Method for controlling chiller unit and control system thereof, device, storage medium and program product

Abstract

The invention provides a method for controlling a water chiller in a water chilling unit and a control system, equipment, a storage medium and a program product thereof, is used for improving the energy efficiency of the water chilling unit, and relates to the technical field of energy management. The method for controlling the water chiller in the water chiller unit comprises the steps that the load of the water chiller is predicted based on the time characteristic, the weather characteristic and the historical cooling load characteristic of the water chiller; according to the load prediction result and the partial load rate constraint condition, a control sequence used for controlling the water chiller is determined, and the control sequence comprises the starting number and the operation sequence of the water chiller in the water chiller unit; selecting an optimal control sequence from the control sequences according to the cooling-water machine model; and controlling the operation of the cooling-water machine based on the optimal control sequence.

Description

(19) State Intellectual Property Office (12) Invention Patent Application (10) Application Publication Number (43) Application Publication Date (21) Application Number 202511426158.X (22) Application Date 2025.09.30 (30) Priority Data 63 / 703,215 2024.10.04 US (71) Applicant Hong Kong University of Science and Technology Address Clear Water Bay, Kowloon, Hong Kong, China (72) Inventor Wang Zhe Li Shuhao Li Siqi (74) Patent Agency Beijing Ying Sai Jia Hua Intellectual Property Agency Co., Ltd. 11204 Patent Attorney Wang Dazuo Wang Yanchun (51) Int.Cl. F24F 11 / 89 (2018.01) F24F 11 / 88 (2018.01) G06F 30 / 27 (2020.01) G06F 30 / 28 (2020.01) G06F 17 / 11(2006.01) G06N 3 / 02(2006.01) G06N 5 / 01(2023.01) G06N 20 / 20(2019.01) G06F 111 / 04(2020.01) G06F 111 / 10(2020.01) G06F 113 / 08(2020.01) F24F 130 / 10(2018.01) (54) Invention Title: Method for Controlling a Chiller and its Control System, Equipment, Storage Medium and Program Product (57) Abstract: This disclosure provides a method for controlling a chiller in a chiller unit and its control system, equipment, storage medium and program product, to improve the energy efficiency of the chiller unit, relating to the field of energy management technology. A method for controlling the chillers in a chiller unit includes: predicting the load of the chillers based on time characteristics, weather characteristics, and historical cooling load characteristics of the chillers; determining a control sequence for controlling the chillers based on the load prediction results and partial load factor constraints, the control sequence including the number of chillers to be turned on and their operating order in the chiller unit; selecting the optimal control sequence from the control sequence based on the chiller model; and controlling the operation of the chillers based on the optimal control sequence. Claims 3 pages, Description 19 pages, Drawings 13 pages, CN 121804052 A 2026.04.07 CN 1 21 80 40 52 A 1. A method for controlling a chiller in a chiller unit, comprising: predicting the load of the chiller based on time characteristics, weather characteristics, and historical cooling load characteristics of the chiller; determining a control sequence for controlling the chiller according to the load prediction results and partial load factor constraints, the control sequence including the number of chillers to be turned on and their operating order in the chiller unit; selecting an optimal control sequence from the control sequence according to a chiller model; and controlling the operation of the chiller based on the optimal control sequence.2. The method according to claim 1, wherein predicting the load of the chiller based on time features, weather features, and historical cooling load features of the chiller comprises: receiving the time features, weather features, and historical cooling load features of the chiller through an XGBoost model; outputting a cooling load demand prediction curve through the XGBoost model; and predicting the load of the chiller based on the prediction curve. 3. The method according to claim 1, wherein the method further comprises: establishing a chiller model based on historical operating data of the chiller. 4. The method according to claim 3, wherein establishing the chiller model based on historical operating data of the chiller comprises: determining target feature parameters based on initial feature parameters related to the chiller; and establishing the chiller model in response to receiving the target feature parameters via an artificial neural network, and outputting the energy consumption and coefficient of performance of the chiller, wherein the initial feature parameters include at least one of the following: chiller unit operation list, chilled water flow rate, cooling water flow rate, chiller unit temperature point data, outdoor ambient temperature, total cooling load, and partial load factor. 5. The control method according to claim 4, wherein determining the target feature parameter based on the initial feature parameter includes: determining the effective range of the initial feature parameter using statistical methods of covariance matrix and confidence ellipse, and determining the initial feature parameter located within the effective range as the target feature parameter. 6. The control method according to claim 1, wherein selecting the optimal control sequence from the control sequence according to the chiller model includes: determining the energy consumption of the chiller unit corresponding to the control sequence within a specific time interval according to the chiller model; calculating the energy consumption of each control process in any control sequence according to the chiller model; and selecting the optimal chiller control sequence based on a comparison result between the energy consumption of multiple chiller units and the energy consumption of each control step in the control sequence controlling the operation of the chiller unit. 7. The control method according to claim 6, wherein, before calculating the energy consumption of each control process in any of the control sequences, the method further comprises: establishing a state transition equation to determine the control sequence and the cumulative energy consumption of the control sequence from a start time to time t, wherein a control action in the control sequence determines the on or off state of the chiller at the next time moment, where t is a positive integer; and storing the tuple if the tuple formed by the control sequence and the cumulative energy consumption is not stored. 8. The control method according to claim 1 or 6, wherein, selecting the optimal control sequence from the control sequences according to the chiller model comprises: searching for a control sequence that satisfies the cooling load and operating constraints when the partial load rate satisfies a first condition; andIf no control sequence is found, a control sequence satisfying the cooling load and operating constraints is searched when the partial load rate meets the second condition; wherein the numerical range of the second condition is greater than the numerical range of the first condition, and the numerical range of the first condition is the range where the partial load rate is close to the maximum value of the performance coefficient. 9. The control method according to claim 2, wherein determining the control sequence for controlling the chiller based on the load forecast result and the partial load rate constraint includes: generating a plurality of initial control sequences based on the cooling load demand forecast curve, the initial control sequences being limited by a minimum on and off interval; and determining the control sequence from the plurality of initial control sequences based on the partial load rate constraint. 10. A control system for controlling a chiller in a chiller unit, comprising: a load prediction module configured to predict the load of the chiller based on time characteristics, weather characteristics, and historical cooling load characteristics of the chiller; a screening module configured to determine a control sequence for controlling the chiller based on the load prediction result and partial load factor constraints, the control sequence including the number of chillers to be turned on and their operating order in the chiller unit; an optimization module configured to select an optimal control sequence from the control sequence based on a chiller model; and a control module configured to control the operation of the chiller based on the optimal control sequence. 11. The control system according to claim 10, wherein the load prediction module is specifically configured to: receive the time characteristics, weather characteristics, and historical cooling load characteristics of the chiller through an XGBoost model, output a cooling load demand prediction curve through the XGBoost model, and predict the load of the chiller based on the prediction curve. 12. The control system of claim 11, wherein the optimization module further comprises a chiller model trainer, configured to determine target feature parameters based on initial feature parameters related to the chiller, and in response to the artificial neural network receiving the target feature parameters, establish the chiller model and output the energy consumption and performance coefficient of the chiller, wherein the initial feature parameters include at least one of chiller operation list, chilled water flow rate, cooling water flow rate, chiller unit temperature point data, outdoor ambient temperature, total cooling load, and partial load factor. 13. The control system of claim 12, wherein the chiller model trainer is specifically configured to determine the effective range of the initial feature parameters using statistical methods of covariance matrix and confidence ellipse, and determine the initial feature parameters located within the effective range as the target feature parameters. 14. The control system of claim 10, wherein the optimization module is specifically configured to determine the energy consumption of the chiller unit corresponding to the control sequence within a specific time interval according to the chiller model, and based on the...The chiller model is described, and the energy consumption of each control process in any of the control sequences is calculated. Based on the comparison results of the energy consumption of multiple chiller units and the energy consumption of each control step in the control sequence controlling the operation of the chiller unit, the optimal chiller control sequence is selected. 15. The control system according to claim 14, wherein the optimization module is further configured to establish a state transition equation to determine the control sequence and the cumulative energy consumption of the control sequence from the start time to time t, wherein a control action in the control sequence determines the on or off state of the chiller at the next time moment, where t is a positive integer, and to store the tuple formed by the control sequence and the cumulative energy consumption if the tuple is not stored. 16. The control system according to claim 14, wherein the optimization module is further configured to search for a control sequence that satisfies the cooling load and operating constraints when the partial load rate satisfies a first condition, and to search for a control sequence that satisfies the cooling load and operating constraints when the partial load rate satisfies a second condition if no control sequence is found; wherein the numerical range of the second condition is greater than the numerical range of the first condition, and the numerical range of the first condition is the range of the partial load rate close to the maximum value of the performance coefficient. 17. The control system according to claim 10, wherein the screening module is specifically configured to generate a plurality of initial control sequences according to the cooling load demand prediction curve, the initial control sequences being limited by a minimum on and off interval, and to determine the control sequence from the plurality of initial control sequences based on the partial load rate constraints. 18. An electronic device, comprising: at least one processor; and a memory communicatively connected to the at least one processor; wherein the memory stores instructions executable by the at least one processor, the instructions being executed by the at least one processor to enable the at least one processor to perform the chiller control method according to any one of claims 1-9. 19. A non-transitory computer-readable storage medium storing computer instructions for causing the computer to perform the chiller control method according to any one of claims 1-9. 20. A computer program product comprising a computer program that, when executed by a processor, implements the steps of the chiller control method according to any one of claims 1-9. Claims 3 / 3 Page 4 CN 121804052 A Method for Controlling a Chiller and its Control System, Device, Storage Medium and Program Product Technical Field

[0001] This disclosure relates to the field of energy management technology, specifically to the field of optimizing chiller operation, and particularly to a method for controlling a chiller in a chiller unit and its control system, device, storage medium and program product. Background Art

[0002] HVAC (Heating, Ventilation and Air Conditioning) systems are comprehensive environmental control systems that integrate heating, ventilation, and air conditioning. Reducing the energy consumption of HVAC systems in large commercial buildings is crucial. Chillers account for more than 50% of the energy consumption of HVAC systems, and are responsible for providing cooling capacity to meet the cooling needs of the building space.

[0003] In large commercial buildings, chiller units comprising multiple chillers are typically used to meet the cooling load. The sequencing of multiple chillers within a chiller unit is a control strategy aimed at optimizing the on / off states of multiple chillers. By strategically controlling the on / off states of the chillers, they can operate at a high COP (Coefficient of Performance), achieving significant energy savings while meeting cooling requirements. Summary of the Invention

[0004] This disclosure proposes a method for controlling chillers in a chiller unit, as well as its control system, equipment, storage medium, and program products, to improve the energy efficiency of the chiller unit.

[0005] In a first aspect, embodiments of this disclosure propose a method for controlling a chiller in a chiller unit, comprising: predicting the load of the chiller based on time characteristics, weather characteristics, and historical cooling load characteristics of the chiller; determining a control sequence for controlling the chiller according to the load prediction results and partial load factor constraints, the control sequence including the number of chillers to be turned on and their operating order in the chiller unit; selecting an optimal control sequence from the control sequence according to a chiller model; and controlling the operation of the chiller based on the optimal control sequence.

[0006] In some embodiments, predicting the load of the chiller based on time characteristics, weather characteristics, and historical cooling load characteristics of the chiller includes: receiving time characteristics, weather characteristics, and historical cooling load characteristics of the chiller through an XGBoost model; outputting a cooling load demand prediction curve through an XGBoost model; and predicting the load of the chiller based on the prediction curve.

[0007] In some embodiments, the method further includes establishing a chiller model based on historical operating data of the chiller.

[0008] In some embodiments, establishing a chiller model based on historical operating data of the chiller includes: determining target characteristic parameters based on initial characteristic parameters related to the chiller; and, in response to the artificial neural network receiving the target characteristic parameters, establishing a chiller model and outputting the energy consumption and performance coefficient of the chiller, wherein the initial characteristic parameters include at least one of the following: chiller unit operation list, chilled water flow rate, cooling water flow rate, chiller unit temperature point data, outdoor ambient temperature, total cooling load, and partial load factor.

[0009] In some embodiments, determining the target characteristic parameters based on the initial characteristic parameters includes: using covariance moments...The statistical methods of arrays and confidence ellipses determine the effective range of the initial feature parameters, and the initial feature parameters within the effective range are determined as the target feature parameters.

[0010] In some embodiments, the optimal control sequence is selected from the control sequences according to the chiller model, including: according to the chiller model, determining the energy consumption of the chiller unit corresponding to the control sequence within a specific time interval; calculating the energy consumption of each control process in any control sequence according to the chiller model; and selecting the optimal chiller control sequence based on the comparison results of the energy consumption of multiple chiller units and the energy consumption of each control step in the control sequence that controls the operation of the chiller unit.

[0011] In some embodiments, before calculating the energy consumption of each control process in any control sequence, the method further includes: establishing a state transition equation to determine the control sequence and the cumulative energy consumption of the control sequence from the start time to time t, wherein a control action in the control sequence determines the on or off state of the chiller at the next time moment, where t is a positive integer; and storing the tuple formed by the control sequence and the cumulative energy consumption if it is not stored.

[0012] In some embodiments, selecting the optimal control sequence from the control sequences according to the chiller model includes: searching for a control sequence that satisfies the cooling load and operating constraints when the partial load rate satisfies a first condition; and searching for a control sequence that satisfies the cooling load and operating constraints when no control sequence is found, provided that the partial load rate satisfies a second condition. The numerical range of the second condition is greater than the numerical range of the first condition, and the numerical range of the first condition is the range where the partial load rate is close to the maximum value of the performance coefficient.

[0013] In some embodiments, determining a control sequence for controlling the chiller based on load forecast results and partial load factor constraints includes: generating multiple initial control sequences based on cooling load demand forecast curves, the initial control sequences being subject to minimum start-up and shut-down intervals; and determining a control sequence from the multiple initial control sequences based on partial load factor constraints.

[0014] In a second aspect, embodiments of this disclosure propose a control system for controlling chillers in a chiller unit, comprising: a load forecasting module, a screening module, an optimization module, and a control module. The load forecasting module is configured to forecast the load of the chiller based on time characteristics, weather characteristics, and historical cooling load characteristics of the chiller. The screening module is configured to determine a control sequence for controlling the chiller based on load forecast results and partial load factor constraints, the control sequence including the number of chillers to be turned on and their operating order in the chiller unit. The optimization module is configured to select the optimal control sequence from the control sequences based on a chiller model. The control module is configured to control the operation of the chiller based on the optimal control sequence.

[0015] In some embodiments, the load prediction module is specifically configured to: receive temporal features through the XGBoost model,Weather characteristics and historical cooling load characteristics of the chiller are used to output a cooling load demand prediction curve through the XGBoost model, and the load of the chiller is predicted based on the prediction curve.

[0016] In some embodiments, the optimization module further includes a chiller model trainer, which is configured to determine target feature parameters based on initial feature parameters related to the chiller, establish a chiller model in response to the artificial neural network receiving the target feature parameters, and output the energy consumption and performance coefficient of the chiller. The initial feature parameters include at least one of the following: chiller operation list, chilled water flow rate, cooling water flow rate, chiller unit temperature point data, outdoor ambient temperature, total cooling load, and partial load factor.

[0017] In some embodiments, the chiller model trainer is specifically configured to determine the effective range of the initial feature parameters using statistical methods of covariance matrix and confidence ellipse, and determine the initial feature parameters within the effective range as target feature parameters.

[0018] In some embodiments, the optimization module is specifically configured to determine the energy consumption of the chiller unit corresponding to the control sequence within a specific time interval according to the chiller model, calculate the energy consumption of each control process in any control sequence according to the chiller model, and select the optimal chiller control sequence based on the comparison results of the energy consumption of multiple chiller units and the energy consumption of each control step in the control sequence that controls the operation of the chiller unit.

[0019] In some embodiments, the optimization module is also configured to establish a state transition equation to determine the cumulative energy consumption of the control sequence and the control sequence from the start time to time t, where a control action in the control sequence determines the on or off state of the chiller at the next time, where t is a positive integer, and store the tuple formed by the control sequence and the cumulative energy consumption if it is not stored.

[0020] In some embodiments, the optimization module is further configured to search for a control sequence that satisfies the cooling load and operating constraints when the partial load rate meets a first condition, and to search for a control sequence that satisfies the cooling load and operating constraints when no control sequence is found, provided that the partial load rate meets a second condition. The numerical range of the second condition is greater than the numerical range of the first condition, and the numerical range of the first condition is the range where the partial load rate is close to the maximum value of the performance coefficient.

[0021] In some embodiments, the screening module is specifically configured to generate multiple initial control sequences based on a cooling load demand prediction curve, the initial control sequences being subject to minimum on / off interval restrictions, and to determine a control sequence from the multiple initial control sequences based on the partial load rate constraints.

[0022] In a third aspect, embodiments of this disclosure provide an electronic device comprising: at least one processor; and a memory communicatively connected to the at least one processor; wherein the memory stores data that can be executed by the at least one processor.The instructions are executed by at least one processor to enable the at least one processor to implement the chiller control method as described in any implementation of the first aspect.

[0023] In a fourth aspect, embodiments of the present disclosure provide a non-transitory computer-readable storage medium storing computer instructions for enabling a computer to implement the chiller control method as described in any implementation of the first aspect.

[0024] In a fifth aspect, embodiments of the present disclosure provide a computer program product including a computer program that, when executed by a processor, can implement the chiller control method as described in any implementation of the first aspect.

[0025] The method for controlling a chiller in a chiller unit provided by the present disclosure predicts the chiller load based on time characteristics, weather characteristics, and historical cooling load characteristics of the chiller. It can use only commonly used chiller sensor points, such as the temperature and flow rate of chilled water and cooling water, and chiller power, without the need for additional measuring equipment, ensuring that the present disclosure can be deployed on a large scale. Furthermore, based on the chiller model, selecting the optimal control sequence from the control sequence can take into account the dynamic impact of each step of the optimization strategy, ensuring accurate acquisition of the optimal control sequence. Obtaining the optimal control sequence can improve the energy efficiency of the chiller unit while meeting cooling requirements.

[0026] It should be understood that the content described in this section is not intended to identify key or important features of the embodiments of this disclosure, nor is it intended to limit the scope of this disclosure. Other features of this disclosure will become readily apparent from the following description.

[0027] Other features, objects, and advantages of this disclosure will become more apparent from the following detailed description of non-limiting embodiments with reference to the accompanying drawings: Figure 1 is a flowchart of a method for controlling a chiller in a chiller unit provided by an embodiment of this disclosure; Figure 2 is a detailed flowchart of the method for controlling a chiller in a chiller unit provided by an embodiment of this disclosure; Figure 3 is a schematic diagram of a chiller model provided by an embodiment of this disclosure; Figure 4 is a schematic diagram of a confidence ellipse provided by an embodiment of this disclosure; Figure 5 is a schematic diagram of a two-step optimization method provided by an embodiment of this disclosure; Figure 6a is a flowchart of a two-step optimization method provided by an embodiment of this disclosure; Figure 6b is a flowchart of a dynamic programming method provided by an embodiment of this disclosure; Figure 7 is a schematic diagram of dynamic programming provided by an embodiment of this disclosure; Figure 8 is a schematic diagram of the control module of the method for controlling a chiller in a chiller unit provided by an embodiment of this disclosure; Figure 9 is a schematic diagram of the simulation results of chiller sorting optimization provided by an embodiment of this disclosure; Figure 10 is a schematic diagram of the average duration calculated using an ANN model and a heuristic model, respectively; Figures 11a and 11b are schematic diagrams comparing the number of feasible control sequences before and after optimization using the two-step optimization method;Figures 12a and 12b are schematic diagrams comparing the computation time before and after using the DP method; Figure 13 is a schematic diagram of the computation time corresponding to the chiller sorting optimization using the DP method, the method combining the DP method with two-step optimization, and the exhaustive search method, respectively; Figure 14 is a schematic diagram of the average and extreme values ​​of the computation time corresponding to the chiller sorting optimization using the DP method, the method combining the DP method with two-step optimization, and the exhaustive search method, respectively. Detailed Embodiments

[0028] The exemplary embodiments of this disclosure are described below with reference to the accompanying drawings. Although various detailed descriptions are provided in these embodiments for ease of understanding, those skilled in the art should understand that these embodiments and their detailed descriptions are merely exemplary, and various modifications can be made to these embodiments without departing from the teachings of this disclosure. Similarly, for clarity and brevity, detailed descriptions of well-known functions and structures will be omitted in the following description. Furthermore, the embodiments and features in the embodiments of this disclosure can be combined with each other without conflict.

[0029] Traditional methods for controlling chillers in chiller units, such as rule-based control (RBC), are often not the optimal control methods. RBC typically regulates the on (i.e., online) and off (i.e., offline) states of chillers in two situations: first, when the measured cooling load exceeds a preset threshold; and second, based on the operator's experience. This method has two problems. The first is that it cannot utilize real-time operating data or weather forecasts to optimize the control strategy in a timely manner. The second is that the chiller's COP and actual cooling capacity are not constant but fluctuate with changing operating conditions. For example, centrifugal chillers are more efficient near full load, while screw chillers typically reach their maximum efficiency at partial load. However, traditional RBC methods, when regulating chillers, usually assume that the chiller's cooling capacity equals its rated capacity. This idealized assumption, in complex and variable actual operating environments, easily leads to the chiller failing to operate optimally, resulting in a decrease in overall efficiency.

[0030] Model Predictive Control (MPC) has become the most advanced solution for optimizing the control sequence of chillers. By integrating system models to predict future cooling loads, MPC provides a dynamic control strategy that minimizes energy consumption while ensuring operational constraints are met. In the current approach, the constructed models may include a load forecasting model and a chiller model. The load forecasting model predicts the chiller load over a future period based on current building conditions and weather forecasts. The chiller model describes the chiller performance and reflects the chiller's operating conditions.The relationship between the component and the chiller's COP is discussed. COP is defined as the ratio of the chiller's output cooling capacity to its input electrical energy. In practice, the chiller's COP usually fluctuates with changes in operating conditions (such as Partial Load Ratio, PLR). The MPC controller prioritizes the chiller with the highest COP in the chiller unit to ensure efficient operation while meeting the predicted cooling requirements. Therefore, the accuracy of the load prediction model and the chiller model directly affects the optimization performance of the MPC. Common models include physical models, gray box models, and black box models.

[0031] The physical model is based on thermodynamics and heat transfer theory and is constructed by collecting the specifications and physical details of the building or chiller equipment components. The constructed model can accurately describe the operating characteristics of the chiller. The gray box model is an improvement on the physical model, mainly relying on operating data and using statistical techniques to determine parameters.

[0032] However, physical models and gray box models are rarely used in practice due to the following problems. The first problem is that physical models and gray box models require an extremely complex parameter system to build and run, and collecting these parameters in large commercial buildings faces enormous challenges. Centralized cooling systems often suffer from insufficient measurement points, making it difficult to obtain the required parameter information comprehensively and accurately. The second problem is that physical models and gray box models require a great deal of expertise to build. The third problem is that physical models and gray box models do not consider the dynamic changes of the building environment and chiller units; they only optimize the chiller control sequence for a single time step without considering the impact of optimization on load changes and chiller performance in subsequent time steps. Therefore, physical models and gray box models often struggle to find the optimal chiller control sequence.

[0033] As building management systems become increasingly advanced, capable of collecting operational data hourly or even minutely, black box models (usually data-driven models) have gained attention. Black box models do not require explicit physical modeling but learn building dynamics from measurement data, offering high flexibility. However, current black box models have the following two problems: the first is the uncertainty in the data fitting model; control performance is only reliable within the trust range, and beyond this range, accuracy drops significantly, i.e., there is a data extrapolation problem. The second problem is the high computational requirements of the model, which may hinder real-time deployment. For example, the computation time may exceed the control interval.

[0034] Another aspect of MPC controller design is the selection of optimization methods. Common optimization methods for chiller control sequences can be broadly classified into heuristic algorithms and global optimization algorithms. Heuristic algorithms typically simulate natural evolution to approach the optimal solution. Although current heuristic algorithms are computationally efficient and well-suited for solving large-scale problems, they often only find suboptimal solutions.The chiller control sequence. Global optimization algorithms, such as linear programming (LP) and mixed-integer linear programming (MILP), can solve the chiller control sequence problem, but they have low computational efficiency. Especially when the number of chillers and the control range increase, the optimization time will be too long, making it unsuitable for practical application. In addition, the increase in model complexity also greatly exacerbates this problem, limiting the real-time deployment of the MPC method.

[0035] Based on the problems existing in the current methods for controlling chillers in chiller units, this disclosure provides a new method for controlling chillers in chiller units to accurately and quickly find the optimal control sequence in the control sequence of chillers, so as to improve the energy efficiency of chiller units.

[0036] The embodiments of this disclosure provide a method for controlling chillers in chiller units, as shown in Figure 1. The method includes: S101, predicting the load of chillers based on time characteristics, weather characteristics and historical cooling load characteristics of chillers.

[0037] S102. Based on the load forecast results and partial load factor constraints, determine the control sequence for controlling the chillers. The control sequence includes the number of chillers to be turned on and their operating order in the chiller unit.

[0038] S103. Based on the chiller model, select the optimal control sequence from the control sequences.

[0039] S104. Control the operation of the chillers based on the optimal control sequence.

[0040] The method for controlling the chillers in a chiller unit provided by the above embodiments predicts the chiller load based on time characteristics, weather characteristics, and the historical cooling load characteristics of the chillers. It can use only commonly used chiller sensor points, such as the temperature and flow rate of chilled water and cooling water, and the chiller power, without the need for additional measuring equipment, ensuring that this disclosure can be deployed on a large scale. In addition, according to the chiller model, selecting the optimal control sequence from the control sequence can take into account the dynamic impact of each step of the optimization strategy, ensuring that the optimal control sequence is accurately obtained. The acquisition of the optimal control sequence can improve the energy efficiency of the chiller unit while meeting the cooling demand.

[0041] The chiller sequencing optimization problem aims to determine the optimal control sequence in the control sequence used to control the chillers, so as to minimize the total energy consumption while meeting the cooling demand and operating constraints. The objective function is to minimize the total power consumption of all chillers within the control range. This requires accurately predicting the cooling load of the chillers and determining the best operating sequence of the chillers in each control time step (i.e., determining the optimal control sequence of the chillers in the chiller unit in each control step), so as to meet the cooling load with the minimum energy consumption. The control step refers to the interval between executing control commands.The optimization objective is shown in formula (1.a):

[0042] Where Et is the total energy consumption of the chiller at time t, derived from the chiller model, and N is the number of time intervals. To ensure the feasibility and practicality of the chiller sorting optimization problem, the optimization problem needs to satisfy the constraints in formulas (1.b)-(1.e).

[0043]

[0044] Where CLt is the cooling load of the chiller unit, PLRt is the partial load rate of the i-th chiller at time t, is the capacity of the i-th chiller, M is the number of chillers, is the predicted power of the chiller model, xt is the input value of the chiller model at time t, and w is the parameter of the chiller model. In addition, the ratio of MUT (Minimum Up Time) and MDT (Minimum Down Time) reflects the physical limitations of the chiller, and this ratio needs to meet certain conditions to prevent frequent start-up and shutdown of the chiller. In addition, the PLR ​​of each chiller must be kept within the specified range to maintain the efficient operation of the chiller.

[0045] The control action vector ut is the control sequence of multiple chillers at time t, which determines the on or off state of the chiller at the next time. Therefore, st+1=ut, where st+1 is the state vector of multiple chillers at time t+1. Assuming that the PLR ​​of all chillers in the on state is the same, the total cooling capacity provided by the chillers at any given time must meet the predicted cooling load, which can ensure that the cooling load of the building is always met.

[0046] As shown in Figure 2, the key to the chiller sorting optimization problem is to quickly and accurately find the optimal control sequence in the control sequence used to control the chillers. The purpose of the screening process is to screen out feasible control sequences from all control sequences, that is, to determine the control sequence used to control the chillers in S102 above. The purpose of the optimization process is to quickly and accurately select the optimal control sequence from the feasible control sequences, that is, to select the optimal control sequence in S103 above. The chiller model in Figure 2 can be seen in Figure 3.

[0047] In some embodiments, the above-mentioned S101, which predicts the chiller load based on time characteristics, weather characteristics, and historical cooling load characteristics of the chiller, includes: receiving time characteristics, weather characteristics, and historical cooling load characteristics of the chiller through an XGBoost model; outputting a cooling load demand prediction curve through an XGBoost model; and predicting the chiller load based on the prediction curve, as shown in Figure 2.

[0048] In the chiller sorting optimization problem, accurate prediction of the chiller load is crucial because accurate load prediction can provide advance insight into future cooling demand, providing data support and decision-making for the efficient operation of the chiller.Policy Basis. The load forecasting model provides crucial input data for the subsequent optimization process. This disclosure selects the Extreme Gradient Boosting (XGBoost) method to construct the load forecasting model. XGBoost is a decision tree-based ensemble algorithm widely used in various forecasting tasks due to its high accuracy and efficient training speed, such as predicting financial electricity load and building energy efficiency. XGBoost effectively handles large-scale data and high-dimensional features in building load forecasting tasks, providing reliable predictions. XGBoost effectively handles large-scale data and high-dimensional features in load forecasting tasks by continuously combining multiple predictors, thus providing reliable predictions. Specifically, new predictors improve prediction accuracy by focusing on and correcting the prediction errors of the previous predictor. To prevent overfitting, the XGBoost model integrates multiple weak predictors (typically shallow decision trees) and regularization terms into the objective. The goal of the XGBoost model is to minimize the objective function of each generation, as shown in Equation (2):

[0049] where i is the i-th sample to be predicted, n is the total number of samples, t is the t-th iteration, is the loss function between the true label and the predicted label, is the base learner added in the t-th iteration, is the feature of the i-th sample, is the regularization term used to avoid overfitting, and represents the objective function in the t-th iteration.

[0050] The XGBoost model describes the relationship between input data and the cooling load of a future building, so the selection of input data is crucial. The cooling load of a building is caused by internal and external heat gain. Internal heat gain is mainly caused by the behavior of occupants, appliances, and lighting, and usually depends on the building's operating schedule. For example, the typical usage time of a commercial building is usually from 10 am to 10 pm, and the use of appliances usually increases on weekends and holidays. External heat gain usually refers to solar radiation, air convection heat transfer, and heat conduction, and is mainly affected by outdoor weather conditions. Furthermore, due to the building's thermal mass, the building's current cooling load is affected by the cooling load of the previous period. Therefore, the input features of the XGBoost model can be time features, weather features, and historical cooling load features of the chiller.

[0051] For example, cooling load exhibits a clear time cycle. Features such as hours, days of the week, and months can provide information on intraday variations, differences between weekdays and weekends, and seasonal variations. Therefore, time features can be selected from features such as hours, days of the week, and months.

[0052] For example, weather directly affects cooling load. Features such as temperature, humidity, and solar radiation help the XGBoost model capture changes in cooling load. Therefore, weather features can be selected from features such as temperature, humidity, and solar radiation.

[0053] For example, cooling load has a strong time dependence. Inputting historical cooling load data can improve...Accuracy and reliability of the XGBoost model.

[0054] Table 1 shows the input features of the XGBoost model. In the time-related features, the hour, weekday, and month data can include data transformed by sine (sin) and cosine (cos) in addition to the original data. In addition, the wind direction, azimuth, and altitude in the weather features can include data transformed by sine (sin) and cosine (cos) in addition to the original data. In addition to using the original data, some feature data in the time-related features and weather features can also use data transformed by sine and data transformed by cosine, in order to better capture the periodic changes of cooling load.

[0055] Table 1

[0056] All features in Table 1 can be normalized using the data standardization method (StandardScaler). The coefficient of variation of root mean square error (CVRMSE) is used to evaluate the accuracy of the XGBoost model, as defined by formulas (3.a)-(3.b). Root mean square error (RMSE) measures the average of prediction errors, while CVRMSE normalizes RMSE, making it a relative measure of prediction accuracy. CVRMSE is used to evaluate the XGBoost model because it provides a standardized metric that allows for more meaningful comparisons across different datasets.

[0057]

[0058] Where y is the actual cooling load value, and is the predicted cooling load value. n is the number of samples. is the average cooling load.

[0059] Hyperparameter tuning is crucial for improving the performance of the XGBoost model. Bayesian optimization (BO), as the most advanced hyperparameter optimization (HPO) algorithm, has higher efficiency and convergence. Specifically, this disclosure selects the Tree-Structured Parzen Estimator (TPE) from the Hyperopt library as the HPO algorithm. When using HPO to optimize the hyperparameters of the XGBoost model, a reasonable search space needs to be defined, including hyperparameters such as maximum depth, learning rate, number of estimators, and regularization parameters. Table 2 shows the range of hyperparameter search.

[0060] Table 2

[0061] In some embodiments, the method of controlling the chiller in the chiller unit of this disclosure further includes: according to the chillerThe historical operating data of the chiller is used to establish a chiller model. For example, the historical operating data of the chiller is updated weekly. The chiller model is used to describe the relationship between the operating conditions of the chiller unit and its COP.

[0062] In some embodiments, establishing a chiller model based on the historical operating data of the chiller includes: determining target feature parameters based on initial feature parameters related to the chiller; and in response to the artificial neural network receiving the target feature parameters, establishing a chiller model and outputting the energy consumption and performance coefficient of the chiller. The initial feature parameters include at least one of the following: chiller unit operation list, chilled water flow rate, cooling water flow rate, chiller unit temperature point data, outdoor ambient temperature, total cooling load, and partial load factor.

[0063] In some embodiments, determining the target feature parameters based on the initial feature parameters includes: using statistical methods of covariance matrix and confidence ellipse to determine the effective range of the initial feature parameters, and determining the initial feature parameters within the effective range as the target feature parameters. The chiller model in this disclosure is constructed based on target feature parameters (i.e., the chiller model in this disclosure is constructed using an artificial neural network method combined with a trust region), thus avoiding data interpolation and improving the accuracy of the chiller model.

[0064] The chiller model in this disclosure is described in detail below.

[0065] For a chiller, COP represents the ratio of cooling capacity to the electrical energy input required for the chiller to generate cooling, as shown in formula (4).

[0066]

[0067] Artificial Neural Network (ANN) is good at handling multivariable and nonlinear problems and can automatically extract features from a large amount of data for dynamic chiller energy consumption prediction. The objective function J(w) of the ANN model is to continuously update the weight parameter w, thereby narrowing the gap between the predicted value and the true value, as shown in formula (5). Specification 9 / 19 pages 13 CN 121804052 A

[0068]

[0069] Where, P is the sample size, x is the input of the ANN model, is the predicted value, and is the true value.

[0070] The inputs (i.e., initial feature parameters) to the ANN model may include at least one of the following: chiller unit operation list(s), chilled water flow rate, cooling water flow rate, chilled water supply temperature, cooling water return temperature, outdoor ambient temperature, total cooling load, and partial load factor (PLR). These parameters directly affect the chiller's energy efficiency and cooling load variation. Table 3 shows an explanation of the chiller model input parameters. By inputting these parameters, the ANN model can effectively predict the energy consumption and COP during chiller operation, thereby optimizing the overall energy consumption of the chiller unit. Figure 3 is a schematic diagram of the chiller unit ANN model. Additionally, the hyperparameters of the ANN model are shown in Table 4.

[0071] Table 3

[0072] Table 4

[0073] However, data-driven models (such as ANN models) often encounter the problem of data extrapolation. When the input data exceeds the range of the training data, the performance of the ANN model becomes very unreliable. In order to increase the reliability of the ANN model, this disclosure will process the initial feature parameters using the statistical methods of covariance matrix and confidence ellipse to determine the effective range of the initial feature parameters. The initial feature parameters within the effective range will be determined as target feature parameters. The target feature parameters will be combined with the ANN model. That is, in order to increase the reliability of the ANN model, this disclosure proposes an ANN method that combines the confidence region, which can avoid the problem of data extrapolation and enhance the stability and reliability of the system.

[0074] The covariance matrix is ​​used to describe the linear relationship between multiple variables. For the two variables of cooling water temperature difference (x) and chilled water temperature difference (y), the definition of the covariance matrix C(x, y) is shown in formula (6).

[0075]

[0076] Where Var(x) and Var(y) are the variances of x and y respectively, and Cov(x,y) is the covariance of x and y, as shown in formula (7).

[0077]

[0078] Based on the covariance matrix, the Pearson correlation coefficient ρ is calculated to control the rotation angle of the confidence ellipse, as shown in formula (8).

[0079]

[0080] The major and minor axes of the confidence ellipse are determined by the standard deviation of the variables:

[0081] Where, and are the eigenvalues ​​of the covariance matrix. x is the confidence level, which is a multiple of the standard deviation (e.g.).

[0082] After completing the affine transformation of the ellipse, multiple confidence ellipses are drawn according to different confidence intervals. The ellipse represents 95% of the data located inside the ellipse, serving as the confidence interval of the variable. Figure 4 is a schematic diagram of the confidence ellipse, where the dashed line represents the confidence range of the chilled water temperature. When the variables of the feasible control sequence exceed this range, the calculation results of the operating conditions will be considered unreliable.

[0083] In some embodiments, this disclosure can also construct a chiller model through a heuristic model, which is a second-order regression model based on empirical parameters. The quadratic term is introduced to better capture the nonlinear relationship between the chiller COP, load rate, chilled water temperature and cooling water temperature, while ensuring the simplicity of the empirical model and reducing the computational complexity. The formula of the heuristic model is shown in formula (10):

[0084] Wherein, PLR is the partial load rate of the chiller, Tccw,s is the cooling water supply temperature, which reflects the heat dissipation capacity of the chiller, and Tchw,s is the chilled water supply temperature, which reflects the cooling capacity of the chiller. The difference between Tccw,s and Tchw,s directly reflects the efficiency of the chiller, so these two variables are introduced into the COP variation model. The coefficients are for regression fitting.The parameters, calibrated using the least squares method based on the unit's operating data, are used to describe the combined impact of PLR and cooling / chilled water temperature difference on COP. Specifically, is the intercept term, is the load first-order term coefficient, quantifying the linear impact of PLR on COP, is the load second-order term coefficient, capturing the nonlinear impact of PLR on COP, is the temperature difference first-order term coefficient, is the temperature difference second-order term coefficient, describing the nonlinear impact of ΔT on COP, and is the interaction term coefficient, describing the coupling effect between PLR and ΔT.

[0085] The initial chiller sequencing optimization method adopted the RBC method, which is the benchmark for comparison with the MPC method. The goal of chiller sequencing optimization is to minimize cumulative energy consumption, so the performance index is the energy saving rate, as shown in formula (11).

[0086]

[0087] Wherein, JRBC is the total energy consumption of chiller ranking optimization using the RBC method, and JMPC is the total energy consumption of chiller ranking optimization using the MPC method.

[0088] In some embodiments, the control sequence for controlling the chiller is determined based on the load forecast results and partial load factor constraints, including: generating multiple initial control sequences based on the cooling load demand forecast curve, wherein the initial control sequences are subject to minimum start-up and shut-down intervals; and determining the control sequence from the multiple initial control sequences based on the partial load factor constraints.

[0089] For example, as shown in FIG2, all control sequences in FIG2 are initial control sequences, and the feasible control sequences in FIG2 are the determined control sequences. Multiple initial control sequences from t+1 to t+k can be generated based on the cooling load demand forecast curve, and the initial control sequence that satisfies the partial load factor constraints is the determined feasible control sequence.

[0090] This disclosure can use dynamic programming and two-step optimization methods to select the optimal control sequence from multiple feasible control sequences. The dynamic programming method can solve the problem of low computational efficiency caused by a large number of feasible control sequences. The two-step optimization method can solve the problem that some load rate constraints may be very strict, and no feasible control sequence can be found, thus leading to the inability to further optimize.

[0091] Each feasible control sequence F in this disclosure is a control sequence of length N, as shown in formula (12).

[0092]

[0093] Where st is the combination of chiller switching states at time t, si,t is the state of the i-th chiller at time t, and si,t can be 0 or 1. N is the time interval, calculated by formula (12.c), M is the total number of chillers, T is the control range, and Δt is the control interval.

[0094] To solve the energy consumption of feasible control sequences, it is necessary to use the ANN model of each chiller at each time. Assuming that the number of feasible control sequences is Q, the total computation is MN.Q. Furthermore, the number of feasible control sequences increases with the increase of time intervals, leading to an exponential increase in computational complexity. Therefore, as the number of time intervals increases, the computation time may become excessively long, potentially exceeding the length of the control time interval. To address the problem of long computation time, this disclosure employs dynamic programming to find the optimal control sequence, which will be described in detail below.

[0095] Additionally, when partial load factor constraints are stringent, feasible control sequences may not be found. For example, to ensure the COP of the chiller unit and limit the number of feasible solutions, the PLR ​​constraint range is narrow, easily resulting in a situation where no feasible control sequence exists. To solve this problem, this disclosure employs a two-step optimization method to screen feasible control sequences, which will be described in detail below. Instruction manual, pages 12 / 19, 16 CN 121804052 A

[0096] In some embodiments, according to the chiller model, the optimal control sequence is selected from the control sequence, including: searching for a control sequence that satisfies the cooling load and operating constraints when the partial load rate meets the first condition; and searching for a control sequence that satisfies the cooling load and operating constraints when no control sequence is found, when the partial load rate meets the second condition. Wherein, the numerical range of the second condition (the second range in Figure 5) is greater than the numerical range of the first condition (the first range in Figure 5), and the numerical range of the first condition is the range where the partial load rate is close to the maximum value of the performance coefficient. This method uses a two-step search to find feasible control sequences, which can improve optimization efficiency and avoid the situation where no feasible control sequence exists.

[0097] Specifically, Figure 5 shows the two-step optimization process. In the first step, the focus of the optimization process is to narrow the range of PLR. By limiting PLR, priority is given to the configuration that is closer to the optimal efficiency point (i.e., higher COP) when the chiller is running, that is, searching when PLR meets the first condition. The second step is to relax the PLR ​​constraints if the initial search under the PLR ​​constraint does not find a control sequence that satisfies the cooling load and operating constraints. This allows for a broader search, specifically searching when the PLR ​​satisfies the second condition. The specific process of the two-step optimization is shown in Figure 6a. The specific method of the two-step optimization includes steps S201 to S204. If the control sequence obtained in the first search is empty, step S203 is executed for the second search. If the control sequence obtained in the first search is not empty, step S204 is executed directly.

[0098] The first step in the two-step search optimization method described above significantly reduces the number of feasible control sequences required, thereby accelerating the optimization process. The PLR ​​constraint range is shown in formula (1.d). The second step in this two-step search optimization method relaxes the PLR ​​constraints to allow for a broader search by considering more chiller operations.The state ensures that a feasible solution is found for subsequent optimization. Specifically, formula (1.d) is replaced with formula (13).

[0099]

[0100] Dynamic Programming (DP) is a method for solving complex optimization problems. It decomposes the original problem into overlapping subproblems and stores the solutions to these subproblems to avoid redundant calculations. DP is particularly suitable for multi-stage, sequential decision problems, especially when the objective function is indivisible. Problems that can be effectively solved by DP usually have the following characteristics: First, optimal substructure, the optimal solution of the problem can be constructed from the optimal solutions of its subproblems. Second, overlapping subproblems, the same subproblems are repeatedly solved during recursive decomposition. Third, no sequelae, once a subproblem is solved, it will not be affected by future decisions. The chiller sorting optimization problem has these characteristics, so it is suitable to use the DP method.

[0101] In some embodiments, selecting the optimal control sequence from the control sequences according to the chiller model includes: determining the energy consumption of the chiller unit corresponding to the control sequence within a specific time interval according to the chiller model; calculating the energy consumption of each control process in any control sequence according to the chiller model; and selecting the optimal chiller control sequence based on the comparison results of the energy consumption of multiple chiller units and the energy consumption of each control step in the control sequence that controls the operation of the chiller unit.

[0102] In some embodiments, before calculating the energy consumption of each control process in any control sequence, the method further includes: establishing a state transition equation to determine the control sequence and the cumulative energy consumption of the control sequence from the start time to time t, wherein a control action in the control sequence determines the on or off state of the chiller at the next time, where t is a positive integer; and storing the tuple formed by the control sequence and the cumulative energy consumption if it is not stored.

[0103] Specifically, according to the chiller model, the optimal control sequence is selected from the control sequence, including the following steps: First, decomposition of the original problem: The original problem is decomposed into several stages, each stage corresponding to multiple sub-problems, and the characteristics of the sub-problems are defined as states. In the chiller sequencing problem, the sub-problems involve determining the energy consumption of the chiller unit within a specific time interval Δt. The state is the combination of the on or off states of the chiller unit at time t, and the decision variable is the on or off operation of each chiller instruction manual (page 13 / 19, CN 121804052 A).

[0104] Second, establishing a recursive relationship: This includes establishing a state transition equation. The optimal strategy has a characteristic that, regardless of the initial state and initial decision, the remaining decisions must constitute the optimal strategy related to the state produced by the first decision. Therefore, the recursive relationship can be determined as:

[0105] Where, ut is the decision operation at time t, i.e., the state switch of the chiller unit, and E(ut) is the state of the chiller unit after executing operation ut.The energy consumption of the interval is calculated by the ANN model, and Vt(ut) is the cumulative energy consumption from the start to time t.

[0106] The third step is to solve the sub-problems in sequence: this includes using the ANN model to solve the energy consumption of each chiller state combination in each time interval. Since this process involves repeatedly solving sub-problems, storing the solutions to previously calculated sub-problems can speed up the overall calculation. Figure 6b shows how the dynamic programming method reduces the amount of computation by storing and memorizing sub-problems.

[0107] In some embodiments, the "solving sub-problems in sequence" includes: solving the energy consumption for each chiller state combination in each control time interval and storing the calculated sub-problem results to avoid repeated calculations and reduce the overall amount of computation; the process is shown in Figure 6b and corresponds to the node-side structure in Figure 7.

[0108] 1. Definition of State and Transition

[0109] In some embodiments, the state is used to represent the operating combination and necessary constraint information of the chiller unit at a given time, specifically including: chiller switching vector; cumulative time quantity related to minimum start / stop time (MUT / MDT); PLR determined by the operating combination and necessary features (such as temperature, flow rate, etc.) for energy consumption estimation.

[0110] In some embodiments, the control action causes the state to transition from to, and the transition is considered feasible and retained in the directed edge of Figure 7 only if the cooling load constraint, PLR constraint, MUT / MDT constraint and parallel / switching operating constraints are satisfied.

[0111] 2. The first calculation step in Figure 6b (corresponding to the first layer of Figure 7)

[0112] In some embodiments, the "first calculation step" is not to calculate the total energy consumption of the entire sequence at once, but rather: generate a set of feasible states at a given time; for each, call the chiller model (e.g., ANN model) to calculate the stage energy consumption of that time interval and obtain the initial value of the cumulative energy consumption.

[0113] Accordingly, the first calculation step in Figure 6b corresponds to the "Stage Energy Consumption Assessment of First Layer Nodes" in Figure 7.

[0114] 3. The second calculation step in Figure 6b (cross-layer recursive merging)

[0115] In some embodiments, the "second calculation step" is used to perform Bellman recursion: - For each feasible state at time t, calculate

[0116] and record the predecessor state pointer prev that reaches the minimum value.

[0117] - Proceed sequentially to t+K in the above manner.

[0118] Accordingly, the second calculation step in Figure 6b corresponds to the "Feasible transition and cumulative value across adjacent time points" in Figure 7, and is not a comparison of parallel energy consumption between the same time points.

[0119] 4. Meaning of the judgment (state) step and state structure

[0120] In some implementations, the "judgment state step" in Figure 6b is used to determine the feasibility and prune each layer state and its transition, including: Cooling load satisfies: The capacity of the unit being turned on (determined by PLR / model) is not less than the predicted load; MUT / MDT satisfies: The cumulative on / off time satisfies the minimum time limit; PLR interval: Prioritize searching within the first condition (PLR window close to the COP peak); If there is no solution, relax the second condition as shown in Figure 6a; - Operational constraints such as parallel upper limit and minimum switching interval.

[0121] When any constraint is not satisfied, the corresponding state or transition is directly discarded to reduce the amount of subsequent calculations. The state structure on which the above judgment is based is the switch vector, timing information and energy consumption assessment features included above.

[0122] 5. Memory storage and sub-problem reuse ("Storage and memory" in Figure 6b)

[0123] In some implementations, in order to avoid repeatedly calling the ANN model when recursively iterating in subsequent layers, cache entries are established for the sub-problems that have been calculated and their results are stored. When the same problem recurs, the cached result is read directly without repeated solving, thus significantly shortening the overall computation time.

[0124] 6. Termination and Search

[0125] In some implementations, the minimum cumulative energy consumption is selected at the predicted terminal time, and the optimal control sequence from t to t is obtained by backtracking along the prev() pointer. Since the optimality principle of dynamic programming (the optimal substructure overlaps with the subproblem) is adopted and a covering search is performed on all feasible states and transitions (if necessary, the PLR ​​is relaxed as shown in Figure 6a to ensure feasibility), the obtained control sequence is globally optimal.

[0126] Through these steps, the DP method can effectively optimize the chiller sequencing problem, ensuring that the cooling requirements are met while minimizing energy consumption. Figure 7 is a schematic diagram of the DP method. After finding the optimal control sequence for the first t+1 time, the subsequent feasible solutions will branch into different control sequences. Calculating the subproblem at the next time requires recalculating the subproblems at all previous time. Therefore, the subproblems at previous time will be stored to avoid repeated calculations and speed up the optimization.

[0127] For example, exhaustive search can be selected as the performance benchmark for DP methods. Exhaustive search obtains the optimal solution by calculating the energy consumption of each feasible solution one by one. This method is chosen because it can find the global optimum like DP programming. However, this method requires calling an ANN model to evaluate each feasible solution, which has a high computational cost.

[0128] For example, the above-described method of this disclosure can be encapsulated into a module program for low-cost deployment and commercialization.

[0129] This disclosure also provides a control system for controlling the chiller in a chiller unit, as shown in FIG2, including load.The load prediction module 101, the screening module 102, the optimization module 103, and the control module (not shown in the figure) are configured to predict the load of the chiller based on time characteristics, weather characteristics, and historical cooling load characteristics of the chiller. The screening module 102 is configured to determine the control sequence for controlling the chiller based on the load prediction results and partial load factor constraints. The control sequence includes the number of chillers to be turned on and the operating order of the chillers in the chiller unit. The optimization module 103 is configured to select the optimal control sequence from the control sequence based on the chiller model. The control module is configured to control the operation of the chiller based on the optimal control sequence. Specification 15 / 19 pages 19 CN 121804052 A

[0130] In some embodiments, the load prediction module 101 is specifically configured to: receive time characteristics, weather characteristics, and historical cooling load characteristics of the chiller through the XGBoost model, output a cooling load demand prediction curve through the XGBoost model, and predict the load of the chiller based on the prediction curve.

[0131] In some embodiments, the optimization module 103 further includes a chiller model trainer, configured to determine target feature parameters based on initial feature parameters related to the chiller, establish a chiller model in response to the artificial neural network receiving the target feature parameters, and output the energy consumption and performance coefficient of the chiller, wherein the initial feature parameters include at least one of the following: chiller operation list, chilled water flow rate, cooling water flow rate, chiller unit temperature point data, outdoor ambient temperature, total cooling load, and partial load factor.

[0132] In some embodiments, the chiller model trainer is specifically configured to determine the effective range of the initial feature parameters using statistical methods of covariance matrix and confidence ellipse, and determine the initial feature parameters within the effective range as target feature parameters.

[0133] In some embodiments, the optimization module 103 is specifically configured to: determine the energy consumption of the chiller unit corresponding to the control sequence within a specific time interval according to the chiller model; calculate the energy consumption of each control process in any control sequence according to the chiller model; and select the optimal chiller control sequence based on the comparison results of the energy consumption of multiple chiller units and the energy consumption of each control step in the control sequence controlling the operation of the chiller unit.

[0134] In some embodiments, the optimization module 103 is further configured to: establish a state transition equation to determine the control sequence and the cumulative energy consumption of the control sequence from the start time to time t; a control action in the control sequence determines the on or off state of the chiller at the next time moment, where t is a positive integer; and store the tuple formed by the control sequence and the cumulative energy consumption if it is not stored.

[0135] In some embodiments, the optimization module 103 is further configured to: search when the partial load rate meets the first condition.A control sequence that satisfies the cooling load and operating constraints, and when no control sequence is found, a control sequence that satisfies the cooling load and operating constraints is searched when the partial load rate satisfies the second condition; wherein, the numerical range of the second condition is greater than the numerical range of the first condition, and the numerical range of the first condition is the range where the partial load rate is close to the maximum value of the performance coefficient.

[0136] In some embodiments, the screening module 102 is specifically configured to generate multiple initial control sequences according to the cooling load demand prediction curve, the initial control sequences being limited by the minimum on and off intervals, and to determine the control sequence from the multiple initial control sequences based on the partial load rate constraints.

[0137] The following further illustrates the optimization method of the chiller provided in this disclosure in conjunction with experimental results.

[0138] For example, the Ruihong Tiandi Sun Palace in Shanghai can be used as the test site, and historical data may include building cooling load data, chiller unit temperature point data, chiller unit operating data, and the corresponding meteorological data of the building in 2023, sampled every 15 minutes. The training set for the load forecasting model and the chiller model comes from historical data from January 1, 2023 to October 15, 2023, excluding data from August 20 to August 27. The test set only includes data from August 20 to August 27.

[0139] Figure 9 shows the simulation results, which includes six subplots from top to bottom. The first subplot compares the cooling load value predicted by the load forecasting model with the actual value observed by the test device. The CVRMSE of the load forecasting model is 11.29%, indicating that the prediction accuracy of the load forecasting model is high. It should be noted that the predicted value is very close to the actual value on weekdays, indicating that the load forecasting model is very effective under normal operating conditions. However, the prediction accuracy decreases on weekends, which may be due to changes in the operation of the mall on weekends.

[0140] The second subplot in Figure 9 compares the actual power (i.e., the initial power), the ANN simulated power, and the optimized power (i.e., the optimal power on pages 16 / 19 of the specification, 20 CN 121804052 A). Among them, actual power refers to the true power under the original control sequence, ANN simulated power refers to the simulated power under the original control sequence, and optimized power refers to the power under the control of the optimal control sequence found using the ANN model, dynamic programming algorithm, and two-step optimization method. The simulated value of the ANN simulated power (red curve in the figure) and the actual value (purple curve in the figure) are very close, indicating that the ANN model effectively simulates the characteristics of the chiller. The optimized power (green curve in the figure) obtained using the ANN model, dynamic programming, and two-step optimization algorithm is lower than the actual value (purple curve in the figure), indicating that the algorithm found a control sequence with lower power, achieving effective energy saving. It should be noted that the ANN model occasionally overestimates energy consumption, for example, on August 21.

[0141] The third subplot in Figure 9 shows the change in the average daily energy saving rate throughout the test period. The average energy saving rate of 9.73% reflects the effectiveness of MPC in reducing energy consumption. Although the daily energy saving rate usually exceeds 10%, Sunday is a significant exception, when the power consumption of MPC exceeds that of the original sequence. Table 5 summarizes the accuracy of the load forecasting model and the chiller model, as well as the energy saving effect achieved by the MPC method during the test period.

[0142] Table 5

[0143] The fourth subplot in Figure 9 shows the change in ambient temperature. It can be seen from the figure that the ambient temperature does not change much on Sunday, but the cooling load prediction error on Sunday is relatively large, which may be related to the reduced energy saving effect.

[0144] It should be noted that the load forecasting model has a great influence on the control performance of MPC. Although the load forecasting model shows high accuracy on weekdays, its prediction accuracy decreases on weekends, especially Friday and Saturday. This may be due to the significant change in the internal load pattern on weekends, rather than the influence of external temperature. Possible solutions include increasing the size of the training dataset and introducing more time-related or lagging features.

[0145] The fifth subplot in Figure 9 shows the average COP of the chiller control sequences before and after optimization. The MPC algorithm always selects the chiller control sequences with a higher COP than the original configuration, thus optimizing energy efficiency. The sixth subplot shows a detailed example of the chiller control sequences before and after optimization on August 23. The MPC algorithm is adaptable to different operating conditions and cooling loads, and can select the chiller control sequence with the highest COP at any given time. For example, at noon, chillers 2 and 4 exhibit the highest COP under the environmental conditions and cooling requirements at that time, while chillers 2 and 3 are more efficient in the evening.

[0146] Figure 10 compares the time required to solve the same feasible solution using the ANN chiller model and the heuristic chiller model. The computational requirements of the ANN chiller model are 250 times that of the heuristic chiller model. Therefore, the computational requirements of the ANN chiller model will increase significantly as the number of feasible solutions increases.

[0147] Figures 11a and 11b show a comparison of the number of feasible schemes (i.e., feasible control sequences) before and after using the two-step optimization method. The results show that the number of feasible control sequences increases rapidly as the prediction range expands and the control interval decreases. Specifically, before optimization, when the control interval was set to 15 min and the prediction interval was set to 4 h, the number of feasible control sequences was approximately 6.76e+7. However, after using the two-step optimization method, the number of feasible control sequences decreased to 3.51e+5, a reduction of 99.47%. This indicates that the two-step optimization method not only significantly reduces the number of feasible control sequences but also accelerates the optimization process. The two-step optimization method first limits the PLR ​​to near the COP peak, and then, when no feasible control sequences are found...In the case of feasible control sequences, the range of PLR is broadened to ensure that feasible control sequences can be found. Therefore, the significant reduction in the number of feasible control sequences also greatly reduces the computational burden of the subsequent DP method.

[0148] Figures 12a and 12b compare the computation time required before and after using the DP method, assuming that the number of feasible control sequences is the same. The DP method avoids repeated computation of the ANN model by storing subproblems, thereby improving computational efficiency. Therefore, even if the number of feasible control sequences remains the same or increases, the computational efficiency will be improved. In the longest control range, the computation time using the DP method is reduced by 99.98%, from nearly 8 days to only 138.24 seconds.

[0149] Figure 13 shows the computation time corresponding to using the DP method alone, using the DP method combined with two-step optimization, and using the exhaustive search method when optimizing chiller sorting. To ensure that control performance is not affected, the computation time of each time step of MPC should not exceed a certain percentage of the control interval, which is set to 10% here. The red dashed line in the figure represents the maximum acceptable computation time for each time step, i.e., 90 seconds. The computation time of using the DP method alone is mostly less than 5 seconds, and the curve fluctuates little, indicating that the method performs well in handling load changes and is always within the acceptable range. The computation time of using the DP method combined with two-step optimization is slightly lower than that of using the DP method alone, indicating that the two-step optimization method can effectively reduce the computation time. In contrast, the computation time of the exhaustive search method is significantly higher, often exceeding the maximum acceptable time in most time steps.

[0150] Figure 14 shows the average and extreme values ​​of the computation time of using the DP method alone, using the DP method combined with two-step optimization, and using the exhaustive search method. The y-axis in Figure 14 is a logarithmic scale. The average computation time of using the DP method combined with two-step optimization is 3.61 seconds, and the maximum value is 6.81 seconds. The average computation time of using the DP method alone is close to that of using the method combined with two-step optimization, but the maximum time of using the DP method alone is 12.29 seconds. The average computation time of using the exhaustive search method is close to 165.26 minutes and fluctuates greatly. The maximum value of using the exhaustive search method is significantly higher, resulting in low computational efficiency in practical applications. Overall, the DP method greatly accelerates the optimization process, and the two-step optimization improves the stability of the DP method and avoids peak computation time. The method combining the DP method with two-step optimization has significant advantages in terms of computation time and stability, and is therefore suitable for practical control scenarios.

[0151] In summary, the method and control system for controlling the chiller in the chiller unit provided in this disclosure have the following beneficial effects:First, this disclosure only uses common chiller sensor points, such as the temperature and flow rate of chilled water and cooling water, as well as the chiller power, without the need for additional measuring equipment, ensuring that this disclosure can be deployed on a large scale.

[0152] Second, this disclosure adopts the MPC method, which can consider the dynamic impact of each step of the optimization strategy and ensure that the optimal control sequence is obtained.

[0153] Third, this disclosure adopts the ANN method combined with trusted regions, avoiding data interpolation and improving the accuracy of the chiller model.

[0154] Fourth, this disclosure adopts a method combining two-step optimization and dynamic programming, which greatly improves the computational efficiency and ensures that the optimal control sequence can be selected quickly, thereby ensuring the feasibility of actual deployment of this disclosure.

[0155] According to an embodiment of this disclosure, this disclosure also provides an electronic device, which includes: at least one processor; and a memory communicatively connected to at least one processor; wherein the memory stores instructions executable by at least one processor, which are executed by at least one processor to enable the chiller control method described in any of the above embodiments when executed by at least one processor.

[0156] According to embodiments of the present disclosure, the present disclosure also provides a readable storage medium storing computer instructions that enable a computer to implement the chiller control method described in any of the above embodiments when executed.

[0157] According to embodiments of the present disclosure, the present disclosure also provides a computer program product that, when executed by a processor, can implement the chiller control method described in any of the above embodiments.

[0158] It should be understood that, based on the teachings of the embodiments of the present disclosure and according to actual needs, the above process steps can be rearranged, or some additional steps can be added, or some of the steps already discussed can be deleted. Furthermore, depending on the actual situation, the steps described in this disclosure can be executed in parallel, sequentially, or in different orders, as long as the desired results of the technical solution disclosed in this disclosure can be achieved, and this is not limited herein.

[0159] The above specific embodiments do not constitute a limitation on the scope of protection of this disclosure. Various modifications can be made to the above embodiments according to design requirements and other factors, including mutual or substitution between features. Any modifications made within the scope of the teachings of this disclosure shall be included within the scope of protection of this disclosure. Specification 19 / 19 pages 23 CN 121804052 A Figure 1 Specification Figure 1 / 13 pages 24 CN 121804052 A Figure 2 Specification Figure 2 / 13 pages 25 CN 121804052 A Figure 3 Figure 4 Specification Figure 3 / 13 pages 26 CN121804052 A Figure 5 Figure 6a Appendix to the Instruction Manual Page 4 / 13 27 CN 121804052 A Figure 6b Figure 7 Appendix to the Instruction Manual Page 5 / 13 28 CN 121804052 A Figure 8 Appendix to the Instruction Manual Page 6 / 13 29 CN 121804052 A Figure 9 Appendix to the Instruction Manual Page 7 / 13 30 CN 121804052 A Figure 10 Appendix to the Instruction Manual Page 8 / 13 31 CN 121804052 A Figure 11a Appendix to the Instruction Manual Page 9 / 13 32 CN 121804052 A Figure 11b Appendix to the Instruction Manual Page 10 / 13 33 CN 121804052 A Figure 12a Appendix to the Instruction Manual Page 11 / 13 34 CN 121804052 A Figure 12b Appendix to the Instruction Manual Page 12 / 13 35 CN 121804052 A Figure 13 Figure 14 Figure 13 / 13 of the specification Page 36 CN 121804052 A Abstract The present disclosure provides a method of controlling a chiller unit, as well as a control system, apparatus, storage medium and program product for improving the energy efficiency of chiller units, and relates to energy management technology. The method for controlling a chiller consists of the steps of: predicting the chiller's load based on time characteristics, weather characteristics, and the chiller's historical cooling load characteristics; and determining a control sequence for the chiller based on the load prediction results and partial loadrate constraints, wherein the control sequence includes the number of to-be-chillers activated within the chiller unit and their operating sequence; selecting an optimal control sequence based on the chiller model; and controlling the chiller's operation based on the optimal control sequence.

Claims

1. A method for controlling a chiller in a chiller unit, comprising: Based on time characteristics, weather characteristics, and historical cooling load characteristics of the chiller, the load of the chiller is predicted; Based on the load forecast results and partial load factor constraints, a control sequence for controlling the chillers is determined, the control sequence including the number of chillers to be turned on and their operating order in the chiller unit; Based on the chiller model, the optimal control sequence is selected from the control sequence; and The operation of the chiller is controlled based on the optimal control sequence.

2. The method according to claim 1, wherein, The method of predicting the load of the chiller based on time characteristics, weather characteristics, and historical cooling load characteristics of the chiller includes: The time characteristics, weather characteristics, and historical cooling load characteristics of the chiller are received through the XGBoost model; The XGBoost model outputs a cooling load demand prediction curve; and Based on the predicted curve, the load of the chiller is predicted.

3. The method according to claim 1, wherein, The method further includes: The chiller model is established based on the historical operating data of the chiller.

4. The method according to claim 3, wherein, The step of establishing the chiller model based on the historical operating data of the chiller includes: Based on the initial characteristic parameters associated with the chiller, the target characteristic parameters are determined; and In response to the artificial neural network receiving the target feature parameters, the chiller model is established, and the energy consumption and coefficient of performance of the chiller are output. The initial characteristic parameters include at least one of the following: chiller unit operation list, chilled water flow rate, cooling water flow rate, chiller unit temperature point data, outdoor ambient temperature, total cooling load, and partial load rate.

5. The control method according to claim 4, wherein, The step of determining the target feature parameters based on the initial feature parameters includes: The effective range of the initial feature parameters is determined by statistical methods using covariance matrix and confidence ellipse, and the initial feature parameters that fall within the effective range are determined as the target feature parameters.

6. The control method according to claim 1, wherein, The step of selecting the optimal control sequence from the control sequence based on the chiller model includes: Based on the chiller model, the energy consumption of the chiller unit corresponding to the control sequence is determined within a specific time interval; Based on the chiller model, calculate the energy consumption of each control process in any of the control sequences; and Based on the comparison results of the energy consumption of multiple chiller units and the energy consumption of each control step in the control sequence that controls the operation of the chiller unit, the optimal chiller control sequence is selected.

7. The control method according to claim 6, wherein, Before calculating the energy consumption of each control process in any of the control sequences, the method further includes: Establish state transition equations to determine the control sequence and the cumulative energy consumption of the control sequence from the start time to time t, where a control action in the control sequence determines the on or off state of the chiller at the next time step, and t is a positive integer; and If the tuple formed by the control sequence and the accumulated energy consumption is not stored, the tuple is stored.

8. The control method according to claim 1 or 6, wherein, The step of selecting the optimal control sequence from the control sequence based on the chiller model includes: When the partial load rate meets the first condition, search for a control sequence that satisfies the cooling load and operating constraints; and If the control sequence is not found, a control sequence that satisfies the cooling load and operating constraints is searched when the partial load rate meets the second condition. The numerical range of the second condition is greater than that of the first condition, and the numerical range of the first condition is the range where the partial load rate is close to the maximum value of the performance coefficient.

9. The control method according to claim 2, wherein, The step of determining the control sequence for controlling the chiller based on load forecast results and partial load factor constraints includes: Based on the cooling load demand prediction curve, multiple initial control sequences are generated, each initial control sequence being limited by a minimum on / off interval; and The control sequence is determined from multiple initial control sequences based on partial load rate constraints.

10. A control system for controlling a chiller in a chiller unit, comprising: The load forecasting module is configured to predict the load of the chiller based on time characteristics, weather characteristics, and historical cooling load characteristics of the chiller. The filtering module is configured to determine a control sequence for controlling the chillers based on load forecast results and partial load factor constraints. The control sequence includes the number of chillers to be turned on and their operating order in the chiller unit. The optimization module is configured to select the optimal control sequence from the control sequence based on the chiller model; as well as The control module is configured to control the operation of the chiller based on the optimal control sequence.

11. The control system according to claim 10, wherein, The load forecasting module is specifically configured as follows: The XGBoost model receives the time characteristics, weather characteristics, and historical cooling load characteristics of the chiller, outputs a cooling load demand prediction curve, and predicts the load of the chiller based on the prediction curve.

12. The control system according to claim 11, wherein, The optimization module also includes a chiller model trainer, configured to determine target feature parameters based on initial feature parameters related to the chiller, and in response to the artificial neural network receiving the target feature parameters, to build the chiller model and output the chiller's energy consumption and coefficient of performance. The initial characteristic parameters include at least one of the following: chiller operation list, chilled water flow rate, cooling water flow rate, chiller unit temperature point data, outdoor ambient temperature, total cooling load, and partial load rate.

13. The control system according to claim 12, wherein, The chiller model trainer is specifically configured to use statistical methods of covariance matrix and confidence ellipse to determine the effective range of the initial feature parameters, and to determine the initial feature parameters that are within the effective range as the target feature parameters.

14. The control system according to claim 10, wherein, The optimization module is specifically configured to determine the energy consumption of the chiller unit corresponding to the control sequence within a specific time interval based on the chiller model, calculate the energy consumption of each control process in any control sequence based on the chiller model, and select the optimal chiller control sequence based on the comparison results of the energy consumption of multiple chiller units and the energy consumption of each control step in the control sequence that controls the operation of the chiller unit.

15. The control system according to claim 14, wherein, The optimization module is further configured to establish a state transition equation to determine a control sequence and the cumulative energy consumption of the control sequence from the start time to time t, wherein a control action in the control sequence determines the on or off state of the chiller at the next time moment, where t is a positive integer, and to store the tuple formed by the control sequence and the cumulative energy consumption if the tuple is not stored.

16. The control system according to claim 14, wherein, The optimization module is further configured to search for a control sequence that satisfies the cooling load and operating constraints when the partial load rate meets the first condition, and to search for a control sequence that satisfies the cooling load and operating constraints when the partial load rate meets the second condition if the control sequence is not found; wherein the numerical range of the second condition is greater than the numerical range of the first condition, and the numerical range of the first condition is the range in which the partial load rate is close to the maximum value of the performance coefficient.

17. The control system according to claim 10, wherein, The filtering module is specifically configured to generate multiple initial control sequences based on the cooling load demand prediction curve. The initial control sequences are subject to minimum on and off intervals and are determined from the multiple initial control sequences based on partial load rate constraints.

18. An electronic device comprising: At least one processor; as well as A memory communicatively connected to the at least one processor; wherein, The memory stores instructions that can be executed by the at least one processor to enable the at least one processor to perform the chiller control method according to any one of claims 1-9.

19. A non-transitory computer-readable storage medium storing computer instructions for causing the computer to perform the chiller control method according to any one of claims 1-9.

20. A computer program product comprising a computer program that, when executed by a processor, implements the steps of the chiller control method according to any one of claims 1-9.

Images