Low-load working condition refrigerator and water pump linkage energy-saving control method

Abstract

The invention provides a cold machine and water pump linkage energy-saving control method under a low-load working condition, which relates to the technical field of information and comprises the following steps of: dynamically predicting load requirements under the low-load working condition by analyzing historical operation data of a refrigeration system and combining weather forecast information, equipment operation parameters and data of building personnel activity rules; obtaining a cooling capacity demand curve and a load demand prediction result in future preset time; establishing a system low-load operation mode optimization model, taking the screened control combination meeting the preset energy-saving operation target as model input, and obtaining an optimized system low-load operation mode through the training model; the optimal energy efficiency interval is determined by analyzing the relation between equipment energy consumption and cold storage capacity, and during load demand prediction and control optimization, the interval serves as a target constraint to guide a refrigerator and water pump equipment to operate in a high-energy-efficiency area. And the operation efficiency and the energy-saving effect of the refrigerating system under the low-load working condition are improved.

Description

Technical Field

[0001] The present invention relates to the field of information technology, and in particular to a method for controlling energy saving by linkage between a chiller and a water pump under low load conditions. Background Art

[0002] Under low-load conditions, the linkage control of chillers and water pumps in refrigeration systems faces many technical challenges. First, it is necessary to obtain and predict the dynamic change law of load demand under low-load conditions, which is affected by many factors, including weather and personnel activities, and the law is difficult to grasp. Secondly, there are many control strategies for the start-stop sequence, frequency regulation, and valve opening of chillers and water pumps, and they affect each other. The optimization of the system operation mode brought about by the control combination is a key issue. Furthermore, in the actual control execution process under the optimized system operation mode, the key direction to solve the system operation under low-load conditions is to set the chilled water temperature of the chiller and adjust the cold storage capacity in combination with the change in load demand. In addition, real-time adjustment of the control strategy to achieve the goal of energy-saving operation is a complex optimization problem involving operating condition prediction, control decision-making, and equipment coordination. Finally, while meeting the energy-saving goals, the chillers and water pump equipment of the refrigeration system must also operate efficiently, which further increases the difficulty of control. Summary of the invention

[0003] In order to solve the above technical problems, the present invention provides a low-load condition chiller and water pump linkage energy-saving control method, which mainly includes: By analyzing the historical operation data of the refrigeration system, combined with weather forecast information, equipment operating parameters, and data on the activities of people in the building, the load demand under low-load conditions is dynamically predicted to obtain the cooling demand curve and load demand forecast results within the preset time in the future; based on the cooling demand curve predicted under low-load conditions, a combined control strategy model of the chiller and water pump is constructed, and the system energy consumption is used as the model input to obtain the combined control strategy of the chiller and water pump under low-load conditions, and determine the optimal start and stop sequence and frequency adjustment plan of the chiller and water pump under different cooling demands; According to the combined control strategy of the chiller and water pump under low-load conditions, in the start-stop optimization sequence of the chiller and water pump, the temperature, pressure and flow parameters in the chilled water pipeline and the cooling water pipeline under low-load conditions are collected in real time to determine the degree of deviation between the actual low-load operating conditions of the system and the preset expected low-load operating conditions. If the deviation exceeds the deviation threshold, the valve opening is dynamically adjusted; The start and stop optimization sequence, frequency regulation scheme and valve opening adjustment of the chiller and water pump are integrated to generate different control combinations. By establishing a digital twin model of the refrigeration system, the low-load system operation status under different control combinations is simulated in a virtual environment to obtain key performance indicators such as energy consumption, cooling capacity and temperature fluctuation, and quantitative analysis and evaluation are performed to screen out control combinations that meet the preset energy-saving operation goals; a low-load operation mode optimization model is established, and the screened control combinations that meet the preset energy-saving operation goals are used as the input of the low-load operation mode optimization model. By training the low-load operation mode optimization model, the optimized system low-load operation mode is obtained; In the actual control execution process under the optimized system low-load operation mode, the operating parameters of the key equipment of the chiller, water pump, and pipeline are collected in real time, combined with the cooling demand curve and load demand forecast results, to obtain the set value of the chilled water temperature under low-load conditions, and adjust the cold storage capacity of the chiller according to the chilled water temperature set value; By analyzing the relationship between the energy consumption and cold storage capacity of chillers and water pumps, the optimal energy efficiency range is determined. During load demand forecasting and control optimization, this range is used as the target constraint to guide chillers and water pump equipment to operate in a high energy efficiency area.

[0004] The technical solution provided by the embodiment of the present invention may have the following beneficial effects: The present invention discloses an energy-saving control method for the linkage between a chiller and a water pump under low-load conditions. The cooling demand curve is dynamically predicted by analyzing historical operating data, weather forecasts, and the patterns of personnel activities. According to the prediction results, a combined control strategy model of the chiller and the water pump is constructed to optimize the start and stop sequence and frequency regulation of the equipment. In actual operation, the present invention monitors system parameters in real time, dynamically adjusts the valve opening, and uses digital twin technology to evaluate the performance of different control combinations. By establishing a low-load operation mode optimization model, the present invention realizes the intelligent regulation of the chilled water temperature and the enhancement of the cold storage capacity of the chiller. At the same time, the present invention analyzes the relationship between the energy consumption of the equipment and the cold storage capacity, determines the optimal energy efficiency range, and guides the equipment to operate in the high energy efficiency area. This method significantly improves the operating efficiency and energy-saving effect of the refrigeration system under low-load conditions. BRIEF DESCRIPTION OF THE DRAWINGS

[0005] Figure 1 The present invention is a flow chart of the energy-saving control method of the chiller and water pump linkage under low-load conditions. DETAILED DESCRIPTION

[0006] In order to make the purpose, technical solutions and advantages of the present invention more clear, the present invention is described in detail below with reference to the accompanying drawings and specific embodiments.

[0007] like Figure 1 The energy-saving control method of the low-load chiller and water pump linkage in this embodiment may specifically include: Step S101, by analyzing the historical operation data of the refrigeration system, combined with weather forecast information, equipment operation parameters, and data on the activities of people in the building, the load demand under low-load conditions is dynamically predicted to obtain the cooling demand curve and load demand forecast results within a preset time in the future.

[0008] Obtain the historical operation data of the system, and use the exponential smoothing method to process the historical operation data of the system to obtain the preliminary load demand forecast result under low-load conditions. According to the preliminary load demand forecast result, combined with weather forecast information, a multivariate linear regression model is established, and the multivariate linear regression model is used to determine the mapping relationship between weather factors and load demand. For the historical operation data of the system, the K-means algorithm is used to perform cluster analysis. If the Euclidean distance between the current working condition and a certain cluster center is the smallest, the sample most similar to the current working condition is extracted from the cluster. Calculate the reference value based on the sample, and perform weighted average of the reference value, the preliminary load demand forecast result and the output of the multivariate linear regression model to obtain the optimized load demand forecast result. Set a fixed-size time window, repeat the prediction process for each time point in the time window, and obtain a continuous cooling demand curve by sliding the time window.

[0009] Specifically, a time series model is constructed based on the historical operation data of the system, and the load demand under low-load conditions is preliminarily predicted using the exponential smoothing method to obtain preliminary prediction results. The exponential smoothing method assigns different weights to historical data and calculates the weighted average as the prediction value. The weight decreases exponentially over time. The N historical data points before the prediction time point are weighted, and the weight coefficient is α(1-α) i , where α is the smoothing coefficient and i is the time interval between the data point and the forecast point. Multiple linear regression is used to extract the features of weather forecast information, establish the mapping relationship between weather factors and load demand, and revise the load demand forecast based on the preliminary forecast results. Collect historical weather data and corresponding load data, select relevant features such as temperature, humidity, wind speed, etc. as independent variables, and load as the dependent variable, and establish the regression equation Y=β0+β1X1+β2X2+...+β n X n , where Y is the load, X i are weather characteristics, they represent different weather characteristics, such as temperature X1, humidity X2... wind speed X n , β iis the regression coefficient. The least squares method is used to calculate each coefficient, and the predicted weather data is substituted into the equation to obtain the corrected result. The K-means algorithm is used to cluster the equipment operating parameters and the data of the activity patterns of building personnel, and similar working conditions are grouped. The Euclidean distance between the current working condition and the center of each cluster is calculated, and the cluster with the closest distance is selected. The 10 samples most similar to the current working condition are extracted from the cluster, and the average load of these samples is calculated as the reference value. The reference value is weighted averaged with the prediction results of the first two steps, and the weight is dynamically adjusted according to the historical prediction accuracy to obtain the optimized load demand prediction result. By sliding the time window, the above prediction process is repeated for each time point in the future preset time, and finally the dynamically updated cooling demand curve and load demand prediction results are obtained. A fixed-size time window of 24 hours is set, and the first three steps of the prediction process are performed for each hour time point in the window. After each prediction is completed, the window is slid back by 1 hour until the entire preset time range is covered. The prediction results at each time point are recorded to form a continuous load demand curve. When predicting load demand, the exponential smoothing method is first used to process the system historical operation data. Assume that there is hourly load data for the past 30 days, and select α=0.2 as the smoothing coefficient. For the data 24 hours before the forecast time point, calculate the weighted average as the preliminary forecast result. For example, if the load at the current time point is 100kW, 95kW one hour ago...105kW 24 hours ago, the forecast value is 100*0.2+95*0.2*(1-0.2)+...+105*0.2(1-0.2) 23 , and the preliminary prediction value is 98.8kW. Next, collect the corresponding historical weather data, including temperature, humidity, wind speed and other factors. Select these factors as independent variables and load as dependent variable to establish a multivariate linear regression equation. Assume that the regression analysis results show that for every 1°C increase in temperature, the load increases by 2kW, and the regression coefficient of X1 is 2; for every 10% increase in humidity, the load increases by 1kW, and the regression coefficient of X2 is 1; for every 1m / s increase in wind speed, the load decreases by 0.5kW, and the regression coefficient of X2 is 1. nThe regression coefficient is -0.5. Substitute the weather data at the prediction time point, such as temperature 25°C, humidity 60%...wind speed 3m / s, into the equation, and the corrected predicted load is 98.8+2*25+0.1*60+...+(-0.5)*3=101.5kW. Then, the K-means algorithm is used to perform cluster analysis on the equipment operating parameters and the data on the regularity of building personnel activities. Assume that the historical data is divided into 5 clusters, calculate the Euclidean distance between the current working condition and the center of each cluster, and select the cluster with the closest distance. Extract the 10 most similar samples from the cluster, and calculate the average load of 99kW. Take this reference value and the prediction results of the first two steps for weighted average, assuming that the weights are 0.3, 0.4, and 0.3 respectively, and the final optimized prediction result is 100.3kW. Finally, use a 24-hour sliding time window to repeat the above prediction process for each hour in the future preset time. For example, starting from the current time, predict the load demand for the next 7 days. After each hour of forecasting is completed, the window slides back by 1 hour until all 168 hours are covered. The forecast results at each time point are recorded to form a continuous load demand curve, and a dynamically updated cooling demand curve and load demand forecast results are obtained.

[0010] Step S102, based on the cooling demand curve predicted under low load conditions, by constructing a combined control strategy model of the chiller and water pump, taking the system energy consumption as the model input, obtaining the combined control strategy of the chiller and water pump under low load conditions, and determining the optimal start and stop sequence and frequency adjustment scheme of the chiller and water pump under different cooling demands.

[0011] According to the cooling demand curve, a mixed integer linear programming method is used to construct a chiller and water pump combination control model; the mixed integer linear programming method takes the system energy consumption as the objective function and the operating status of the chiller and water pump as the decision variable. According to the chiller and water pump combination control model, a genetic algorithm is used to optimize the search process to calculate the optimal chiller and water pump combination under different cooling demands; wherein the genetic algorithm generates a new population by initializing the population, calculating the fitness, selecting, crossing, and mutating, and iteratively optimizes until the fitness converges. For the optimal chiller and water pump combination, an optimal start and stop sequence of the chiller and water pump is formulated; wherein the start and stop optimization sequence is defined by establishing fuzzy rules, the input variables are defined as the cooling demand change rate and the current load rate, and the output variable is the start and stop priority. According to the historical operation data, polynomial regression fitting is used to obtain the efficiency curves of the chiller and water pump under different loads; combined with the real-time cooling demand and the efficiency curve, the gradient descent method is used to calculate the optimal operating frequency; the gradient descent method updates the frequency value iteratively until the total energy consumption change is less than the total energy consumption change threshold or reaches the maximum number of iterations, and outputs the optimal operating frequency of each device to form a frequency adjustment plan.

[0012] Specifically, according to the cooling demand curve predicted under low-load conditions, a mixed integer linear programming method is used to construct a chiller and water pump combination control model, taking the system energy consumption as the objective function, the operating status of the chiller and water pump as the decision variable, and establishing constraints. The decision variables are defined as the switch status and operating frequency of the chiller and water pump, and the constraints are set to meet the cooling demand and the equipment operating range, and the objective function of minimizing the total energy consumption is constructed. The genetic algorithm is used to optimize the search process to calculate the optimal chiller and water pump combination under different cooling demands. Initialize the population, set the population size to 100, the chromosome length to the number of devices, and the gene value to represent the device status and frequency. Calculate the fitness of each chromosome, that is, the inverse of the total energy consumption of the system. Generate a new population through roulette selection, single-point crossover, and random mutation. Iterate the optimization for 500 generations or when the fitness converges to obtain the optimal chiller and water pump combination for each cooling demand point. According to the obtained optimal chiller and water pump combination, formulate the start and stop optimization sequence of the chiller and water pump. The input variables are defined as the rate of change of cooling demand and the current load rate, and the output variable is the start-stop priority. Fuzzy rules are designed, such as "If the rate of change of cooling demand is positive and the current load rate is high, the start priority is high" and "If the rate of change of cooling demand is negative and the current load rate is low, the shutdown priority is high". The start-stop priority of each device is calculated by fuzzy reasoning to generate a priority list. According to the historical operation data, the efficiency curves of the chiller and water pump under different loads are obtained by polynomial regression fitting. Combined with the real-time cooling demand and efficiency curve, the gradient descent method is used to calculate the optimal operating frequency. The learning rate is set to 0.01, the number of iterations is 1000, and the frequency value is updated each iteration until the total energy consumption change is less than the preset threshold or the maximum number of iterations is reached. The optimal operating frequency of each device is output to form a frequency adjustment plan to achieve dynamic optimization of system energy consumption. When implementing the combined control of the chiller and the water pump, the mixed integer linear programming method is first used to construct the combined control model of the chiller and the water pump based on the predicted cooling demand curve. Assume that the system contains 3 chillers and 4 water pumps, and the cooling demand range is 0-1000kW. The decision variables include the switch state of each device, 0 or 1, and the operating frequency, 20-50Hz. The constraints are set as the total cooling output must meet the demand, the maximum cooling capacity of a single chiller is 400kW, and the water pump flow is proportional to the frequency. The objective function is to minimize the total energy consumption, where the energy consumption of the chiller is related to the cooling capacity and COP, and the energy consumption of the water pump is related to the flow and head. The solver is used to calculate the preliminary optimal combination solution. Then, the genetic algorithm is used to further optimize the search process. The population size is set to 100, the chromosome length is 7 (corresponding to 7 devices), and the gene value represents the device state and frequency. The fitness function is defined as the inverse of the total energy consumption. Through 500 generations of iterations, 50 individuals with higher fitness are selected in each generation, and single-point crossover with a crossover probability of 0.8 and random mutation with a mutation probability of 0.05 are performed to generate a new population.Finally, the optimal equipment combination under different cooling demands is obtained. For example, when the demand is 500kW, two chillers are turned on with frequencies of 35Hz and 30Hz and three water pumps with frequencies of 40Hz, 35Hz and 30Hz. Then, the optimal start and stop sequence of the equipment is formulated. The fuzzy input variables are defined as the cooling demand change rate from -10% / h to +10% / h and the current load rate from 0 to 100%; the output variable is the start and stop priority from 0 to 10. Fuzzy rules are designed, such as "If the cooling demand change rate is +5% / h and the current load rate is 80%, the start priority is 8". The priority of each device is calculated through fuzzy reasoning to generate a priority list. Finally, the equipment efficiency curve is fitted using historical data. For example, the COP curve of the chiller is a quadratic function of COP=-0.001x. 2 +0.05x+3, x is the load rate. The pump efficiency curve is η=-0.0005y 2 +0.03y+0.6, y is the flow rate, η is the pump efficiency. Combined with the real-time cooling demand, the gradient descent method is used to calculate the optimal operating frequency. Set the learning rate to 0.01 and iterate 1000 times. Update the frequency value at each iteration until the total energy consumption changes by less than 0.1kW or the maximum number of iterations is reached. Finally, the optimal operating frequency of each device is output, such as 38Hz for chiller 1, 32Hz for chiller 2, 42Hz for pump 1, 37Hz for pump 2, and 33Hz for pump 3, forming a complete frequency regulation solution.

[0013] Step S103, according to the combined control strategy of the chiller and the water pump under low-load conditions, in the start-stop optimization sequence of the chiller and the water pump, by real-time collection of the temperature, pressure and flow parameters in the chilled water pipeline and the cooling water pipeline under low-load conditions, the degree of deviation between the actual low-load operating condition of the system and the preset expected low-load operating condition is determined. If the deviation exceeds the deviation threshold, the valve opening is dynamically adjusted.

[0014] The temperature, pressure and flow parameters in the chilled water pipeline and the cooling water pipeline are collected in real time through a distributed sensor network to obtain the original data under the low-load condition of the chiller. According to the original data, the moving average filtering method is used to remove noise and obtain the parameter data after smoothing. For the parameter data after smoothing, the minimum and maximum normalization method is used to scale it to a range of zero to one, and then the weighted average method is used to construct the characteristic vector of the actual low-load operating condition of the system, and the actual low-load operating condition index of the system is calculated according to the preset temperature, pressure and flow weights. The actual low-load operating condition index of the system is compared with the preset expected low-load operating condition index to determine whether the relative error exceeds the deviation threshold. If the relative error exceeds the deviation threshold, the adjustment is triggered. Based on the feedback control principle, combined with the system deviation degree and the current valve opening, the proportional integral differential control method is used to calculate the adjustment amount of the valve opening. According to the adjustment amount of the valve opening, the valve opening is dynamically adjusted by the actuator.

[0015] Specifically, according to the combined control strategy of chillers and water pumps under low-load conditions, the temperature, pressure, and flow parameters in the chilled water pipeline and cooling water pipeline under low-load conditions are collected in real time through a distributed sensor network. A 5-minute moving average filter is applied to the collected data to remove noise, and then the minimum and maximum normalization is used to scale the data to the range of 0-1 to obtain a standardized system operation parameter data set. According to the standardized system operation parameter data set, the weighted average method is used to construct the feature vector of the actual low-load operation condition of the system. The weights are pre-set according to the importance of the parameters, such as temperature 0.4, pressure 0.3, and flow 0.3, and the actual low-load operation condition index of the system is calculated. Using the relative error calculation method, the comprehensive index is compared with the preset expected low-load operation condition index to compare the actual low-load operation condition of the system with the preset expected low-load operation condition, and the degree of deviation between the two is calculated. The relative error between the actual comprehensive index and the expected comprehensive index is calculated, and the formula is |(actual value-expected value) / expected value|. The relative error threshold is set to 10%. If the deviation exceeds the relative error threshold, the adjustment mechanism is triggered. Based on the feedback control principle, combined with the system deviation and the current valve opening, the proportional integral differential control algorithm is used to calculate the adjustment amount of the valve opening. The Ziegler-Nichols tuning method is used to determine the PID parameters, and the proportional gain K p Increase until the system oscillates continuously, and record K at this time p K u , the oscillation period is T u . Calculate PID parameters according to the empirical formula, K p =0.6K u , T i =0.5T u , T d =0.125Tu . The valve opening is dynamically adjusted by the actuator to achieve closed-loop control of the system parameters. When implementing the combined control strategy of the chiller and the water pump, the data under low-load conditions is first collected once a minute through the distributed sensor network. For example, the chilled water supply temperature is 7°C, the return water temperature is 12°C, the cooling water inlet temperature is 28°C, the outlet water temperature is 33°C, the chilled water pressure is 300kPa, the cooling water pressure is 400kPa, the chilled water flow rate is 50m³ / h, and the cooling water flow rate is 60m³ / h. A 5-minute moving average filter is applied to these data, and the average of the last 5 data points is taken to eliminate short-term fluctuations. Then use minimum-maximum normalization to scale the temperature, pressure, and flow rate to the range of 0-1 respectively. If the temperature normalization formula is (TT min ) / (T max -T min ), where T min and T max They are the lowest and highest temperatures in history, respectively. Next, the weighted average method is used to calculate the comprehensive index of the system operating status. The weights are set to 0.4 for temperature, 0.3 for pressure, and 0.3 for flow. Assuming that the normalized temperature is 0.6, the pressure is 0.7, and the flow is 0.5, the comprehensive index = 0.6*0.4+0.7*0.3+0.5*0.3=0.6. Compare this comprehensive index with the preset expected low-load operating condition index, such as 0.55, and the calculated relative error is |(0.6-0.55) / 0.55|=9.1%. If this deviation exceeds the preset threshold of 10%, the adjustment mechanism is triggered. Finally, the proportional integral differential (PID) control algorithm is used to adjust the valve opening. The PID parameters are determined by the Ziegler-Nichols tuning method: gradually increase the proportional gain K p Until the system continues to oscillate, record K p is 100, the oscillation period T u is 30 seconds. K is calculated according to the empirical formula p =60, T i = 15 seconds, T d = 3.75 seconds. Based on these parameters and the current deviation, the PID controller calculates the adjustment amount of the valve opening, such as an increase of 5%. The actuator then adjusts the valve opening from the current 50% to 55%, achieving closed-loop control of the system parameters.

[0016] Step S104, comprehensively consider the start and stop optimization sequence, frequency regulation scheme and valve opening adjustment of the chiller and water pump, generate different control combinations, and establish a digital twin model of the refrigeration system to simulate the low-load system operation status under different control combinations in a virtual environment, obtain key performance indicators such as energy consumption, cooling capacity and temperature fluctuation, conduct quantitative analysis and evaluation, and screen out control combinations that meet the preset energy-saving operation goals.

[0017] According to the start and stop status, frequency level and valve opening parameters of the chiller and water pump, the enumeration method is used to generate a control combination scheme to construct the multidimensional control matrix; according to the multidimensional control matrix, the energy balance equation and the mass balance equation are used to establish a digital twin model of the refrigeration system; the parameters in the multidimensional control matrix are input into the digital twin model, and the system operation status under low load conditions is obtained by a numerical solution method; the key performance indicators of energy consumption, cooling capacity and temperature fluctuation are extracted from the system operation status, and the weighted summation method is used to calculate the comprehensive performance score, and the calculation formula of the comprehensive performance score is comprehensive score = energy consumption weight * energy consumption normalized value + cooling capacity weight * cooling capacity normalized value + temperature fluctuation weight * temperature fluctuation normalized value; if the comprehensive performance score meets the preset energy-saving operation target threshold, the corresponding control combination scheme is determined as a scheme that meets the conditions; the schemes that meet the conditions are arranged in descending order according to the comprehensive performance score, and the control combination with the highest comprehensive performance score is selected as the optimal scheme.

[0018] Specifically, different control combination schemes are generated by enumeration method according to the start and stop optimization sequence, frequency regulation and valve opening adjustment parameters of the chiller and water pump. All possible start and stop states, frequency levels and valve openings are listed, and all possible combinations are generated through nested loops to construct a multidimensional control matrix containing the start and stop states of the equipment, operating frequencies and valve openings. The digital twin model of the refrigeration system is constructed using the energy balance equation and the mass balance equation. The chiller cooling capacity calculation formula is used, Q=m*c*(T1-T2), where Q is the cooling capacity, m is the chilled water flow, c is the specific heat capacity, T1 is the return water temperature, and T2 is the supply water temperature. The parameters in the control combination scheme are input into the virtual environment, and the system operation status under low load conditions is calculated by the numerical solution method. Key performance indicators such as energy consumption, cooling capacity, and temperature fluctuation are extracted from the virtual environment simulation results, and multi-objective optimization is performed using the weighted summation method. Weights are set for each indicator, such as 40% for energy consumption, 30% for cooling capacity, and 30% for temperature fluctuation. After normalization, each indicator is weighted and summed to obtain a comprehensive performance score. The calculation formula is comprehensive score = 0.4*(1-energy consumption / maximum energy consumption)+0.3*cooling capacity / maximum cooling capacity+0.3*(1-temperature fluctuation / maximum temperature fluctuation). Based on the preset energy-saving operation target threshold, the threshold screening method is used to screen and sort the control combination schemes. Set a specific energy-saving target threshold, such as energy consumption not exceeding 80% of the baseline value. Screen out the schemes that meet the threshold and arrange them in descending order according to the comprehensive performance score. Select the control combination that meets the energy-saving target and has the highest comprehensive performance score as the optimal solution to achieve the energy-saving operation target under low-load conditions. When implementing the low-load optimization control of the refrigeration system, the enumeration method is first used to generate the control combination scheme. Assume that the system contains 2 chillers and 3 water pumps. Each device has two states: on and off. The operating frequency is divided into three levels: 30Hz, 40Hz and 50Hz. The valve opening is divided into three gears: 25%, 50% and 75%. Through nested loops, 2 are generated. 5 *3 5=7776 possible combinations. Next, the energy balance equation is used to construct a digital twin model. Taking the calculation of the cooling capacity of the chiller as an example, assuming that the chilled water flow rate is 100m³ / h, the specific heat capacity is 4.2kJ / (kg·℃), the return water temperature is 12℃, and the supply water temperature is 7℃, the cooling capacity Q=100*1000*4.2*(12-7) / 3600=583.3kW. Input each control combination into the model to calculate the system operation status. Extract key performance indicators from the simulation results, such as energy consumption 500kWh, cooling capacity 550kW, and temperature fluctuation ±0.5℃. The weighted summation method is used for multi-objective optimization, and the weights are set to 40% for energy consumption, 30% for cooling capacity, and 30% for temperature fluctuation. Assuming the maximum energy consumption is 800kWh, the maximum cooling capacity is 700kW, and the maximum temperature fluctuation is ±1°C, the comprehensive score of the combination = 0.4*(1-500 / 800)+0.3*550 / 700+0.3*(1-0.5 / 1)=0.6357. Finally, the selection is based on the energy-saving target threshold, such as setting the energy consumption to not exceed 80% of the baseline value of 600kWh, that is, 480kWh. The solutions that meet the conditions are selected and arranged in descending order according to the comprehensive score. The combination with the highest score is selected as the optimal solution, such as chiller No. 1 is turned on (40Hz), chiller No. 2 is turned off, water pumps No. 1 and No. 2 are turned on (50Hz), water pump No. 3 is turned off, and the valve opening is 50%.

[0019] Step S105, establish a low-load operation mode optimization model, use the screened control combination that meets the preset energy-saving operation target as the input of the low-load operation mode optimization model, and obtain the optimized system low-load operation mode by training the low-load operation mode optimization model.

[0020] The selected control combinations are preprocessed by Z-score standardization method. By calculating the mean and standard deviation of each feature, each data point is converted to (x-μ) / σ, where x is the original value, μ is the mean, and σ is the standard deviation. Based on the preprocessed data, the principal component analysis method is used for feature extraction. By calculating the feature covariance matrix and solving the eigenvalues ​​and eigenvectors, the principal components with a cumulative variance contribution rate of a preset percentage are selected. For the selected principal components, a linear regression model is constructed, with the principal components as independent variables and the performance indicators as dependent variables, and the regression coefficients are estimated by the least squares method. The constructed linear regression model is used as the low-load operation mode optimization model, and the grid search method is used for tuning. The range of values ​​of the regularization coefficient λ is defined, and the performance of each group of chiller, water pump and valve operating parameters is evaluated by cross-validation. According to the optimized model, the prediction results are converted into equipment operating parameters of the chiller start-stop state, water pump frequency and valve opening, and the system low-load operation mode is obtained.

[0021] Specifically, according to the selected control combinations that meet the preset energy-saving operation goals, the Z-score standardization method is used to preprocess the input data. The mean and standard deviation of each feature are calculated, and each data point is transformed: (x-μ) / σ, where x is the original value, μ is the mean, and σ is the standard deviation. Through this operation, outliers and duplicate data are eliminated, and parameters of different dimensions are unified to the same scale. The principal component analysis method is used to extract features from the preprocessed data to reduce the data dimension. The feature covariance matrix is ​​calculated, the eigenvalues ​​and eigenvectors are solved, and the number of principal components is selected according to the cumulative variance contribution rate. The threshold is set to 95%, and the characteristic variables with the most significant impact on the low-load operation mode of the system are selected as model inputs. A linear regression model is constructed as a low-load operation mode optimization model, with the selected principal components as independent variables and performance indicators as dependent variables. The least squares method is used to estimate the regression coefficient, and the relationship between the low-load operation mode of the system and the control parameters is established. The normal equation is solved by matrix operation to obtain the estimated value of the regression coefficient. The grid search method is used for model tuning. Define the parameter grid, such as the range of different values ​​of the regularization coefficient λ is [0.001, 0.01, 0.1, 1, 10]. Use 5-fold cross validation to evaluate the performance of each group of parameters, and calculate the mean square error and determination coefficient. Select the best parameter combination and update the model parameters. Apply the optimized low-load operation mode optimization model to the actual system control, convert the model prediction results into specific equipment operating parameters, such as the start and stop status of the chiller, the pump frequency and the valve opening, and output the optimized system low-load operation mode. When implementing the system low-load operation mode optimization, first perform Z-score standardization on the screened control combination data. Assume that there are 100 groups of control combinations, each containing 10 feature parameters. Calculate the mean and standard deviation of each feature, such as the mean of the chiller frequency is 40Hz and the standard deviation is 5Hz. Convert each data point, for example, the original frequency of 45Hz is converted to (45-40) / 5=1. Repeat this process for all features to obtain a standardized data set. Next, use principal component analysis to extract features. Calculate the 10×10 characteristic covariance matrix and solve the eigenvalues ​​and eigenvectors. Assuming that the first four principal components are required when the cumulative variance contribution rate reaches 95%, select these four principal components as new characteristic variables. Convert the 100×10 original data matrix to a 100×4 dimension reduction matrix. Then, construct a linear regression model with the four principal components as independent variables and the energy consumption index as the dependent variable. Use the least squares method to estimate the regression coefficients and obtain the regression equation. For example, energy consumption = 500-20×PC1+15×PC2-10×PC3+5×PC4, where PC1 to PC4 are the values ​​of the four principal components. Finally, the grid search method is used to tune the model. Set the value range of the regularization coefficient λ to [0.001, 0.01, 0.1, 1, 10], and use 5-fold cross validation to evaluate the performance of each set of parameters.Assume that the model performance is best when λ=0.1, the minimum mean square error is 10kWh², and the maximum determination coefficient is 0.95. Use the optimal parameters to update the model and convert the prediction results into specific operating parameters. For example, when the predicted energy consumption is the lowest, the corresponding principal component value is inversely solved to obtain chiller 1 on (frequency 35Hz), chiller 2 off, water pumps 1 and 2 on, with frequencies of 40Hz and 30Hz respectively, water pump 3 off, and valve opening 60%. This set of parameters is the optimized low-load operation mode.

[0022] Step S106, during the actual control execution process under the optimized system low-load operation mode, the operating parameters of the key equipment such as the chiller, water pump, and pipeline are collected in real time, combined with the cooling demand curve and the load demand forecast results, to obtain the set value of the chilled water temperature under low-load conditions, and adjust the cold storage capacity of the chiller according to the chilled water temperature set value.

[0023] Receive the collected equipment operation parameters, the equipment operation parameters include the cooling capacity of the chiller, compressor power, water pump flow, pipeline pressure and temperature; construct a real-time operation status data set according to the equipment operation parameters; use the exponential smoothing method to process the real-time operation status data set, wherein the smoothing coefficient α is a preset value, and obtain a short-term cooling demand prediction model; calculate the cooling demand change trend within a future preset time period according to the short-term cooling demand prediction model; if the cooling demand change trend increases and the current load rate is lower than the load rate threshold, the chilled water temperature set value is reduced by a preset temperature value; if the cooling demand change trend decreases and the current load rate is higher than the load rate threshold, the chilled water temperature set value is increased by a preset temperature value; obtain the temperature deviation between the chilled water temperature set value and the actual chilled water temperature; calculate the control amount according to the temperature deviation, the control amount is composed of a proportional term, an integral term and a differential term; convert the control amount into a compressor speed and a valve opening of the chiller; adjust the cold storage capacity of the chiller according to the compressor speed and the valve opening of the chiller.

[0024] Specifically, according to the optimized system low-load operation mode, the operating parameters of key equipment such as chillers, water pumps, and pipelines are collected in real time through a distributed sensor network, including chiller cooling capacity, compressor power, water pump flow, pipeline pressure and temperature, etc., to build a real-time operation status data set. Using the exponential smoothing method, combined with the historical cooling demand curve and the predicted load demand results, a short-term cooling demand prediction model is established. The smoothing coefficient α is selected as 0.3, and according to the formula S t =0.3*Y t +0.7*S t-1 Calculate the smoothed value where Y t is the actual observed value, S tis a smoothed value, and the trend of cooling demand changes in the next 2 hours is calculated. The if-then rule is used, and the real-time operating status data and short-term cooling demand forecast results are used as input variables. If the cooling demand is predicted to increase and the current load rate is lower than 50%, the chilled water temperature set value is reduced by 0.5°C; if the cooling demand is predicted to decrease and the current load rate is higher than 80%, the chilled water temperature set value is increased by 0.5°C to obtain the chilled water temperature set value under low load conditions. Based on the proportional-integral-differential control principle, the calculated chilled water temperature set value is compared with the actual chilled water temperature, and the temperature deviation e=set value-actual value is calculated. Use the PID controller to calculate the control quantity u=K p e+K i ∫edt+K d *de / dt, where K p , K i , K d are preset proportional, integral, and differential coefficients. The control quantity u is converted into specific values ​​of the compressor speed and valve opening to achieve dynamic adjustment of the cold storage capacity of the chiller. When the temperature decreases, the cold storage capacity is increased; when the temperature rises, the cold storage capacity is reduced, thereby adjusting the cold storage capacity of the chiller. When implementing the optimized system low-load operation mode, the key equipment operating parameters are first collected once a minute through the distributed sensor network. For example, the chiller cooling capacity is 200kW, the compressor power is 50kW, the water pump flow is 40m³ / h, the pipeline pressure is 300kPa, the chilled water supply temperature is 7°C, and the return water temperature is 12°C. These data constitute a real-time operating status data set. Next, the exponential smoothing method is used to predict the short-term cooling demand. Assume that the actual cooling demand Y at the current time t t is 210kW, and the smoothing value S at the previous moment t-1 The power is 200kW, and the smoothing coefficient α is selected as 0.3. The calculated smoothing value S t =0.3*210+0.7*200=203kW. Repeat this process to predict the trend of cooling demand in the next 2 hours. Then, apply the if-then rule to determine the chilled water temperature set point. If the predicted cooling demand rises to 220kW and the current load rate is 40% (less than 50%), reduce the chilled water temperature set point from 7℃ to 6.5℃. Finally, use the PID controller to adjust the chiller's cold storage capacity. Assuming the temperature deviation e=6.5-7=-0.5℃, preset K p =10, K i = 0.1, K d=1, and the control quantity u=10*(-0.5)+0.1*∫(-0.5)dt+1*d(-0.5) / dt=-5 is calculated. This control quantity is converted into an adjustment instruction to increase the compressor speed by 5% and the valve opening by 3%. By reducing the chilled water temperature and increasing the cold storage capacity, the cold storage capacity of the chiller can be dynamically adjusted.

[0025] In order to adjust the cold storage capacity of the chiller according to the chilled water temperature set value, the system reduces the chiller outlet temperature and adjusts the water pump flow according to the cooling demand curve to actively reduce the chilled water temperature set value. At the same time, the chiller's cold storage device is used to store cold energy and continuously optimize the chilled water temperature to enhance the cold storage capacity.

[0026] Obtain a cooling demand curve, calculate the average cooling demand within the time window based on the cooling demand curve, and obtain a predicted value of the future cooling change trend; use a proportional-integral controller to calculate the adjustment amount of the chilled water temperature setting value based on the predicted value of the cooling change trend, and determine the reduction range of the chiller outlet temperature; use a flow characteristic curve method to calculate the adjustment amount of the water pump flow based on the reduction range of the chiller outlet temperature, and determine the required valve opening change; use thermodynamic principles to establish a linear relationship model between cold storage and temperature change, and calculate the cold storage capacity of the current cold storage device; obtain the actual chilled water temperature, the energy storage state of the cold storage device, and the system load rate, and calculate the optimal adjustment amount of the chilled water temperature based on the actual chilled water temperature, the energy storage state of the cold storage device, and the system load rate, and determine whether the chilled water temperature setting value needs to be adjusted.

[0027] Specifically, based on the cooling demand curve, the moving average method is used to predict the cooling demand trend in the next 2 hours. A 30-minute time window is selected, the average cooling demand within the window is calculated, and the sliding window is used to predict the trend. The prediction result is converted into the adjustment amount of the chilled water temperature set value through the proportional integral controller, and the reduction range of the chiller outlet temperature is output. The flow characteristic curve method is used to calculate the adjustment amount of the water pump flow combined with the reduction range of the chiller outlet temperature. According to Formula, where For flow, is the valve flow coefficient, The required change in valve opening is calculated based on the pressure difference. At the same time, the pipeline pressure loss is considered to ensure the hydraulic balance of the system and realize the active reduction of the chilled water temperature. Using the principle of thermodynamics, based on the heat capacity of the chiller storage device and the characteristics of phase change materials, a linear relationship model between the cold storage capacity and temperature change is established. The formula ΔE=m*c*ΔT is adopted, where ΔE is the stored cold capacity, m is the mass of the cold storage medium, c is the specific heat capacity, and ΔT is the temperature change, to calculate the cold storage capacity of the current cold storage device. A feedback control mechanism based on fuzzy logic is constructed, and the actual chilled water temperature, the energy storage state of the cold storage device and the system load rate are used as input variables. Specific fuzzy rules are set. If the actual temperature is higher than the set value and the energy storage of the cold storage device is insufficient, the chilled water temperature is greatly reduced; if the actual temperature is close to the set value and the energy storage of the cold storage device is sufficient, the chilled water temperature is slightly adjusted. The optimal adjustment amount of the chilled water temperature is obtained through fuzzy reasoning, and the chilled water temperature setting value is continuously optimized, thereby increasing the temperature difference between the cold storage device and the environment, improving the heat exchange efficiency, and enhancing the cold storage capacity of the system. When implementing the optimal control of the chiller's cold storage capacity, the moving average method is first used to predict the trend of cooling capacity changes. Assuming that the current cooling demand is 500kW, the system collects data every 5 minutes, and selects a 30-minute time window. The average value of the last 6 data points is calculated to be 520kW, and it is predicted that the cooling demand will increase in the next 2 hours, and the peak value may reach 600kW. Based on this trend, the proportional-integral controller calculates that the chilled water temperature needs to be reduced by 0.5℃, and adjusts the chiller outlet temperature from 7℃ to 6.5℃. Next, the flow characteristic curve method is used to adjust the water pump flow. Assuming that the current valve opening is 60%, the flow rate is 100m³ / h, and the stress difference is 100kPa. According to Formula, it is calculated that the valve opening needs to be increased to 65%, and the corresponding flow rate needs to be increased to 110m³ / h to adapt to the change in flow demand caused by the temperature drop. Then, the cold storage capacity is calculated using the characteristics of the cold storage device. Assuming that the mass of the cold storage medium is 10,000kg, the specific heat capacity is 4.2kJ / (kg·K), and the cold storage capacity that can be stored for a temperature change of 1°C is ΔE=10000*4.2*1=42000kJ, which is approximately 11.67kWh of cold. Finally, the chilled water temperature is optimized based on the fuzzy logic control mechanism. The fuzzy rule is set that if the actual temperature is 7°C higher than the set value of 6.5°C, and the current energy storage of the cold storage device is only 5kWh (less than 50% of the maximum capacity), the chilled water temperature will be further reduced by 0.3°C to 6.2°C, which increases the temperature difference between the cold storage device and the environment, improves the heat exchange efficiency, and enables the system to cope with the possible 600kW cold demand peak in the next 2 hours, effectively enhancing the system's cold storage capacity.

[0028] Step S107, by analyzing the relationship between the energy consumption and cold storage capacity of the chiller and water pump, the optimal energy efficiency range is determined. During load demand forecasting and control optimization, this range is used as a target constraint to guide the chiller and water pump equipment to operate in a high energy efficiency area.

[0029] The input features including chiller frequency, water pump flow, chilled water temperature parameters, and output labels of energy consumption and cold storage are obtained; the support vector regression method is used to select the radial basis function kernel, and the kernel parameters and regularization coefficients are optimized through grid search to establish a relationship model between the input features and the output labels; the genetic algorithm is used to search for the optimal solution under the constraints of the relationship model, and the energy efficiency ratio is used as the fitness function. The boundary conditions of the optimal energy efficiency range are obtained through selection, crossover, and mutation operations; it is determined whether the predicted load falls within the optimal energy efficiency range. If so, the equipment operating parameters corresponding to the predicted load are directly used. If not, the nearest boundary value is selected as the operating parameter to generate an equipment operation plan; a controller based on rolling time domain optimization is constructed, and the optimal energy efficiency range is used as a constraint condition; in each control cycle, the controller dynamically adjusts the chiller frequency and water pump flow parameters by solving the optimization problem, and guides the equipment to operate in the high energy efficiency area in real time.

[0030] Specifically, the support vector regression method is used to establish a relationship model between equipment energy consumption and cold storage capacity. Historical operation data, including parameters such as chiller frequency, water pump flow, and chilled water temperature, are collected as input features, and energy consumption and cold storage capacity are used as output labels. The radial basis function kernel is selected, and the kernel parameters and regularization coefficients are optimized through grid search to train the support vector regression model to capture the nonlinear relationship between parameters. The genetic algorithm is used to search for the optimal solution under the constraints of the support vector regression model with the energy efficiency ratio as the fitness function. The chromosome encoding is defined to represent the operating parameters of the chiller and water pump, and the population size is set to 100 and the number of iterations is set to 1000. The boundary conditions of the optimal energy efficiency interval are obtained through selection, crossover, and mutation operations, and the optimal operating parameter range of the chiller and water pump is determined. Based on time series analysis, combined with historical load data and weather forecast information, the exponential smoothing method is used to predict the load demand curve for the next 24 hours. The predicted load is mapped to the optimal energy efficiency interval. If the predicted load falls within the interval, the equipment operating parameters corresponding to the predicted load are directly used; if it falls outside the interval, the nearest boundary value is selected as the operating parameter to generate the equipment operation plan. A controller based on rolling horizon optimization was constructed, with the optimal energy efficiency interval as a constraint. The prediction horizon was set to 1 hour and the control cycle was 5 minutes. An optimization problem was solved in each control cycle. The objective function was to minimize the total energy consumption in the next hour. The constraints included equipment operation restrictions and the optimal energy efficiency interval. By dynamically adjusting parameters such as chiller frequency and water pump flow, the equipment was guided to operate in the high energy efficiency zone in real time. When implementing energy efficiency optimal control, the support vector regression method was first used to establish a relationship model between equipment energy consumption and cold storage capacity. Operation data for the past three months were collected, including parameters such as chiller frequency 20-50Hz, water pump flow 50-150m³ / h, chilled water temperature 5-10℃, and corresponding energy consumption 100-500kWh and cold storage 200-1000kWh. A radial basis function kernel was selected, and the kernel parameter γ (0.01-1) and regularization coefficient C (1-100) were optimized by grid search. Finally, γ=0.1 and C=10 were determined. The trained model can predict energy consumption and cold storage with 90% accuracy. Next, the genetic algorithm is used to search for the optimal energy efficiency interval. The population size is set to 100, and the iteration is 1000 times. The chromosome encoding includes the chiller frequency, pump flow rate and chilled water temperature. After evolution, the optimal energy efficiency interval is obtained as the chiller frequency of 35-40Hz, the pump flow rate of 80-100m³ / h, and the chilled water temperature of 6-7℃. Then, the exponential smoothing method is used to predict the load demand curve for the next 24 hours, and the smoothing coefficient α=0.2 is selected. Assuming that the current load is 300kW, the predicted peak load is 450kW. Mapping the predicted load to the optimal energy efficiency interval, the equipment operation plan is obtained as follows: the chiller runs at 40Hz, the pump flow rate is 100m³ / h, and the chilled water temperature is 6℃ during the peak period; the valley period is adjusted to 35Hz, 80m³ / h and 7℃.Finally, a rolling time domain optimization controller is constructed to perform optimization calculations every 5 minutes to predict the energy consumption in the next hour. Assuming that in a certain control cycle, the predicted total energy consumption is 450kWh, it can be reduced to 420kWh by adjusting the parameters, achieving continuous operation of the equipment in the high energy efficiency area.

[0031] The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions and improvements made within the spirit and principles of the present invention should be included in the protection scope of the present invention.

Claims

1. A low-load chiller and water pump linkage energy-saving control method, characterized in that: The method comprises: By analyzing the historical operation data of the refrigeration system, combined with weather forecast information, equipment operating parameters, and data on the activities of people in the building, the load demand under low-load conditions is dynamically predicted to obtain the cooling demand curve and load demand forecast results within the preset time in the future; based on the cooling demand curve predicted under low-load conditions, a combined control strategy model of the chiller and water pump is constructed, and the system energy consumption is used as the model input to obtain the combined control strategy of the chiller and water pump under low-load conditions, and determine the optimal start and stop sequence and frequency adjustment plan of the chiller and water pump under different cooling demands; According to the combined control strategy of the chiller and water pump under low-load conditions, in the start-stop optimization sequence of the chiller and water pump, the temperature, pressure and flow parameters in the chilled water pipeline and the cooling water pipeline under low-load conditions are collected in real time to determine the degree of deviation between the actual low-load operating conditions of the system and the preset expected low-load operating conditions. If the deviation exceeds the deviation threshold, the valve opening is dynamically adjusted; The start and stop optimization sequence, frequency regulation scheme and valve opening adjustment of the chiller and water pump are integrated to generate different control combinations. By establishing a digital twin model of the refrigeration system, the low-load system operation status under different control combinations is simulated in a virtual environment to obtain key performance indicators such as energy consumption, cooling capacity and temperature fluctuation, and quantitative analysis and evaluation are performed to screen out control combinations that meet the preset energy-saving operation goals; a low-load operation mode optimization model is established, and the screened control combinations that meet the preset energy-saving operation goals are used as the input of the low-load operation mode optimization model. By training the low-load operation mode optimization model, the optimized system low-load operation mode is obtained; In the actual control execution process under the optimized system low-load operation mode, the operating parameters of the key equipment of the chiller, water pump, and pipeline are collected in real time, combined with the cooling demand curve and load demand forecast results, to obtain the set value of the chilled water temperature under low-load conditions, and adjust the cold storage capacity of the chiller according to the chilled water temperature set value; By analyzing the relationship between the energy consumption and cold storage capacity of chillers and water pumps, the optimal energy efficiency range is determined. During load demand forecasting and control optimization, this range is used as the target constraint to guide chillers and water pump equipment to operate in a high energy efficiency area.

2. The method according to claim 1, characterized in that The above method analyzes the historical operation data of the refrigeration system, combines weather forecast information, equipment operation parameters, and data on the activity patterns of building personnel, and dynamically predicts the load demand under low-load conditions to obtain the cooling demand curve and load demand prediction results within a preset time in the future, including: Acquire historical system operation data, process the historical system operation data using an exponential smoothing method, and obtain a preliminary load demand forecast result under a low-load condition; According to the preliminary load demand forecast result, combined with weather forecast information, a multiple linear regression model is established, wherein the multiple linear regression model is used to determine the mapping relationship between weather factors and load demand; The K-means algorithm is used to perform cluster analysis on the equipment operating parameters and the activity patterns of building personnel. If the Euclidean distance between the current working condition and a cluster center is the smallest, the sample most similar to the current working condition is extracted from the cluster. Calculating a reference value based on the sample, performing weighted averaging of the reference value, the preliminary load demand forecast result, and the output of the multivariate linear regression model to obtain an optimized load demand forecast result; A time window of fixed size is set, and the prediction process is repeated for each time point within the time window, and a continuous cooling demand curve is obtained by sliding the time window.

3. The method according to claim 1, characterized in that: The cooling demand curve predicted under low load conditions is used to construct a combined control strategy model of the chiller and water pump, and the system energy consumption is used as the model input to obtain the combined control strategy of the chiller and water pump under low load conditions, and determine the start and stop optimization sequence and frequency adjustment scheme of the chiller and water pump under different cooling demands, including: According to the cooling demand curve, a combined control model of a chiller and a water pump is constructed by using a mixed integer linear programming method; The mixed integer linear programming method uses system energy consumption as the objective function and the operating status of the chiller and the water pump as the decision variables; According to the control model of chiller and water pump combination, the genetic algorithm is used to optimize the search process to calculate the optimal chiller and water pump combination under different cooling requirements; According to the optimal chiller and water pump combination, an optimized start and stop sequence of the chiller and the water pump is formulated; Based on historical operating data, polynomial regression fitting is used to obtain the efficiency curves of chillers and water pumps under different loads; In combination with the real-time cooling demand and the efficiency curve, the optimal operating frequency is calculated using the gradient descent method to form a frequency adjustment plan.

4. The method according to claim 1, characterized in that: The combined control strategy of the chiller and the water pump under low-load conditions is described, in the start-stop optimization sequence of the chiller and the water pump, by real-time collection of the temperature, pressure, and flow parameters in the chilled water pipeline and the cooling water pipeline under low-load conditions, to determine the degree of deviation between the actual low-load operating condition of the system and the preset expected low-load operating condition, and dynamically adjust the valve opening if the deviation exceeds the deviation threshold, including: The temperature, pressure and flow parameters in the chilled water pipeline and cooling water pipeline are collected in real time through the distributed sensor network to obtain the original data under the low-load condition of the chiller; According to the original data, a moving average filtering method is used to remove noise to obtain parameter data after smoothing; For the smoothed parameter data, the minimum and maximum normalization method is used to scale it to a range of zero to one, and then the weighted average method is used to construct the characteristic vector of the actual low-load operating condition of the system, and the actual low-load operating condition index of the system is calculated according to the preset temperature, pressure and flow weights; Compare the actual low-load operating condition index of the system with the preset expected low-load operating condition index to determine whether the relative error exceeds the deviation threshold; If the relative error exceeds a deviation threshold, an adjustment is triggered; Based on the feedback control principle, combined with the system deviation degree and the current valve opening, the proportional integral differential control method is used to calculate the adjustment amount of the valve opening; The valve opening is dynamically adjusted by the actuator according to the adjustment amount of the valve opening.

5. The method according to claim 1, characterized in that The integrated start-stop optimization sequence, frequency regulation scheme and valve opening adjustment of the chiller and water pump are used to generate different control combinations. By establishing a digital twin model of the refrigeration system, the low-load system operation state under different control combinations is simulated in a virtual environment, and key performance indicators such as energy consumption, cooling capacity and temperature fluctuation are obtained. Quantitative analysis and evaluation are performed to screen out control combinations that meet the preset energy-saving operation goals, including: According to the start / stop status and frequency level of the chiller and water pump as well as the valve opening parameters, the enumeration method is used to generate the control combination scheme and construct a multi-dimensional control matrix. According to the multi-dimensional control matrix, a digital twin model of the refrigeration system is established using an energy balance equation and a mass balance equation; Inputting the parameters in the multidimensional control matrix into the digital twin model, and obtaining the system operating state under low-load conditions by a numerical solution method; Extracting the key performance indicators of energy consumption, cooling capacity and temperature fluctuation from the operating state of the system, and calculating the comprehensive performance score by weighted summation method, the calculation formula of the comprehensive performance score is comprehensive score = energy consumption weight * energy consumption normalized value + cooling capacity weight * cooling capacity normalized value + temperature fluctuation weight * temperature fluctuation normalized value; If the comprehensive performance score meets the preset energy-saving operation target threshold, the corresponding control combination scheme is determined as a scheme that meets the conditions; The solutions that meet the conditions are arranged in descending order according to their comprehensive performance scores, and the control combination with the highest comprehensive performance score is selected as the optimal solution.

6. The method according to claim 1, characterized in that The low-load operation mode optimization model is established, and the screened control combination that meets the preset energy-saving operation target is used as the input of the low-load operation mode optimization model, and the optimized system low-load operation mode is obtained by training the low-load operation mode optimization model, including: The screened control combinations were preprocessed using the Z-score standardization method, and each data point was converted to (x-μ) / σ by calculating the mean and standard deviation of each feature, where x is the original value, μ is the mean, and σ is the standard deviation; Based on the preprocessed data, feature extraction is performed using principal component analysis. By calculating the feature covariance matrix and solving the eigenvalues ​​and eigenvectors, the principal components whose cumulative variance contribution rate reaches a preset percentage are selected. For the selected principal components, a linear regression model is constructed, with the principal components as independent variables and the performance indicators as dependent variables, and the regression coefficients are estimated by the least squares method; The constructed linear regression model is used as the low-load operation mode optimization model and is tuned using the grid search method. According to the tuned model, the prediction results are converted into equipment operating parameters such as the start and stop status of the chiller, the water pump frequency, and the valve opening to obtain the low-load operation mode of the system.

7. The method according to claim 1, characterized in that In the actual control execution process under the optimized system low-load operation mode, the operating parameters of the chiller, water pump, and pipeline key equipment are collected in real time, and the set value of the chilled water temperature under low-load conditions is obtained in combination with the cooling demand curve and the load demand prediction result, and the cold storage capacity of the chiller is adjusted according to the chilled water temperature set value, including: Receiving collected equipment operating parameters, the equipment operating parameters including chiller cooling capacity, compressor power, water pump flow, pipeline pressure and temperature; Constructing a real-time operation status data set according to the equipment operation parameters; The real-time operation status data set is processed by exponential smoothing method to obtain a short-term cooling demand prediction model; Calculate the change trend of cooling demand within a preset time period in the future according to the short-term cooling demand prediction model; If the cooling demand change trend increases and the current load rate is lower than the load rate threshold, the chilled water temperature setting value is reduced by a preset temperature value; If the cooling demand change trend decreases and the current load rate is higher than the load rate threshold, the chilled water temperature setting value is increased by a preset temperature value; Obtaining a temperature deviation between the chilled water temperature setting value and the actual chilled water temperature; Calculating a control quantity according to the temperature deviation, wherein the control quantity is composed of a proportional term, an integral term and a differential term; The cold storage capacity of the refrigerator is adjusted according to the control amount.

8. The method according to claim 7, characterized in that In the actual control execution process under the optimized system low-load operation mode, the operating parameters of the chiller, water pump, and pipeline key equipment are collected in real time, and the set value of the chilled water temperature under low-load conditions is obtained in combination with the cooling demand curve and the load demand prediction result, and the cold storage capacity of the chiller is adjusted according to the chilled water temperature set value, which also includes: In order to adjust the cold storage capacity of the chiller according to the chilled water temperature set value, the system reduces the chiller outlet temperature and adjusts the water pump flow according to the cooling demand curve to actively reduce the chilled water temperature set value. At the same time, the chiller's cold storage device is used to store cold energy and continuously optimize the chilled water temperature to enhance the cold storage capacity.

9. The method according to claim 8, characterized in that The system adjusts the cold storage capacity of the chiller according to the chilled water temperature setting value, reduces the chiller outlet temperature and adjusts the water pump flow according to the cooling demand curve, and actively reduces the chilled water temperature setting value. At the same time, the chiller cold storage device is used to store cold, and the chilled water temperature is continuously optimized to enhance the cold storage capacity, including: Obtaining a cooling demand curve, calculating an average cooling demand in a time window according to the cooling demand curve, and obtaining a predicted value of a future cooling change trend; According to the predicted value of the cooling capacity change trend, a proportional-integral controller is used to calculate the adjustment amount of the chilled water temperature setting value to determine the reduction range of the chiller outlet temperature; According to the reduction of the outlet temperature of the chiller, the adjustment amount of the water pump flow is calculated by using the flow characteristic curve method to determine the required change in valve opening; The linear relationship model between cold storage capacity and temperature change is established using thermodynamic principles to calculate the cold storage capacity of the current cold storage device; The actual chilled water temperature, the energy storage status of the cold storage device and the system load rate are obtained, and the optimal adjustment amount of the chilled water temperature is calculated according to the actual chilled water temperature, the energy storage status of the cold storage device and the system load rate, to determine whether the chilled water temperature setting value needs to be adjusted.

10. The method according to claim 1, characterized in that The method analyzes the relationship between the energy consumption and cold storage capacity of the chiller and the water pump to determine the optimal energy efficiency range, and uses the range as a target constraint during load demand forecasting and control optimization to guide the chiller and water pump equipment to operate in a high energy efficiency area, including: Obtain input features including chiller frequency, pump flow and chilled water temperature parameters, and output labels of energy consumption and cold storage; Using a support vector regression method, selecting a radial basis function kernel, optimizing kernel parameters and regularization coefficients through grid search, and establishing a relationship model between the input features and the output labels; A genetic algorithm is used to search for the optimal solution under the constraints of the relationship model, and the energy efficiency ratio is used as the fitness function. The boundary conditions of the optimal energy efficiency range are obtained through selection, crossover and mutation operations. Determine whether the predicted load falls within the energy efficiency optimal range. If so, directly use the equipment operation parameters corresponding to the predicted load. If not, select the nearest boundary value as the operation parameter to generate an equipment operation plan. Constructing a controller based on rolling horizon optimization, taking the energy efficiency optimal interval as a constraint condition; In each control cycle, the controller dynamically adjusts the chiller frequency and water pump flow parameters by solving the optimization problem, and guides the equipment to operate in a high energy efficiency area in real time.

Images