each pipeline to achieve efficient switching of cold energy between cold storage and cold release equipment.

10. An artificial intelligence-based refrigeration equipment operation energy-saving control system, applied to an artificial intelligence-based refrigeration equipment operation energy-saving control method according to any one of claims 1 to 8, characterized in that: include: Multi-sensor module, used to collect environmental variable data and equipment operation data; Data fusion module, used to conduct credibility assessment and fusion of multi-sensor data to generate comprehensive environmental variables; A deep learning prediction module for predicting future cooling load demand based on comprehensive environmental variables and equipment operating status; An optimization control module, used to generate an optimal control strategy for equipment operating parameters based on an optimization objective function; The feedback module is used to dynamically update the deep learning model and optimize the control parameters based on real-time monitoring data to achieve closed-loop control.

Images