Efficient and energy-saving type intelligent management and control system and method for integrated cold station

Abstract

The invention discloses an efficient and energy-saving type intelligent management and control system and method for an integrated cold station, and particularly relates to the technical field of intelligent management and control of cold stations. A plurality of sensors are arranged in a cold station system, the operation state of each component and the external environment change are monitored in real time, historical data and external environment information are combined, and a load fluctuation prediction model is established by adopting a long-short-term memory network, so that the load demand is accurately predicted, and the operation parameters of each component of the system are adjusted in advance; meanwhile, on the basis of real-time monitoring data and load prediction deviation, dynamic optimization is conducted on the cold station system through fuzzy logic, and therefore the equipment response speed and the energy-saving efficiency are balanced; abnormal detection is carried out on the system through a machine learning model, potential fault risks are found in time, and early warning is given out; the system can quickly and accurately adjust the operation state when the load fluctuates, the stability, the energy efficiency and the fault prevention capability of the cold station system are improved, and the overall operation efficiency and the energy-saving effect are remarkably improved.

Description

Technical Field

[0001] The present invention relates to the technical field of intelligent control of cold stations, and particularly relates to an intelligent control system and method for an energy-efficient integrated cold station in a complete set. Background Art

[0002] With the increasingly serious problem of energy consumption and the continuous improvement of environmental protection requirements, the demand for energy-efficient cooling systems has become more urgent. Traditional cold station systems often rely on fixed operating parameters and simple control methods, resulting in energy waste and low operating efficiency. Especially in large-scale commercial and industrial applications, the energy efficiency optimization of cold stations has become a key technical problem. To solve this problem, integrated cold station systems have emerged, which reduce space occupancy by integrating multiple components and adjust the operating status of equipment in real time through an intelligent control system, so as to achieve the goal of energy conservation and high efficiency. However, the current intelligent control technology has not been able to fully and effectively meet the energy efficiency optimization requirements under different load conditions. Especially in the case of complex dynamic loads and environmental changes, how to achieve precise control is still a technical problem that needs to be overcome urgently.

[0003] The existing technologies have the following deficiencies: In the existing intelligent control system for energy-efficient cold stations, when the system responds to sudden load fluctuations, it is difficult to balance the equipment response speed and energy conservation efficiency. Especially in the case of a large range of rapid load changes, traditional control systems often have a situation of response lag or over-regulation, which not only reduces the efficiency of the cold station system, but also may cause a significant increase in energy consumption, ultimately affecting the energy-saving effect of the system. Therefore, how to quickly and accurately adjust the operating status of each component during sudden load changes to ensure the stable and efficient operation of the system has become a major challenge for current technologies. Summary of the Invention

[0004] The purpose of the present invention is to provide an intelligent control system and method for an energy-efficient integrated cold station in a complete set to solve the deficiencies in the background art.

[0005] To achieve the above purpose, the present invention provides the following technical solutions: An intelligent control method for an energy-efficient integrated cold station in a complete set, including: Arrange a plurality of sensors in the cold station system to monitor the operating status of each component of the system and external environmental changes in real time, including temperature, pressure, flow rate, and power consumption data, and transmit the real-time monitoring data to the central control system; Utilize historical operation data and external environmental data to establish a prediction model of load fluctuations through a long short-term memory network, predict the load demand of the cold station system at different time periods, and according to the load prediction results, the central control system pre-adjusts the operating parameters of each component of the cold station in advance; During real-time operation, based on the deviation between the current load and the predicted load and real-time monitoring data, the operating states of each component of the chilled water plant are dynamically optimized using fuzzy logic; Based on the optimized operating states of each component of the chilled water plant, anomaly detection is performed on the real-time monitoring data through a machine learning model. Once signs of potential faults in the chilled water plant system are detected, a fault warning is immediately issued and the load prediction model and optimization control strategy are automatically updated.

[0006] Preferably, historical operation data is collected and a dataset is established. The dataset is divided into a training set, a validation set, and a test set. According to the time series characteristics of the input data, the input and output structures of the LSTM model are designed. The input of the model is the load and environmental data of the past few time steps, and the output is the load prediction of a future time step. The LSTM model is trained, and the trained LSTM model is used to predict the load demand in the future time period, generating a predicted value of the future load. The predicted value is compared with the actual load demand, the prediction error is calculated, and the calculated prediction error is compared with the set error threshold. If the calculated prediction error is greater than or equal to the set error threshold, the operating parameters of each component of the chilled water plant are pre-adjusted; if the calculated prediction error is less than the set error threshold, no additional adjustment is required.

[0007] Preferably, the deviation between the current load and the predicted load and the real-time monitoring data are used as input items of fuzzy logic, and they are respectively divided into different fuzzy sets; The operating states of each component of the chilled water plant are used as output items of fuzzy logic, and they are divided into different fuzzy sets; Fuzzy rules are formulated to describe the influence of the deviation between the current load and the predicted load and the real-time monitoring data on the operating states of each component of the chilled water plant; According to the fuzzy inference results, the operating states of each component of the chilled water plant are optimized.

[0008] Preferably, based on the optimized operating states of each component of the chilled water plant, anomaly detection is performed on the real-time monitoring data through a machine learning model; the real-time monitoring data includes the abnormal power consumption index of the compressor and the temperature difference fluctuation index between the condenser and the evaporator. Among them, the method for obtaining the abnormal power consumption index of the compressor is: obtain the original power consumption data, calculate the covariance matrix of the data. Suppose there is an n×m data matrix X, where n is the number of samples and m is the number of features, and each column represents a feature. The calculation formula for the covariance matrix C is: ; where C is an m×m covariance matrix, is the transpose of the data matrix X; perform eigenvalue decomposition on the covariance matrix to obtain eigenvalues and eigenvectors. The eigenvalue decomposition formula for the covariance matrix is: ; where, is the eigenvector, is the corresponding eigenvalue. Assuming the first k principal components are selected, the new data representation is: ; where, is the matrix containing the first k eigenvectors, is the data representation in the principal component space. The reconstruction error refers to the error when the original data is projected into the principal component space and then reconstructed back to the original space from this space. First, the reconstructed power consumption data is expressed as: ; then, calculate the reconstruction error , and the expression is: ; calculate the power consumption anomaly index , and the expression is: ; where: is the original power consumption data of the i-th sample point, is the i-th sample point after reconstruction by the PCA model, is the standard deviation of the reconstruction error.

[0009] Preferably, among them, the method for obtaining the temperature difference fluctuation index of the condenser and the evaporator is: obtain the temperature difference data of the condenser and the evaporator. For time series data, first set a sliding window with a window size of w, which is used to calculate the mean and variance within the window, and detect the change of the mean by calculating the cumulative deviation. The cumulative sum calculation formula is: ; where, is the cumulative deviation at time t, is the temperature difference data at time t, is the expected mean of the data; if exceeds a certain threshold τ, it is considered that a change point occurs at time t, and the change point is selected by minimizing the total cost function: ; where, is the position of the change point, is the mean in each sub-interval, and λ is the penalty coefficient used to control the number of detected change points; after completing the change point detection, calculate the standard deviation of the temperature difference before and after the change of the change point as the temperature difference fluctuation index.

[0010] Preferably, convert the power consumption anomaly index and the temperature difference fluctuation index into a comprehensive feature vector, use the comprehensive feature vector as the input of the machine learning model. The machine learning model takes predicting the potential fault value label existing in the cold station system for each group of comprehensive feature vectors as the prediction target, and takes minimizing the sum of the prediction errors of the potential fault value labels existing in all cold station systems as the training target, and trains the machine learning model until the sum of the prediction errors reaches convergence and then stops the model training, and determines the potential fault value existing in the cold station system according to the model output result, where the machine learning model is a polynomial regression model.

[0011] Preferably, compare the obtained potential fault value of the chilled water station system with the reference threshold of the potential fault value of the chilled water station system set in advance. If the potential fault value of the chilled water station system is greater than or equal to the reference threshold of the potential fault value of the chilled water station system set in advance, it indicates that the risk of potential faults in the chilled water station system is high. At this time, a warning signal is generated, and the load prediction model and the optimal control strategy are automatically updated; if the potential fault value of the chilled water station system is less than the reference threshold of the potential fault value of the chilled water station system set in advance, it indicates that the risk of potential faults in the chilled water station system is low. At this time, no warning signal is generated.

[0012] The present invention also provides an intelligent management and control system for an energy-efficient integrated prefabricated chilled water station, including a data acquisition module, a load prediction module, an optimization module, and a fault warning module; Data acquisition module: Arrange multiple sensors in the chilled water station system to monitor the operating status of each component of the system and changes in the external environment in real time, including temperature, pressure, flow rate, and power consumption data, and transmit the real-time monitoring data to the central control system; Load prediction module: Use historical operation data and external environment data to establish a prediction model of load fluctuations through a long short-term memory network, predict the load demand of the chilled water station system at different time periods, and according to the load prediction results, the central control system pre-adjusts the operating parameters of each component of the chilled water station in advance; Optimization module: During the real-time operation process, based on the deviation between the current load and the predicted load and the real-time monitoring data, dynamically optimize the operating status of each component of the chilled water station after using fuzzy logic; Fault warning module: Based on the optimized operating status of each component of the chilled water station, perform anomaly detection on the real-time monitoring data through a machine learning model. Once signs of potential faults in the chilled water station system are detected, immediately issue a fault warning and automatically update the load prediction model and the optimal control strategy.

[0013] In the above technical solution, the technical effects and advantages provided by the present invention: 1. Through the innovative intelligent management and control system and method for an energy-efficient integrated prefabricated chilled water station of the present invention, the problems of lagging equipment response or over-regulation in the prior art under sudden load fluctuations can be effectively solved. By real-time monitoring the operating status of each component of the system and changes in the external environment, combining historical data with the load prediction model, the central control system can adjust the equipment operating parameters in advance and optimize the working status of each component of the chilled water station. At the same time, the operation of the chilled water station system is dynamically optimized through fuzzy logic, improving the response speed and energy-saving efficiency. On this basis, the machine learning model is further used for anomaly detection and potential fault warning to ensure that the chilled water station system can still maintain efficient and stable operation under load fluctuations and fault risks.

[0014] 2. By integrating the abnormal power consumption index and the temperature difference and pressure fluctuation index, the present invention generates a comprehensive feature vector and uses a polynomial regression model to predict the fault value, realizing the intelligent management of the system. By comparing the predicted fault value with the preset threshold, the system can issue a warning signal in time to prevent the expansion of the fault and improve the reliability and safety of the system. More importantly, the system can automatically update the load prediction model and optimize the control strategy after the fault warning to ensure that the cold station system always maintains the best operating state, thus significantly improving the energy-saving effect and reducing the operating cost. BRIEF DESCRIPTION OF THE DRAWINGS

[0015] In order to more clearly illustrate the technical solutions in the embodiments of the present application or the prior art, the following will briefly introduce the drawings required to be used in the embodiments. Obviously, the drawings described below are only some embodiments recorded in the present invention, and those of ordinary skill in the art can also obtain other drawings based on these drawings.

[0016] Figure 1 It is a flowchart of the method of the present invention.

[0017] Figure 2 It is a system module diagram of the present invention. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0018] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the following will clearly and completely describe the technical solutions in the embodiments of the present invention with reference to the drawings in the embodiments of the present invention. Obviously, the described embodiments are some, but not all, of the embodiments of the present invention. All other embodiments obtained by those of ordinary skill in the art without creative efforts based on the embodiments of the present invention belong to the scope of protection of the present invention.

[0019] Example 1, please refer to Figure 1 As shown, the intelligent control method for an efficient energy-saving integrated cold station in this embodiment includes: Arrange a plurality of sensors in the cold station system to real-time monitor the operating status of each component of the system and the changes in the external environment, including temperature, pressure, flow rate, and power consumption data, and transmit the real-time monitoring data to the central control system; Using historical operation data and external environment data, establish a prediction model of load fluctuation through a long short-term memory network to predict the load demand of the cold station system at different time periods. According to the load prediction results, the central control system pre-adjusts the operating parameters of each component of the cold station in advance; During the real-time operation process, based on the deviation between the current load and the predicted load and the real-time monitoring data, dynamically optimize the operating status of each component of the cold station after using fuzzy logic; Based on the optimized operating states of the components in the cold station, anomaly detection is performed on the real-time monitoring data through a machine learning model. Once signs of potential faults in the cold station system are detected, a fault warning is immediately issued and the load prediction model and optimization control strategy are automatically updated.

[0020] The specific steps for arranging multiple sensors in the cold station system to monitor the operating states of system components and external environment changes in real time are as follows: According to the requirements of the cold station system, select appropriate types of sensors. Common sensors include temperature sensors, pressure sensors, flow sensors, and power consumption sensors. For each type of sensor, a model and specification suitable for the cold station operating environment should be selected.

[0021] According to the component layout of the cold station system, determine the installation locations of the sensors. Generally, temperature sensors should be arranged at the inlets and outlets of the evaporator, condenser, and compressor, pressure sensors on the condenser, evaporator, compressor suction, and discharge pipelines, flow sensors on the key pipelines of the refrigeration cycle, and power consumption sensors at the power input ends of equipment such as compressors, pumps, and fans.

[0022] Install various types of sensors at the determined locations and ensure that the sensors are stably connected to the cold station equipment and can accurately collect data. The sensors need to be electrically and data-connected to the cold station's monitoring system to ensure the stability and real-time nature of the transmitted data.

[0023] After starting the cold station system, each sensor begins to collect real-time data on the equipment operating states, including important parameters such as temperature, pressure, flow, and power consumption. Each sensor regularly records data according to a preset sampling period and generates real-time monitoring data.

[0024] Before uploading the collected data, perform data preprocessing. This includes operations such as removing noise, correcting errors, and filtering out invalid data to ensure the accuracy and reliability of the transmitted data.

[0025] To ensure the stability and real-time nature of data transmission, select a suitable communication protocol (such as Modbus, BACnet, Zigbee, or Wi-Fi, etc.) to transmit the real-time data collected by the sensors to the central control system.

[0026] At a predetermined time interval, transmit the processed data to the central control system through wired or wireless communication methods to ensure that the data is transmitted to the central data center in real time and continuously. The data transmission can be relayed through a sensor gateway or edge computing device to achieve unified data transmission for multiple sensors.

[0027] After the central control system receives the transmitted data, it first conducts data verification and integrity checks to ensure the accuracy and reliability of the data. If packet loss or data errors occur during the transmission process, the system will automatically request a retransmission.

[0028] Store the received real-time monitoring data in the database and manage it according to the timestamps. Classify and store the data according to different parameters (temperature, pressure, flow rate, power consumption, etc.) for convenient subsequent analysis and query.

[0029] The central control system displays the received data to the operator through a graphical interface for intuitive monitoring of the real-time status of the cold station system. For example, the system can display current parameters such as temperature, pressure, and flow rate through real-time curve graphs and dashboards, and mark the normal range and abnormal status.

[0030] Specific steps for load forecasting: Collect historical operation data (e.g., historical load of the cold station, external environmental parameters such as temperature, humidity, wind speed, etc.) and relevant timestamp information. Normalize the collected time series data to improve training efficiency and avoid the influence of dimensional differences of certain parameters on the model performance. Divide the dataset into training set, validation set, and test set.

[0031] According to the time series characteristics of the input data, design the input and output structures of the LSTM model. For example, the input of the model can be the load and environmental data of the past few time steps, and the output is the load prediction of a future time step. Set hyperparameters such as the number of layers of the LSTM network, the number of neurons in each layer, and the activation function. Use a gradient descent algorithm (such as the Adam optimizer) to train the LSTM model to minimize the loss function (such as the mean squared error MSE).

[0032] Use the training set data to train the model, adjust the weights and biases of the network to minimize the error between the predicted load and the actual load. Use the validation set for mid-term evaluation to prevent overfitting, and improve the generalization ability of the model by adjusting hyperparameters (such as the learning rate, batch size, etc.).

[0033] Use the trained LSTM model to predict the load demand of the test set or future time periods. The model generates predicted values for future loads based on the current input data, including historical load, external temperature, and humidity data. The predicted load values can be single-step predictions (e.g., predicting the load demand for the next 1 hour) or multi-step predictions (e.g., predicting the load demand for the next day).

[0034] Compare the predicted value with the actual load demand, calculate the prediction error, and compare the calculated prediction error with the set error threshold. If the calculated prediction error is greater than or equal to the set error threshold, pre-adjust the operating parameters of each component of the cold station; if the calculated prediction error is less than the set error threshold, no additional adjustment is required.

[0035] If the predicted load demand increases, the central control system can pre-adjust the operating parameters of equipment (such as compressors, pumps, fans, etc.) in advance to increase the cooling capacity of the cold station. For example, the system can increase the speed of the compressor through variable frequency control, adjust the valve opening, or increase the refrigerant flow rate. When the predicted load is low: If the predicted load demand decreases, the central control system can reduce energy consumption and avoid energy waste by reducing the compressor speed and decreasing the cooling water flow rate.

[0036] The central control system pre-adjusts the equipment parameters according to the load prediction result and real-time monitors the difference between the actual load and the predicted load. If a deviation is found, the system can further adjust the equipment operation to ensure that the cold station operates in an optimal state. By pre-adjusting the operating parameters of each equipment in advance, it is ensured that the cold station can operate smoothly during the load fluctuation process, reduce energy waste, and improve the energy-saving effect.

[0037] During the real-time operation process, based on the deviation between the current load and the predicted load and the real-time monitoring data, the operating states of each component of the cold station are dynamically optimized using fuzzy logic. Specifically: First, the system needs to collect the working parameters of each component of the cold station in real time through sensors, such as temperature, pressure, flow rate, and power consumption, and obtain the values of the current load and the predicted load. These data will be used as the input of the fuzzy logic control system. Current load: The actual load of the cold station at the current moment (for example, refrigeration load, cooling load, etc.). Predicted load: The future load demand calculated by the LSTM model or other prediction methods. Real-time monitoring data: Includes the operating state data of the cold station equipment, such as the speed of the compressor, the frequency of the pump, the opening of the valve, and the power consumption. According to these data, the load deviation (i.e., the difference between the current load and the predicted load) and other real-time state parameters can be calculated. These data will become the input of the fuzzy logic control system.

[0038] The first step of the fuzzy logic system is to convert the accurate numerical data into fuzzy language variables. For example, the load deviation can be represented by the following fuzzy sets: Load deviation: It can be defined as fuzzy sets such as "low load deviation (negative)", "moderate load deviation (medium)", "high load deviation (positive)", etc.

[0039] Real-time monitoring data: Fuzzify parameters such as temperature, pressure, and flow rate. Common fuzzy sets include "low temperature", "medium temperature", "high temperature", etc.

[0040] To convert this data into fuzzy variables, it is first necessary to define the fuzzy sets and their membership functions. For example, the membership function of the load deviation can be represented as a triangular function, trapezoidal function, or Gaussian function, and the membership degree of the actual load deviation value in each fuzzy set is determined accordingly.

[0041] Fuzzy inference is the core step in fuzzy logic control. It uses the fuzzy rule base to perform inference and make decisions based on the input data. By setting a series of fuzzy rules, the fuzzy inference system calculates the output based on the input membership degrees.

[0042] In this step, the fuzzy rule base needs to be designed in advance according to the operating characteristics of the cold station, the load fluctuation law, and the equipment response ability. Common fuzzy rules can include: Rule example: If the load deviation is "high load deviation" and the current temperature is "high temperature", then adjust the speed of the compressor to "high speed". If the load deviation is "low load deviation" and the current pressure is "low pressure", then adjust the flow rate of the condenser to "low flow rate". If the load deviation is "moderate load deviation" and the current power consumption is "medium power consumption", then maintain the current state of the equipment. Each rule consists of multiple input conditions and one output result, and the system determines the output item through fuzzy operations. The fuzzy rule base combines the current input data and membership degrees for inference and calculation.

[0043] After the fuzzy inference obtains the result, the output is still a fuzzy value. To convert these fuzzy results into specific control instructions, defuzzification is required.

[0044] Common methods of defuzzification are the Center of Gravity (COG) method or the MaxMembership Method. Center of Gravity (COG) method: By calculating the center of gravity position of the fuzzy result area, the fuzzy value is converted into an actual control quantity. For example, for the control of the compressor speed, the COG method can calculate a weighted average value and use this value as the actual speed command of the compressor. MaxMembership Method: Select the specific control quantity corresponding to the output value with the largest membership degree. It is suitable for control methods that require more direct and rapid responses.

[0045] The result after defuzzification is the optimized control parameters for each component. For example, the control instructions may include: Compressor regulation: Determine the speed or power adjustment range of the compressor based on the load deviation and the current operating status. If the load deviation is positive and large, the system may increase the speed of the compressor; if the load deviation is negative, the system may decrease the speed.

[0046] Pump frequency regulation: Adjust the operating frequency of the pump according to the load change. If the load demand increases, the frequency of the pump may increase, and vice versa.

[0047] Condenser valve regulation: Adjust the flow rate of the condenser based on the changes in load and environment to optimize the heat exchange efficiency.

[0048] Once the cold station equipment adjusts its operating status according to the fuzzy control system, the new system data will be fed back to the central control system in real time. Through continuous feedback, the system can continuously optimize the operation of the equipment based on the difference between the latest operating status and the load prediction.

[0049] For example, the system can perform secondary adjustment on the load with a large deviation to avoid the risk of over-regulation. According to the real-time performance changes of the equipment, further optimize the operating mode of the cold station, reduce energy waste and improve system stability.

[0050] Fuzzy logic control is a dynamic process. The system will perform fuzzy reasoning and optimization adjustment in a cycle according to the changes of multiple factors such as real-time load, predicted load, temperature, pressure, and flow rate. This closed-loop control mode ensures that the cold station can always maintain an efficient and stable operating state under load fluctuations.

[0051] Based on the optimized operating status of each component of the cold station, perform anomaly detection on the real-time monitoring data through a machine learning model. The real-time monitoring data includes the abnormal power consumption index of the compressor and the temperature difference fluctuation index of the condenser and evaporator.

[0052] Among them, the method for obtaining the abnormal power consumption index of the compressor is as follows: Obtain the original power consumption data, calculate the covariance matrix of the data. Suppose there is an n×m data matrix X, where n is the number of samples and m is the number of features, and each column represents a feature; the calculation formula for the covariance matrix C is: ; where C is an m×m covariance matrix, is the transpose of the data matrix X; perform eigenvalue decomposition on the covariance matrix to obtain eigenvalues and eigenvectors. The eigenvectors represent the principal components of the data, and the eigenvalues represent the variances (i.e., importance) of the principal components. The eigenvalue decomposition formula of the covariance matrix is: ; where, are eigenvectors, is the corresponding eigenvalue. Select the eigenvectors corresponding to the largest several eigenvalues to construct a new low-dimensional space. Suppose the first k principal components are selected, then the new data representation is: ; where, is the matrix containing the first k eigenvectors, is the data representation in the principal component space. The reconstruction error refers to the error when the original data is projected into the principal component space and then reconstructed back to the original space from this space. This reconstruction error can be used as the basis for the anomaly index. First, the reconstructed power consumption data can be expressed as: ; then, calculate the reconstruction error , and the expression is: ; The reconstruction error measures the gap between the original data and the reconstruction through the principal component space. The larger the reconstruction error, the less the data point matches the representative pattern of the principal component space, and it may be abnormal data. We use the reconstruction error as the anomaly index to evaluate the anomaly degree of each data point. Calculate the power consumption anomaly index , and the expression is: ; where: is the original power consumption data of the i-th sample point, is the i-th sample point after reconstruction through the PCA model, is the standard deviation of the reconstruction error, which can be obtained by calculating all samples in the training set.

[0053] By calculating the anomaly index of all data points, the abnormal points can be distinguished from the normal points. Usually, a threshold T can be set to judge whether it is an abnormal point: if >T, then the sample is considered an abnormal point. If ≤T, then the sample is considered a normal point. The threshold T can be selected through statistical analysis (such as the reconstruction error distribution based on the training set) or cross-validation.

[0054] Among them, the method for obtaining the temperature difference fluctuation index of the condenser and the evaporator is as follows: Obtain the temperature difference data of the condenser and the evaporator. For time series data, first set a sliding window with a window size of w, which is used to calculate the mean and variance within the window, and detect the change of the mean by calculating the cumulative deviation. The cumulative sum calculation formula is: ; where, is the cumulative deviation at time t, is the temperature difference data at time t, is the expected mean of the data (i.e., the mean temperature difference during normal operation).

[0055] If If it exceeds a certain threshold τ, it is considered that a change point occurs at time t, and the change point is selected by minimizing the total cost function: ; where is the position of the change point, is the mean value in each sub - interval, and λ is the penalty coefficient used to control the number of detected change points. After completing the change point detection, the standard deviation of the temperature difference pressure difference before and after the change of the change point is calculated as the temperature difference pressure difference fluctuation index.

[0056] Convert the power consumption anomaly index and the temperature difference pressure difference fluctuation index into a comprehensive feature vector, and use the comprehensive feature vector as the input of the machine learning model. The machine learning model takes predicting the potential fault value label existing in the cold station system for each group of comprehensive feature vectors as the prediction target, and takes minimizing the sum of the prediction errors of the potential fault value labels existing in all cold station systems as the training target to train the machine learning model until the sum of the prediction errors reaches convergence and then stops the model training. Determine the potential fault value existing in the cold station system according to the model output result, where the machine learning model is a polynomial regression model.

[0057] The method for obtaining the potential fault value existing in the cold station system is: obtain the corresponding function expression from the comprehensive feature vector training data of the trained machine learning model: ; in the formula is the output function of the model, SX is the power consumption anomaly index, GH is the temperature difference pressure difference fluctuation index, q is the potential fault value existing in the cold station system.

[0058] Compare the obtained potential fault value existing in the cold station system with the reference threshold of the potential fault value existing in the cold station system set in advance. If the potential fault value existing in the cold station system is greater than or equal to the reference threshold of the potential fault value existing in the cold station system set in advance, it indicates that the risk of potential faults in the cold station system is high. At this time, generate a warning signal and automatically update the load prediction model and the optimal control strategy; if the potential fault value existing in the cold station system is less than the reference threshold of the potential fault value existing in the cold station system set in advance, it indicates that the risk of potential faults in the cold station system is low. At this time, no warning signal is generated.

[0059] If the system fault risk is high, generate a warning signal to remind the operation and maintenance personnel to check the cold station system. The warning signal can be pushed through the real - time monitoring system or the automatic alarm system. When the potential fault risk is relatively high, the system can automatically update the load prediction model and the optimal control strategy, adjust the operation parameters of the cold station system to reduce the probability of faults and optimize energy consumption.

[0060] Example 2, please refer to Figure 2As shown in the figure, an intelligent control system for an efficient and energy-saving integrated cold station in this embodiment includes a data acquisition module, a load prediction module, an optimization module, and a fault warning module; Data acquisition module: A plurality of sensors are arranged in the cold station system to monitor the operating status of each component of the system and changes in the external environment in real time, including temperature, pressure, flow rate, and power consumption data, and transmit the real-time monitoring data to the central control system; Load prediction module: Using historical operation data and external environment data, a prediction model of load fluctuations is established through a long short-term memory network to predict the load demand of the cold station system at different time periods. According to the load prediction results, the central control system pre-adjusts the operating parameters of each component of the cold station in advance; Optimization module: During the real-time operation process, based on the deviation between the current load and the predicted load and the real-time monitoring data, the operating status of each component of the cold station is dynamically optimized after using fuzzy logic; Fault warning module: Based on the optimized operating status of each component of the cold station, abnormal detection is performed on the real-time monitoring data through a machine learning model. Once a potential fault sign in the cold station system is detected, a fault warning is immediately issued and the load prediction model and the optimized control strategy are automatically updated.

[0061] The above formulas are all dimensionless and take their numerical values for calculation. The formulas are obtained by collecting a large amount of data for software simulation to obtain a formula that is closest to the actual situation. The preset parameters in the formulas are set by those skilled in the art according to the actual situation.

[0062] It should be understood that the term "and / or" in this article is only a description of the association relationship of associated objects, indicating that there can be three relationships. For example, A and / or B can represent: A exists alone, A and B exist simultaneously, and B exists alone. These three situations, where A and B can be singular or plural. In addition, the character " / " in this article generally represents an "or" relationship between the associated objects before and after, but it may also represent an "and / or" relationship. For specific understanding, reference can be made to the context before and after.

[0063] Those of ordinary skill in the art can realize that the units and algorithm steps of each example described in combination with the embodiments disclosed in this article can be implemented by electronic hardware, or by a combination of computer software and electronic hardware. Whether these functions are executed in a hardware or software manner depends on the specific application and design constraints of the technical solution. Professional technicians can use different methods to implement the described functions for each specific application, but such implementation should not be considered to exceed the scope of this application.

[0064] As described above, it is only the specific implementation manner of this application. However, the protection scope of this application is not limited thereto. Any person skilled in the art within the technical scope disclosed by this application can easily think of changes or substitutions, which should all be covered within the protection scope of this application.

Claims

1. An intelligent management and control method for an energy-efficient integrated cold storage station, characterized in that: include: Multiple sensors are deployed in the cold station system to monitor the operating status of each system component and external environment changes in real time, including temperature, pressure, flow and power consumption data, and transmit the real-time monitoring data to the central control system; Using historical operation data and external environment data, a load fluctuation prediction model is established through a long short-term memory network to predict the load demand of the cooling station system in different time periods. Based on the load prediction results, the central control system pre-adjusts the operating parameters of each component of the cooling station in advance; In real-time operation, based on the deviation between current load and forecasted load and real-time monitoring data, fuzzy logic is used to dynamically optimize the operating status of each component of the cold station; Based on the operating status of each component of the optimized cold station, the real-time monitoring data is detected for anomalies through a machine learning model. Once signs of potential failure in the cold station system are detected, a fault warning is immediately issued and the load forecasting model and optimization control strategy are automatically updated.

2. According to claim 1, a highly efficient and energy-saving integrated cold storage intelligent management and control method is characterized by: Collect historical operation data and establish a data set, and divide the data set into a training set, a validation set, and a test set; design the input and output structure of the LSTM model based on the time series characteristics of the input data. The input of the model is the load and environmental data of the past few time steps, and the output is the load forecast of a certain time step in the future. Train the LSTM model, use the trained LSTM model to predict the load demand in the future time period, generate the predicted value of the future load, compare the predicted value with the actual load demand, calculate the prediction error, and compare the calculated prediction error with the set error threshold. If the calculated prediction error is greater than or equal to the set error threshold, pre-adjust the operating parameters of each component of the cold station; If the calculated prediction error is less than the set error threshold, no additional adjustment is required.

3. According to claim 1, a highly efficient and energy-saving integrated cold storage intelligent management and control method is characterized by: The deviation between the current load and the predicted load and the real-time monitoring data are used as the input items of fuzzy logic and divided into different fuzzy sets respectively; The operating status of each component of the cold station is taken as the output item of fuzzy logic and divided into different fuzzy sets; Formulate fuzzy rules to describe the deviation between current load and forecast load and the impact of real-time monitoring data on the operating status of each component of the cooling station; According to the fuzzy reasoning results, the operating status of each component of the cold station is optimized.

4. According to claim 1, a highly efficient and energy-saving integrated cold storage intelligent management and control method is characterized by: Based on the optimized operating status of each component of the cold station, the real-time monitoring data is detected for anomalies through a machine learning model; the real-time monitoring data includes the power consumption anomaly index of the compressor and the temperature and pressure difference fluctuation index of the condenser and evaporator; the method for obtaining the power consumption anomaly index of the compressor is: obtaining the original power consumption data, calculating the covariance matrix of the data, and setting an n×m data matrix X, where n is the number of samples, m is the number of features, and each column represents a feature; the calculation formula of the covariance matrix C is: ; where C is the m×m covariance matrix, is the transpose of the data matrix X; perform eigenvalue decomposition on the covariance matrix to obtain eigenvalues ​​and eigenvectors. The eigenvalue decomposition formula of the covariance matrix is: ;in, is the eigenvector, is the corresponding eigenvalue, and if the first k principal components are selected, the new data is expressed as: ;in, is the matrix containing the first k eigenvectors, is the data representation in the principal component space. The reconstruction error refers to the error when the original data is projected into the principal component space and then reconstructed back to the original space from this space. First, the reconstructed power consumption data It is expressed as: ; Then, calculate the reconstruction error , the expression is: ; Calculate the power consumption abnormality index , the expression is: ;in: is the original power consumption data of the i-th sample point, is the i-th sample point reconstructed by the PCA model, is the standard deviation of the reconstruction error.

5. According to claim 4, a highly efficient and energy-saving integrated cold storage intelligent management and control method is characterized by: in, The method for obtaining the temperature and pressure difference fluctuation index of the condenser and evaporator is as follows: obtain the temperature and pressure difference data of the condenser and evaporator. For time series data, first set a sliding window with a window size of w to calculate the mean and variance within the window. The change of the mean is detected by calculating the cumulative deviation. The cumulative sum calculation formula is: ;in, is the cumulative deviation at time t, is the temperature and pressure difference data at time t, is the expected mean of the data; if If a certain threshold τ is exceeded, it is considered that a change point has occurred at time t, and the change point is selected by minimizing the total cost function: ;in, is the location of the change point, is the mean value in each subinterval, and λ is the penalty coefficient, which is used to control the number of change points detected. After completing the change point detection, the standard deviation of the temperature and pressure difference before and after the change point is calculated as the temperature and pressure difference fluctuation index.

6. According to claim 5, a highly efficient and energy-saving integrated cold storage intelligent management and control method is characterized by: The power consumption anomaly index and the temperature and pressure difference fluctuation index are converted into comprehensive feature vectors, and the comprehensive feature vectors are used as the input of the machine learning model. The machine learning model predicts the potential fault value labels of the cold station system with each group of comprehensive feature vectors as the prediction target, and minimizes the sum of prediction errors of the potential fault value labels of all cold station systems as the training target. The machine learning model is trained until the sum of prediction errors converges, and the model training is stopped. The potential fault value of the cold station system is determined according to the model output results, wherein the machine learning model is a polynomial regression model.

7. According to claim 6, a highly efficient and energy-saving integrated cold storage intelligent management and control method is characterized by: The obtained potential fault value of the cooling station system is compared with the preset reference threshold of the potential fault value of the cooling station system. If the potential fault value of the cooling station system is greater than or equal to the preset reference threshold of the potential fault value of the cooling station system, it indicates that the risk of potential fault of the cooling station system is high. At this time, an early warning signal is generated, and the load forecasting model and the optimization control strategy are automatically updated. If the potential failure value of the cooling station system is less than a preset reference threshold of the potential failure value of the cooling station system, it indicates that the risk of potential failure of the cooling station system is low, and no warning signal is generated at this time.

8. An efficient and energy-saving integrated cold station intelligent management and control system, used to implement an efficient and energy-saving integrated cold station intelligent management and control method according to any one of claims 1 to 7, characterized in that: Including data acquisition module, load forecasting module, optimization module and fault warning module; Data acquisition module: multiple sensors are arranged in the cold station system to monitor the operating status of each component of the system and changes in the external environment in real time, including temperature, pressure, flow and power consumption data, and transmit the real-time monitoring data to the central control system; Load forecasting module: Using historical operation data and external environment data, a load fluctuation forecasting model is established through a long short-term memory network to forecast the load demand of the cooling station system in different time periods. Based on the load forecasting results, the central control system pre-adjusts the operating parameters of each component of the cooling station in advance; Optimization module: In real-time operation, based on the deviation between current load and forecasted load and real-time monitoring data, fuzzy logic is used to dynamically optimize the operating status of each component of the cold station; Fault warning module: Based on the optimized operating status of each component of the cold station, the real-time monitoring data is detected for anomalies through a machine learning model. Once signs of potential failure in the cold station system are detected, a fault warning is immediately issued and the load forecasting model and optimization control strategy are automatically updated.

Images