Two-stage optimization control method and device for air conditioning cold station under low load operation condition

Abstract

This invention discloses a two-stage optimized control method and apparatus for air conditioning chiller plants under low-load operation conditions, relating to the field of air conditioning chiller plant optimized control technology. The method includes: periodically determining whether to switch to a first-stage low-load operation control mode; if so, shutting down the target chiller with the smallest capacity and its associated equipment; in the first-stage mode, determining whether to switch to a second-stage low-load operation control mode based on the total cooling load demand and the network's cold storage capacity within a preset future cooling supply period; if so, storing cold in the network in the second-stage mode, shutting down all operating chillers and their associated equipment after cold storage is completed, and using a secondary chilled water pump to drive low-temperature chilled water in the network to supply cooling to the terminals; continuously monitoring the supply water temperature; if it remains above a preset temperature detection value for a preset monitoring period, starting the last shut-down chiller; through two-stage optimized control, the system's operating energy consumption under low-load conditions is effectively reduced.

Description

Technical Field

[0001] This invention relates to the field of air conditioning chiller plant optimization control technology, and in particular to a two-stage optimization control method and device for air conditioning chiller plants under low load operating conditions. Background Technology

[0002] Air conditioning chiller plants are among the most energy-intensive systems in the operation of large public buildings. Their equipment and subsystems are highly coupled, and their energy consumption typically exhibits a trade-off between different components. Optimizing and controlling air conditioning chiller plants through reasonable optimization methods to reduce system energy consumption while meeting end-load demands is a powerful entry point for achieving low-carbon and energy-saving practices in public buildings.

[0003] Most existing research methods only address routine daytime operating conditions, where the terminal cooling load is at a normal level and multiple chillers work together to provide cooling. They lack consideration for low-load conditions during off-peak hours and at night, when only one or two chillers are operational. This leads to reduced system efficiency under low-load conditions, impacting overall system energy efficiency. Furthermore, some existing technologies employ pipe network cold storage for optimized system control, but these rely solely on empirical rules to determine start-up and shutdown logic, lacking quantitative basis and scalability, or depend on precise modeling of the pipe network system, which is often difficult to implement in practical applications. Summary of the Invention

[0004] The purpose of this invention is to provide a two-stage optimized control method and device for air conditioning chiller plants under low-load operating conditions. It does not rely on precise modeling of the pipeline network, can identify low-load operating conditions in a timely manner, avoids the chiller from running continuously in the inefficient range, and effectively reduces the system operating energy consumption under low-load conditions.

[0005] In a first aspect, the present invention provides a two-stage optimized control method for an air conditioning chiller plant under low-load operation conditions, comprising: periodically determining whether the operating conditions of the chiller plant meet preset conditions based on the number of chillers in operation and the load rate of the chillers; if so, switching to a preset first-stage low-load operation control mode and shutting down the target chiller and its supporting equipment; wherein the target chiller is the chiller with the smallest capacity currently in operation; under the first-stage low-load operation control mode, determining whether to switch to a preset second-stage low-load operation control mode based on the total cooling load demand within a preset future self-cooling time period and the cold storage capacity of the pipeline network; if so, under the second-stage low-load operation control mode, storing cold in the pipeline network, and after completing the cold storage, shutting down all operating chillers and their supporting equipment, so that the low-temperature chilled water in the pipeline network is driven by the chilled water secondary pump to supply cooling to the terminal; continuously monitoring the water supply temperature in the pipeline network, and if the water supply temperature continues to be greater than the preset temperature detection value within a preset monitoring time period, starting the last shut-down chiller and its corresponding supporting equipment.

[0006] In some preferred embodiments of the present invention, the self-cooling period is 1 hour; determining whether to switch to the preset two-stage low-load operation control mode based on the total cooling load demand within the preset future self-cooling period and the cold storage capacity of the pipeline network includes: predicting the total cooling load demand within the next 1 hour based on a pre-established ultra-short-term building load prediction model, and determining the cold storage capacity of the pipeline network based on the parameters of the pipeline network; if the total cooling load demand within the next 1 hour is less than the cold storage capacity of the pipeline network, then determining to switch to the two-stage low-load operation control mode; otherwise, maintaining the one-stage low-load operation control mode.

[0007] In some preferred embodiments of the present invention, predicting the total cooling load demand for the next hour based on a pre-established ultra-short-term building load prediction model includes: using the hourly cooling load of the past 24 hours and the cooling load of every 5 minutes in the past hour as inputs to the model, and using the cooling load value of the next 5 minutes as the predicted output value of the model; using a rolling prediction method, the previous predicted value is used as the input variable for the new round, and the calculation is repeated until the total cooling load demand for the self-supply period is obtained; wherein, the total cooling load demand is determined by the following formula: ;in, This represents the total cooling load demand for the next hour. This represents the cooling load for the i-th 5-minute interval within the next hour.

[0008] In some preferred embodiments of the present invention, the cold storage capacity of the pipeline network is determined by the following formula: ;in, To enhance the cold storage capacity of the pipeline network; This is the preset temperature rise of the cold network; Cold network pipe section The diameter, Cold network pipe section Length, The density of water, The specific heat capacity of water at constant pressure. Cold network pipe section The capacity factor.

[0009] In some preferred embodiments of the present invention, the cold storage of the pipeline network includes: calculating the chilled water outlet temperature setpoint of the chiller through a reinforcement learning controller, and controlling the operation of the chiller according to the chilled water outlet temperature setpoint to achieve cold storage; wherein, the reinforcement learning controller is established using a deep Q-learning algorithm, including a Q-value network, a target network, and an experience playback pool; both the Q-value network and the target network are constructed using deep neural networks with consistent structures, with the input being a state variable vector and the output being a Q-value vector for each possible action.

[0010] In some preferred embodiments of the present invention, determining the setpoint for the chilled water outlet temperature includes: inputting the current state variable into a Q-value network to obtain the Q-value of each possible action, and taking the action corresponding to the maximum Q-value as the optimal setpoint for the chilled water outlet temperature, as shown in the following formula: ; ; ;in, Set the chilled water outlet temperature; For the i-th action Q value; Let k be the state variable in round k. For state variables; The chilled water supply temperature for the pipeline network; This refers to the return temperature of the chilled water in the pipe network. This is the difference between the cold storage capacity of the pipeline network and the total cooling load demand in the future period. This represents the current operating power of the chiller.

[0011] In some preferred embodiments of the present invention, the conditions for determining the completion of cold storage include: the actual supply temperature of chilled water reaches the set value of the chiller's chilled water supply temperature, and the actual supply and return temperature difference of chilled water is less than a preset threshold.

[0012] In some preferred embodiments of the present invention, the method further includes: updating the parameters of the reinforcement learning controller by the following steps: storing the current round's state variable, action variable, reward function value, and the new round's state variable in the form of a quadruple in the experience replay pool; wherein the reward function value is determined by the following formula: Where R is the reward function value; The time required for the cold network to complete cold storage. The preset weighting coefficients, During the cold storage period The chiller's operating power at any given time; The time interval depends on the sensor's acquisition frequency; every preset number of rounds, a preset number of samples are randomly selected from the experience playback pool, and the parameters of the Q-value network are updated using the following formula: ; ;in, The loss function; For preset quantity; For Q-value networks, the current state and actions The predicted Q value; Discount factor; For the target network to determine the next round state The maximum Q value that can be output, For learning rate, The loss function with respect to network parameters The gradient; whenever the Q-value network completes an update a preset number of times, the parameters of the target network are updated to the current parameters of the Q-value network.

[0013] In some preferred embodiments of the present invention, in the two-stage low-load operation control mode, after the steps of storing cold in the pipeline network and shutting down all operating chillers and their supporting equipment after the cold storage is completed, the method further includes: if the duration for which the water supply temperature is greater than the preset temperature detection value does not reach the preset monitoring time period, the two-stage low-load operation control mode is maintained until the normal operation stage of the next day is reached, and then the control mode under normal operation conditions is restored.

[0014] Secondly, the present invention provides a two-stage optimization control device for an air conditioning chiller plant under low-load operation conditions, comprising: a first-stage control module, used to periodically determine whether the operating conditions of the chiller plant meet preset conditions based on the number of chillers in operation and the load rate of the chillers; if so, switching to a preset first-stage low-load operation control mode and shutting down the target chiller and its supporting equipment; wherein the target chiller is the chiller with the smallest capacity currently in operation; and a cooling data processing module, used to, under the first-stage low-load operation control mode, calculate the total cooling load within a preset future cooling time period. The system determines whether to switch to the preset two-stage low-load operation control mode based on demand and the cold storage capacity of the pipeline network. The two-stage control module is used to store cold in the pipeline network under the two-stage low-load operation control mode if the condition is met. After the cold storage is completed, all operating chillers and their supporting equipment are shut down, so that the low-temperature chilled water in the pipeline network is driven by the chilled water secondary pump to supply cooling to the terminal. The cooling data monitoring module is used to continuously monitor the water supply temperature in the pipeline network. If the water supply temperature is continuously higher than the preset temperature detection value within the preset monitoring period, the last shut-down chiller and its corresponding supporting equipment are started.

[0015] This invention brings the following beneficial effects: This invention provides a two-stage optimized control method and apparatus for an air conditioning chiller plant under low-load operation conditions. The method includes: periodically determining whether the operating conditions of the chiller plant meet preset conditions based on the number of chillers in operation and the chiller load rate; if so, switching to a preset first-stage low-load operation control mode and shutting down the target chiller and its supporting equipment; wherein the target chiller is the chiller with the smallest capacity currently in operation; under the first-stage low-load operation control mode, determining whether to switch to a preset second-stage low-load operation mode based on the total cooling load demand during a preset future self-cooling period and the cold storage capacity of the pipeline network. Control mode: If so, in the two-stage low-load operation control mode, the pipeline network is used for cold storage. After the cold storage is completed, all operating chillers and their supporting equipment are shut down, and the low-temperature chilled water in the pipeline network is driven by the chilled water secondary pump to supply cooling to the terminal. The supply water temperature in the pipeline network is continuously monitored. If the supply water temperature is continuously higher than the preset temperature detection value within the preset monitoring period, the last shut-down chiller and its corresponding supporting equipment are started. Through two-stage optimized control, low-load conditions can be identified in a timely manner, avoiding continuous operation of chillers in the inefficient range and effectively reducing the system's operating energy consumption under low-load conditions. Attached Figure Description

[0016] To more clearly illustrate the specific embodiments of the present invention or the technical solutions in the prior art, the drawings used in the description of the specific embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are some embodiments of the present invention. For those skilled in the art, other drawings can be obtained from these drawings without creative effort.

[0017] Figure 1 A schematic diagram of an air conditioning chiller system provided in an embodiment of the present invention; Figure 2 A flowchart of a two-stage optimization control method for an air conditioning chiller under low-load operating conditions provided by an embodiment of the present invention; Figure 3 For reference Figure 3 The diagram shown is a comparison of daily operating energy consumption of different methods provided by embodiments of the present invention under low-load operating conditions of the system during a certain week. Figure 4 This is a schematic diagram of the structure of a two-stage optimization control device for an air conditioning chiller plant under low-load operating conditions, provided in an embodiment of the present invention. Figure 5 This is a schematic diagram of the structure of an electronic device provided in an embodiment of the present invention.

[0018] Icons: 310 - Phase 1 control module; 320 - Cooling data processing module; 330 - Phase 2 control module; 340 - Cooling data monitoring module; 400 - Memory; 401 - Processor; 402 - Bus; 403 - Communication interface. Detailed Implementation

[0019] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. The components of the embodiments of the present invention described and shown in the accompanying drawings can generally be arranged and designed in various different configurations.

[0020] Therefore, the following detailed description of the embodiments of the invention provided in the accompanying drawings is not intended to limit the scope of the claimed invention, but merely to illustrate selected embodiments of the invention. All other embodiments obtained by those skilled in the art based on the embodiments of the invention without inventive effort are within the scope of protection of the invention.

[0021] It should be noted that similar labels and letters in the following figures indicate similar items. Therefore, once an item is defined in one figure, it does not need to be further defined and explained in subsequent figures.

[0022] In the description of this invention, it should be noted that the terms "center," "upper," "lower," "left," "right," "vertical," "horizontal," "inner," and "outer," etc., indicate the orientation or positional relationship based on the orientation or positional relationship shown in the accompanying drawings, or the orientation or positional relationship commonly used when the product of this invention is in use. They are only for the convenience of describing this invention and simplifying the description, and do not indicate or imply that the device or element referred to must have a specific orientation, or be constructed and operated in a specific orientation, and therefore should not be construed as a limitation of this invention. In addition, the terms "first," "second," "third," etc., are only used to distinguish descriptions and should not be construed as indicating or implying relative importance.

[0023] Furthermore, terms such as "horizontal," "vertical," and "sag" do not imply that components must be absolutely horizontal or suspended, but rather that they can be slightly tilted. For example, "horizontal" simply means that its direction is more horizontal relative to "vertical," and does not mean that the structure must be completely horizontal, but can be slightly tilted.

[0024] In the description of this invention, it should also be noted that, unless otherwise explicitly specified and limited, the terms "set," "install," "connect," and "link" should be interpreted broadly. For example, they can refer to a fixed connection, a detachable connection, or an integral connection; they can refer to a mechanical connection or an electrical connection; they can refer to a direct connection or an indirect connection through an intermediate medium; and they can refer to the internal connection of two components. Those skilled in the art can understand the specific meaning of the above terms in this invention based on the specific circumstances.

[0025] The following detailed description of some embodiments of the present invention is provided in conjunction with the accompanying drawings. Unless otherwise specified, the following embodiments and features can be combined with each other.

[0026] This invention provides a two-stage optimized control method for air conditioning chiller plants under low-load operating conditions, applicable to the controller of an air conditioning chiller plant system. (See also...) Figure 1 The schematic diagram shown in this embodiment of the invention illustrates an air conditioning chiller system. The system comprises five chiller units, five primary chilled water pumps, three secondary chilled water pumps, five cooling water pumps, and five cooling towers. The chiller units have a rated cooling capacity of 1600kW and a rated COP of 5.71; the primary chilled water pumps have a rated power of 27kW and a rated flow rate of 288m³ / h. 3 / h; the rated power of the chilled water secondary pump is 55kW, and the rated flow rate is 456m³ / h. 3 / h; The rated power of the cooling water pump is 40kW and the rated flow rate is 324m³ / h. 3 / h; the rated heat exchange capacity of the cooling tower is 2194kW, the rated power of the fan is 22kW, and the rated flow rate of the fan is 4200m³ / h. 3 / min. The primary chilled water loop is used to generate low-temperature chilled water, and the secondary chilled water loop is used to deliver the generated chilled water to each load terminal. The three secondary chilled water pumps are always kept on during system operation.

[0027] See Figure 2 The flowchart shown in this embodiment of the invention provides a two-stage optimization control method for an air conditioning chiller plant under low-load operating conditions. The method includes: Step S102: Periodically determine whether the operating conditions of the chiller station meet the preset conditions based on the number of chillers turned on and the load rate of the chillers. If so, switch to the preset first-stage low-load operation control mode and shut down the target chiller and its supporting equipment. The target chiller is the chiller with the smallest capacity currently in operation.

[0028] Specifically, in this embodiment, the controller performs a judgment every 5 minutes. The preset condition is that the number of operating chillers is less than or equal to two and the load rate of each chiller is less than the preset lower limit of the high-efficiency operating range. The lower limit of the high-efficiency operating range can be set to 70% of the rated capacity. As an optional implementation, the lower limit of the high-efficiency operating range can also be set to other values ​​based on the actual performance curve of the chiller unit, such as 60%, 65%, 75%, or 80% of the rated capacity. As another optional implementation, the preset condition can also be based solely on the number of operating chillers being less than or equal to two without judging the load rate, or based on the ratio of the total cooling capacity of the chillers to the total load of the terminals being lower than a certain preset threshold. The supporting equipment of the target chiller includes a chilled water primary pump, a cooling water pump, and a cooling tower. If the number of operating chillers is greater than two, or if the load rate of any chiller is greater than or equal to the lower limit of the high-efficiency operating range, the preset condition is not met, and the control mode under the current normal operating condition should be maintained. In this embodiment, the control mode under normal operating conditions is set as a rule-based control mode based on cooling load measurement. That is, the total cooling load at the terminal is calculated based on the real-time supply and return water temperature difference and flow rate. Then, the number of chiller units to be started is determined based on the magnitude of the total cooling load. Simultaneously, each chiller unit is interlocked with its corresponding chilled water primary pump, cooling water pump, and cooling tower for start-stop operation. As another optional implementation, the control mode under normal operating conditions can also be set as an optimization-based control mode, such as using differential evolution algorithms or genetic algorithms to optimize chiller load distribution. It should be noted that any feasible control mode can be set in practice; this embodiment is merely an illustrative example. By periodically judging based on the number of chillers started and the load rate and automatically switching to a low-load operation control mode, low-load conditions can be identified in a timely manner, preventing chillers from continuously operating in inefficient ranges and effectively reducing system energy consumption.

[0029] Step S104: In the first-stage low-load operation control mode, determine whether to switch to the preset second-stage low-load operation control mode based on the total cooling load demand during the preset future self-cooling time period and the cold storage capacity of the pipeline network.

[0030] Specifically, the self-cooling time period can be set to 1 hour. As an optional implementation, the self-cooling time period can also be set to 0.5 hours, 2 hours, or 3 hours depending on the actual building thermal inertia and pipeline network scale. In this embodiment, 1 hour is preferred. The decision to switch to the preset two-stage low-load operation control mode is based on the total cooling load demand within the preset future self-cooling time period and the pipeline network's cold storage capacity. This specifically includes: firstly, predicting the total cooling load demand (Cooling Load, CL) for the next hour based on a pre-established ultra-short-term building load prediction model, and simultaneously determining the pipeline network's cold storage capacity (Cap) based on pipeline network parameters. Then, the two are compared: if the total cooling load demand for the next hour is less than the pipeline network's cold storage capacity, then it is determined to switch to the two-stage low-load operation control mode, proceeding to step S106; otherwise, the first-stage low-load operation control mode is maintained, and the remaining single chiller continues to provide cooling load to the terminal. This quantitative comparison method provides a precise decision-making basis for whether to enter the two-stage cold storage mode, avoiding reliance solely on empirical rules for switching decisions, and improving the reliability and scalability of the control strategy.

[0031] Furthermore, in some preferred embodiments of the present invention, the self-cooling period is 1 hour; determining whether to switch to the preset two-stage low-load operation control mode based on the total cooling load demand within the preset future self-cooling period and the cold storage capacity of the pipeline network includes: predicting the total cooling load demand within the next 1 hour based on a pre-established ultra-short-term building load prediction model, and determining the cold storage capacity of the pipeline network based on the parameters of the pipeline network; if the total cooling load demand within the next 1 hour is less than the cold storage capacity of the pipeline network, then determining to switch to the two-stage low-load operation control mode; otherwise, maintaining the one-stage low-load operation control mode.

[0032] Specifically, based on a pre-established ultra-short-term building load forecasting model, the total cooling load demand for the next hour is predicted. A Long Short-Term Memory (LSTM) recurrent neural network is used to establish the ultra-short-term building load forecasting model. The model network parameters are listed in Table 1 below. The input variables are the hourly cooling load of the past 24 hours and the cooling load every five minutes of the past hour; the output variable is the cooling load value for the next five minutes. The training data consists of the hourly and five-minute cooling load values ​​from the past year. The root mean square error is selected as the loss function, and the Adam algorithm is selected as the optimizer. The training iterations are 800, and the sample size per iteration is 64.

[0033] Table 1

[0034] When forecasting total cooling load demand for the next hour, the hourly cooling load of the past 24 hours is used as a reference. and the cooling load value every five minutes in the past hour Input the data into the model to calculate the predicted cooling load for the next 5 minutes. Then, the oldest five-minute cooling load value is removed, and this predicted value is used as a new input variable into the model to calculate the latest cooling load forecast. For example, after the forecast is obtained... Then, replace the original model input with it. This allows for the calculation of the predicted cooling load for the next 10 minutes. Repeat this rolling forecasting step until you obtain the cooling load forecast every five minutes for the next hour. By employing a long short-term memory recurrent neural network and a rolling forecasting method, the time-series characteristics of building cooling load can be accurately captured, enabling high-precision prediction of cooling load every 5 minutes for the next hour.

[0035] The cold storage capacity of the pipeline network is determined based on its parameters, specifically by calculating the cold storage capacity Cap using the following formula: ; in, To enhance the cold storage capacity of the pipeline network; The preset temperature rise of the cold network is set to 4 K in this embodiment; Cold network pipe section The diameter, Cold network pipe section Length, The density of water, The specific heat capacity of water at constant pressure. Cold network pipe section The capacity factor. The system pipeline parameters for this embodiment are shown in Table 2 below: Table 2

[0036] As an optional implementation method, the allowable temperature rise of the cooling network is... T can also be set to 3K, 5K, or 6K according to actual system operation requirements and end-point comfort requirements. By accurately calculating the cold storage capacity based on the pipeline network's geometric and physical properties, the total cold capacity that the pipeline network can store under the current allowable temperature rise can be accurately assessed, providing an accurate quantitative basis for whether to enter the second-stage cold storage mode.

[0037] Furthermore, in some preferred embodiments of the present invention, predicting the total cooling load demand for the next hour based on a pre-established ultra-short-term building load prediction model includes: using the hourly cooling load of the past 24 hours and the cooling load of every 5 minutes in the past hour as inputs to the model, and using the cooling load value of the next 5 minutes as the predicted output value of the model; through a rolling prediction method, using the previous predicted value as a new round of input variables, and repeating the calculation until the total cooling load demand for the self-supply period is obtained; wherein, the total cooling load demand is determined by the following formula: ;in, This represents the total cooling load demand for the next hour. This represents the cooling load for the i-th 5-minute interval within the next hour.

[0038] Specifically, the cooling load forecast is obtained every five minutes for the next hour based on rolling forecasts. Then, the total cooling load demand for the next hour can be obtained by summing the results. The specific formula is as follows: ; in, This represents the total cooling load demand for the next hour. This represents the cooling load for the i-th 5-minute interval within the next hour. As an alternative implementation, if the model's prediction time step differs, for example, if predictions are made every 10 minutes or every 15 minutes, the upper limit of the summation and the value of 5 minutes are adjusted accordingly to match the total length of the self-supply cooling period. Using the above formula, the predicted cooling load for each 5-minute interval can be accurately converted into the total cooling demand for the next hour, providing a unified dimension for subsequent comparisons with the network's cooling storage capacity.

[0039] Furthermore, in some preferred embodiments of the present invention, the cold storage capacity of the pipeline network is determined by the following formula: ;in, To enhance the cold storage capacity of the pipeline network; This is the preset temperature rise of the cold network; Cold network pipe section The diameter, Cold network pipe section Length, The density of water, The specific heat capacity of water at constant pressure. Cold network pipe section The capacity factor.

[0040] Specifically, this formula calculates the total cooling capacity that the network can absorb and store when the current temperature rises to the maximum allowable temperature by summing the volumes of each pipe segment in the network and multiplying them by the density of water, specific heat capacity at constant pressure, allowable temperature rise, and capacity factor. The capacity factor is used to calculate this total cooling capacity. This is used to correct the difference between the actual usable cold storage capacity and the theoretical volumetric capacity of each pipe section due to factors such as insulation conditions and heat loss. As an optional implementation method, if the insulation conditions of each pipe section in the network are good and the heat loss is small, the capacity coefficient of each pipe section is [not specified]. All values ​​can be set to 1.0. As another optional implementation, different capacity coefficients can be set for pipe sections with different insulation levels. For example, the secondary loop of an air conditioning chiller station, being located indoors and with good insulation, A value of 0.95 can be used; for long-distance pipelines transporting goods to the terminal area, a value of 0.85 to 0.9 can be used depending on the actual situation. By introducing this formula and the capacity coefficient, the actual cold storage potential of the pipeline network can be reflected more accurately, providing a reliable guarantee for comparing cold load demand with cold storage capacity.

[0041] Step S106: If so, in the second-stage low-load operation control mode, the pipeline network is used for cold storage, and after the cold storage is completed, all operating chillers and their supporting equipment are shut down, so that the low-temperature chilled water in the pipeline network is driven by the chilled water secondary pump to supply cooling to the terminal.

[0042] Specifically, upon entering the second-stage low-load operation control mode, the controller first performs a chilled water storage operation on the pipe network. During this process, the controller dynamically calculates and outputs the chilled water outlet temperature setpoint for the chiller through reinforcement learning. The chiller operates at this target temperature, injecting low-temperature chilled water into the pipe network until it is fully chilled. After chilled water storage is complete, the controller issues a command to shut down the last operating chiller and its corresponding auxiliary equipment (including the primary chilled water pump, cooling water pump, and cooling tower). At this point, the chiller completely stops operating, and the chilled water that has accumulated in the pipe network is circulated by the always-on secondary chilled water pump, providing the required cooling load to each terminal area. Through the above control logic, in the second-stage mode, the operation mode completely replaces the direct cooling supply of the chiller by utilizing pipe network chilled water storage, minimizing the energy consumption of the chiller under extremely low load conditions.

[0043] Furthermore, in some preferred embodiments of the present invention, the cold storage of the pipeline network includes: calculating the chilled water outlet temperature setpoint of the chiller through a reinforcement learning controller, and controlling the operation of the chiller according to the chilled water outlet temperature setpoint to achieve cold storage; wherein, the reinforcement learning controller is established using a deep Q-learning algorithm, including a Q-value network, a target network, and an experience playback pool; both the Q-value network and the target network are constructed using deep neural networks with consistent structures, the input being a state variable vector, and the output being a Q-value vector for each possible action.

[0044] Specifically, the reinforcement learning controller is built using the Deep Q-Network (DQN) algorithm. This controller consists of three core components: a Q-value network, a target network, and an experience replay pool. Both the Q-value network and the target network are constructed using deep neural networks, and their network structures are completely identical. The input variables are state variable vectors, and the output variables are Q-value vectors for each possible action. The network parameter settings are shown in Table 3 below. The experience replay pool stores a quadruple consisting of the current round's state variable, action variable, reward function value, and the state variable for the next round. By using a random sampling mechanism, the temporal correlation between samples is broken, thereby improving training stability.

[0045] Table 3

[0046] As an optional implementation, the reinforcement learning controller can also be built using other reinforcement learning algorithms such as Deep Deterministic Policy Gradient (DDPG) or Proximal Policy Optimization (PPO). When using the DDPG algorithm, the action space can be continuous, for example, the chilled water outlet temperature setpoint can be continuously adjustable between 5.0℃ and 12.0℃. In this case, the Q-value network is replaced by an Actor-Critic network structure. This embodiment uses the deep Q-learning algorithm as an example, as its discrete action space is easier to train and converge. By dynamically optimizing the outlet temperature setpoint through the reinforcement learning controller, the chilled water storage process can adapt online to different loads and pipeline conditions, continuously optimizing the chilled water storage strategy.

[0047] Furthermore, in some preferred embodiments of the present invention, determining the setpoint for the chilled water outlet temperature includes: inputting the current state variable into a Q-value network to obtain the Q-value of each possible action, and taking the action corresponding to the maximum Q-value as the optimal setpoint for the chilled water outlet temperature, as shown in the following formula: ; ; ;in, Set the chilled water outlet temperature; For the i-th action Q value; Let k be the state variable in round k. For state variables; The chilled water supply temperature for the pipeline network; This refers to the return temperature of the chilled water in the pipe network. This is the difference between the cold storage capacity of the pipeline network and the total cooling load demand in the future period. This represents the current operating power of the chiller.

[0048] Specifically, the state variable S is defined as follows: ; in, Set the chilled water outlet temperature; This refers to the return temperature of the chilled water in the pipe network. To enhance the cold storage capacity of the pipeline network Total cooling load demand in the future period The difference; The current operating power of the chiller is represented by these four state variables. These four state variables comprehensively reflect the current thermal state of the pipeline network, the balance between cold storage supply and demand, and the operating status of the chiller, enabling the reinforcement learning agent to fully perceive the system state and make reasonable decisions.

[0049] Action variable A is defined as follows: ; in, This is the setpoint for the chilled water outlet temperature. In this embodiment, the action space is a discrete value. The selectable values ​​are as follows: ; There are a total of 10 possible actions, each corresponding to a different outlet water temperature setpoint. As an optional implementation, the action space can also be set to more finely discrete values, such as values ​​between 5.0℃ and 12.0℃ in 0.5℃ intervals, in which case the number of output layer nodes increases to 15. Alternatively, the action space can also be adjusted according to the actual chiller unit's minimum outlet water temperature capability and end-point comfort requirements.

[0050] Optimal setpoint for chilled water outlet temperature The specific calculation method is as follows: In the k-th round of decision-making, the state variables of this round are... The input is fed into a Q-value network, where the possible actions are calculated via forward propagation by a deep neural network. The corresponding Q value Then, the maximum value among all Q values ​​is taken. The action corresponding to this maximum value is the optimal chilled water outlet temperature setpoint for the current state, as shown in the following formula: ; in, Set the chilled water outlet temperature; For the i-th action Q value; Let be the state variable for the k-th round. Using the above method, the controller can select the optimal outlet water temperature setpoint from a predefined action space based on the current system state at each decision moment, achieving dynamic online optimization of the outlet water temperature during the cold storage process. By defining a multi-dimensional state including supply and return water temperatures, the cold storage supply-demand difference, and chiller power, as well as a discretized outlet water temperature setpoint action space, the reinforcement learning agent can accurately perceive the system state and balance the cold storage speed with energy-saving effects.

[0051] Furthermore, in some preferred embodiments of the present invention, the conditions for determining the completion of cold storage include: the actual supply temperature of chilled water reaches the set value of the chiller's chilled water supply temperature, and the actual supply and return temperature difference of chilled water is less than a preset threshold.

[0052] Specifically, the criteria for completing chilled water storage are described as follows: continuously monitoring the actual chilled water supply temperature during the chilled water storage process. and actual return water temperature The cold storage process is considered complete when both of the following conditions are met simultaneously: First, the actual supply temperature of the chilled water... The chiller water supply temperature setpoint has been reached. ,Right now Second, the actual temperature difference between the supply and return chilled water. Less than the preset threshold ,Right now In this embodiment, Set to 5.5℃. Set to 2.5K. As an optional implementation, the threshold... It can also be set to 2.0K, 3.0K, or 3.5K depending on the network size and cold storage accuracy requirements. By judging the dual conditions, it can be ensured that not only has the outlet water temperature dropped to the target value, but the terminal load has also been basically met (the supply and return water temperature difference is small enough), and the low-temperature chilled water in the network has been fully accumulated, thereby avoiding insufficient cooling capacity due to premature cessation of cold storage, or excessive cold storage that wastes the energy consumption of the chiller operation.

[0053] Furthermore, in some preferred embodiments of the present invention, the method further includes: updating the parameters of the reinforcement learning controller by the following steps: storing the current round's state variable, action variable, reward function value, and the new round's state variable in the form of a quadruple in the experience replay pool; wherein, the reward function value is determined by the following formula: Where R is the reward function value; The time required for the cold network to complete cold storage. The preset weighting coefficients, During the cold storage period The chiller's operating power at any given time; The time interval is defined as follows: every preset number of rounds, a preset number of samples are randomly selected from the experience replay pool, and the parameters of the Q-value network are updated using the following formula: ; ;in, The loss function; For preset quantity; For Q-value networks, the current state and actions The predicted Q value; Discount factor; For the target network to determine the next round state The maximum Q value that can be output, For learning rate, The loss function with respect to network parameters The gradient; whenever the Q-value network completes an update a preset number of times, the parameters of the target network are updated to the current parameters of the Q-value network.

[0054] Specifically, after each round of cold storage is completed, the controller will collect the state information, action information, and reward information from that round of cold storage to update the parameters of the reinforcement learning controller. First, the state variables of that round... Actual action variables The calculated reward function value and the new round of state variables reached after the action is performed. With quadruplets The data is stored in the experience replay pool in the form of 200 records. The capacity of the experience replay pool is set to 200 records. If the experience replay pool is full, the oldest set of data is replaced according to the first-in, first-out principle.

[0055] The reward function value R is determined by the following formula: ; Where R is the reward function value; The time required for the cold network to complete cold storage. The preset weighting coefficients, During the cold storage period The chiller's operating power at any given time; The time interval is defined. The reward function is designed to consider both the time efficiency and energy efficiency of the cold storage process: the first term... Encourage the controller to complete the cold storage as soon as possible, the second item The cumulative energy consumption during the cold storage process is penalized. A weighting coefficient ω is used to balance the relative importance of the two factors; the larger the value of ω, the heavier the penalty for energy consumption. This reward function guides the reinforcement learning controller to learn a fast and energy-efficient outlet water temperature setting strategy.

[0056] Every preset number of rounds (every hundred rounds in this embodiment), a preset number of samples (N is set to 32 in this embodiment) are randomly selected from the experience replay pool, and the parameters θ of the Q-value network are updated using the following formula: ; ; in, The loss function; The preset quantity is set to 32 in this embodiment; For Q-value networks, the current state and actions The predicted Q value; The discount factor is set to 0.98 in this embodiment; For the target network to determine the next round state The maximum Q value that can be output, The learning rate is set to 0.0001 in this embodiment; The loss function with respect to network parameters The gradient is calculated; after the Q-value network completes a preset number of updates, the parameters of the target network are updated to the current parameters of the Q-value network. This loss function is based on temporal difference error and minimizes the mean squared error between the predicted Q-value of the Q-value network and the target Q-value (obtained by adding the maximum Q-value discount of the next state to the reward).

[0057] After the Q-value network completes a preset number of updates (five updates in this embodiment), the parameters of the target network are... Update to the current parameters of the Q-value network, i.e. By periodically synchronizing the parameters of the target network with the parameters of the Q-value network, training instability caused by frequent changes in the target value during training is avoided. As an optional implementation, the preset number of rounds can be set to update once every fifty or two hundred rounds; the preset number of samples N can be set to 64 or 128. As another optional implementation, the discount factor γ can be set to 0.95 or 0.99, and the learning rate α can be set to 0.0005 or 0.002. By employing empirical replay, the target network, and a loss function based on temporal difference error, the deep Q-learning network can be trained stably and efficiently, enabling the reinforcement learning controller to gradually converge to the optimal outlet water temperature setting strategy, achieving comprehensive optimization of energy consumption and time during the cooling process.

[0058] Step S108: Continuously monitor the water supply temperature in the pipeline network. If the water supply temperature continues to exceed the preset temperature detection value within the preset monitoring time period, start the last shut-down chiller and corresponding supporting equipment.

[0059] Specifically, during the process of supplying cooling to the terminal units by the low-temperature chilled water stored in the pipeline network driven by the secondary chilled water pump, the controller continuously monitors the chilled water supply temperature in the pipeline network every 5 minutes. When the detected water supply temperature consistently exceeds the preset temperature detection value within a preset monitoring time period, it indicates that the stored cooling capacity in the pipe network has been largely depleted and is insufficient to maintain the cooling quality at the terminal. At this point, the controller issues a command to restart the most recently shut-down chiller and its corresponding auxiliary equipment (including the chilled water primary pump, cooling water pump, and cooling tower), restoring the chiller's cooling capacity. In this embodiment, the preset monitoring time period is set to 15 minutes, and the preset temperature detection value is set to 10°C. As an optional implementation, the preset temperature detection value can also be set to 9°C, 11°C, or 12°C; the preset monitoring time period can also be set to 10 minutes or 20 minutes. By setting both temperature and duration as dual judgment conditions, false restarts caused by instantaneous temperature fluctuations or measurement noise can be effectively avoided, ensuring that the system only resumes chiller operation when the stored cooling capacity in the pipe network is truly depleted, thus guaranteeing the stability of system operation and the reliability of terminal cooling.

[0060] Furthermore, in some preferred embodiments of the present invention, after the steps of storing cold in the pipeline network and shutting down all operating chillers and their supporting equipment after the cold storage is completed in the two-stage low-load operation control mode, the method further includes: if the duration for which the water supply temperature is greater than the preset temperature detection value does not reach the preset monitoring time period, the two-stage low-load operation control mode is maintained until the normal operation stage of the next day is reached, at which point the control mode under normal operation conditions is restored.

[0061] Specifically, during the process of supplying cooling to the terminal using the cold storage capacity of the pipe network, if the monitored water temperature rises but the duration of the rise exceeds the preset temperature detection value has not yet reached the preset monitoring time period, it indicates that there is still a certain amount of residual cold storage capacity available in the pipe network, or that the current temperature rise is only a short-term fluctuation. At this time, the controller maintains the two-stage low-load operation control mode, continuing to drive the chilled water in the pipe network to supply cooling to the terminal via the chilled water secondary pump, without starting the chiller. This state can continue until the start of the normal operation phase the next day, at which time the controller automatically exits the low-load control mode and returns to the control mode under normal operating conditions. In this embodiment, the start time of the normal operation phase the next day is set to 8:00 AM the next day. As an optional implementation, the start time of the normal operation phase the next day can also be set to 6:00 AM, 7:00 AM, or 9:00 AM the next day according to the actual usage schedule of the building. By clearly defining the exit conditions, the control method of the present invention can operate completely throughout the low-load period (such as at night) and automatically and seamlessly switch to the normal operation control mode when the daytime load recovers, without manual intervention, ensuring the efficient operation of the system around the clock.

[0062] Furthermore, to verify the optimized control performance of the method of the present invention, a performance comparison was conducted using a control method under normal operating conditions. Under this method, the control strategy of normal operating conditions is still adopted under low-load operating conditions. One year of operating data was selected for performance testing, with data intervals of 5 minutes. Data from January to August was used for training, and data from September to December was used for validation.

[0063] See Figure 3 The diagram illustrates a comparison of daily energy consumption under low-load operating conditions in a given week, using different methods provided by embodiments of the present invention. Traditional methods still employ normal operating control strategies under low-load conditions. Therefore, even when the cooling load demand is significantly lower than the cooling capacity of one or two chillers, the chillers remain running, resulting in low chiller efficiency and wasted cooling capacity. In contrast, the present invention uses a two-stage operating control strategy to automatically identify low-load operating conditions. Under low-load conditions, it effectively reduces the energy consumption of the air conditioning chiller plant by shutting down the chillers and using pipe network cold storage to supply cooling to the terminals, achieving an average energy saving rate of 17.5% and significantly improving the system's operating efficiency under low-load conditions.

[0064] This invention provides a two-stage optimized control method for an air conditioning chiller plant under low-load operation conditions, comprising: periodically determining whether the chiller plant's operating conditions meet preset conditions based on the number of chillers in operation and their load rate; if so, switching to a preset first-stage low-load operation control mode and shutting down the target chiller and its associated equipment; wherein the target chiller is the chiller with the smallest capacity currently in operation; under the first-stage low-load operation control mode, determining whether to switch to a preset second-stage low-load operation control mode based on the preset total cooling load demand during the future self-cooling period and the network's cold storage capacity. In the two-stage low-load operation control mode, the pipeline network is used for cold storage. After the cold storage is completed, all operating chillers and their supporting equipment are shut down, and the low-temperature chilled water in the pipeline network is driven by the chilled water secondary pump to supply cooling to the terminal. The supply water temperature in the pipeline network is continuously monitored. If the supply water temperature is continuously higher than the preset temperature detection value within the preset monitoring period, the last shut-down chiller and its corresponding supporting equipment are started. Through two-stage optimized control, low-load conditions can be identified in a timely manner, avoiding continuous operation of chillers in the inefficient range and effectively reducing the system's energy consumption under low-load conditions.

[0065] Based on the above embodiments, this invention provides a two-stage optimized control device for air conditioning chiller plants under low-load operating conditions. (See attached image.) Figure 4 The diagram shown is a structural schematic of a two-stage optimization control device for an air conditioning chiller plant under low-load operating conditions, provided by an embodiment of the present invention. The device includes: The first-stage control module 310 is used to periodically determine whether the operating conditions of the chiller station meet the preset conditions based on the number of chillers turned on and the load rate of the chillers. If so, it switches to the preset first-stage low-load operation control mode and shuts down the target chiller and its supporting equipment. The target chiller is the chiller with the smallest capacity currently in operation.

[0066] The cooling data processing module 320 is used to determine whether to switch to the preset second-stage low-load operation control mode based on the total cooling load demand and the cold storage capacity of the pipeline network during the preset future cooling time period under the first-stage low-load operation control mode.

[0067] The second-stage control module 330 is used to store cold in the pipeline network under the second-stage low-load operation control mode. After the cold storage is completed, all operating chillers and their supporting equipment are shut down so that the low-temperature chilled water in the pipeline network can be driven by the chilled water secondary pump to supply cold to the terminal.

[0068] The cooling data monitoring module 340 is used to continuously monitor the water supply temperature in the pipeline network. If the water supply temperature continues to exceed the preset temperature detection value within the preset monitoring period, the last shut-down chiller and corresponding supporting equipment will be started.

[0069] Furthermore, in some preferred embodiments of the present invention, the self-cooling time period is 1 hour; the cooling data processing module 320 is used to predict the total cooling load demand in the next 1 hour based on a pre-established ultra-short-term building load prediction model, and to determine the cooling storage capacity of the pipeline network based on the parameters of the pipeline network; if the total cooling load demand in the next 1 hour is less than the cooling storage capacity of the pipeline network, then it is determined to switch to the second-stage low-load operation control mode; otherwise, the first-stage low-load operation control mode is maintained.

[0070] Furthermore, in some preferred embodiments of the present invention, the cooling data processing module 320 is used to take the hourly cooling load of the past 24 hours and the cooling load of every 5 minutes of the past 1 hour as input to the model, and take the cooling load value of the next 5 minutes as the predicted value of the model output; through a rolling prediction method, the previous predicted value is used as the input variable for a new round, and the calculation is repeated until the total cooling load demand within the cooling period is obtained; wherein, the total cooling load demand is determined by the following formula: ;in, This represents the total cooling load demand for the next hour. This represents the cooling load for the i-th 5-minute interval within the next hour.

[0071] Furthermore, in some preferred embodiments of the present invention, the cooling data processing module 320 is used to determine the cooling capacity of the pipeline network using the following formula: ;in, To enhance the cold storage capacity of the pipeline network; This is the preset temperature rise of the cold network; Cold network pipe section The diameter, Cold network pipe section Length, The density of water, The specific heat capacity of water at constant pressure. Cold network pipe section The capacity factor.

[0072] Furthermore, in some preferred embodiments of the present invention, the two-stage control module 330 is used to calculate the chilled water outlet temperature setpoint of the chiller through a reinforcement learning controller, and control the operation of the chiller according to the chilled water outlet temperature setpoint to achieve cold storage; wherein, the reinforcement learning controller is established using a deep Q-learning algorithm, including a Q-value network, a target network, and an experience replay pool; both the Q-value network and the target network are constructed using deep neural networks with consistent structures, the input being a state variable vector, and the output being a Q-value vector for each possible action.

[0073] Furthermore, in some preferred embodiments of the present invention, the two-stage control module 330 is used to input the current round state variables into the Q-value network to obtain the Q-values ​​of each possible action, and take the action corresponding to the maximum Q-value as the optimal setpoint for the chilled water outlet temperature, as shown in the following formula: ; ; ;in, Set the chilled water outlet temperature; For the i-th action Q value; Let k be the state variable in round k. For state variables; The chilled water supply temperature for the pipeline network; This refers to the return temperature of the chilled water in the pipe network. This is the difference between the cold storage capacity of the pipeline network and the total cooling load demand in the future period. This represents the current operating power of the chiller.

[0074] Furthermore, in some preferred embodiments of the present invention, the conditions for determining the completion of cold storage include: the actual supply temperature of chilled water reaches the set value of the chiller's chilled water supply temperature, and the actual supply and return temperature difference of chilled water is less than a preset threshold.

[0075] Furthermore, in some preferred embodiments of the present invention, the two-stage control module 330 is used to update the parameters of the reinforcement learning controller through the following steps: storing the current round's state variables, action variables, reward function values, and the new round's state variables in the form of quadruples into the experience replay pool; wherein, the reward function value is determined by the following formula: Where R is the reward function value; The time required for the cold network to complete cold storage. The preset weighting coefficients, During the cold storage period The chiller's operating power at any given time; The time interval is defined as follows: every preset number of rounds, a preset number of samples are randomly selected from the experience replay pool, and the parameters of the Q-value network are updated using the following formula: ; ;in, The loss function; For preset quantity; For Q-value networks, the current state and actions The predicted Q value; Discount factor; For the target network to determine the next round state The maximum Q value that can be output, For learning rate, The loss function with respect to network parameters The gradient; whenever the Q-value network completes an update a preset number of times, the parameters of the target network are updated to the current parameters of the Q-value network.

[0076] Furthermore, in some preferred embodiments of the present invention, the cooling data monitoring module 340 is also used to maintain a two-stage low-load operation control mode if the duration of the water supply temperature being greater than the preset temperature detection value does not reach the preset monitoring time period, until the normal operation stage of the next day is reached, and then restore the control mode under normal operation conditions.

[0077] Those skilled in the art will clearly understand that, for the sake of convenience and brevity, the specific working process of the two-stage optimization control device for air conditioning chiller under low-load operating conditions described above can be referred to the corresponding process in the embodiments of the two-stage optimization control method for air conditioning chiller under low-load operating conditions, and will not be repeated here.

[0078] This invention also provides an electronic device for implementing a two-stage optimized control method for air conditioning chiller plants under low-load operating conditions; see [link to related documentation]. Figure 5 The schematic diagram of an electronic device provided by the embodiment of the present invention shown above includes a memory 400 and a processor 401. The memory 400 is used to store one or more computer instructions, which are executed by the processor 401 to realize the above-mentioned two-stage optimization control method for air conditioning chiller under low load operating conditions.

[0079] Furthermore, Figure 5The electronic device shown also includes a bus 402 and a communication interface 403. The processor 401, the communication interface 403 and the memory 400 are connected via the bus 402.

[0080] The memory 400 may include high-speed random access memory (RAM) and may also include non-volatile memory, such as at least one disk storage device. Communication between this system network element and at least one other network element is achieved through at least one communication interface 403 (which can be wired or wireless), such as the Internet, wide area network, local area network, metropolitan area network, etc. The bus 402 can be an ISA bus, PCI bus, or EISA bus, etc. The bus can be divided into address bus, data bus, control bus, etc. For ease of representation, Figure 5 The symbol is represented by a single double-headed arrow, but this does not mean that there is only one bus or one type of bus.

[0081] Processor 401 may be an integrated circuit chip with signal processing capabilities. In implementation, each step of the above method can be completed by the integrated logic circuitry in the hardware of processor 401 or by instructions in software form. Processor 401 can be a general-purpose processor, including a Central Processing Unit (CPU), a Network Processor (NP), etc.; it can also be a Digital Signal Processor (DSP), an Application Specific Integrated Circuit (ASIC), a Field-Programmable Gate Array (FPGA), or other programmable logic devices, discrete gate or transistor logic devices, or discrete hardware components. It can implement or execute the methods, steps, and logic block diagrams disclosed in the embodiments of this invention. The general-purpose processor can be a microprocessor or any conventional processor. The steps of the methods disclosed in the embodiments of this invention can be directly manifested as execution by a hardware decoding processor, or execution by a combination of hardware and software modules in the decoding processor. The software module can reside in a readily available storage medium in the art, such as random access memory, flash memory, read-only memory, programmable read-only memory, electrically erasable programmable memory, or registers. This storage medium is located in memory 400, and processor 401 reads information from memory 400 and, in conjunction with its hardware, completes the steps of the method described in the foregoing embodiments.

[0082] This invention also provides a computer-readable storage medium storing computer-executable instructions. When these computer-executable instructions are called and executed by a processor, they cause the processor to implement the aforementioned two-stage optimized control method for air conditioning chiller stations under low-load operating conditions. For specific implementation details, please refer to the method embodiments, which will not be repeated here.

[0083] The computer program product of the two-stage optimization control method, device and electronic equipment for air conditioning chiller under low load operation provided in the embodiments of the present invention includes a computer-readable storage medium storing program code. The instructions included in the program code can be used to execute the methods in the preceding method embodiments. For specific implementation, please refer to the method embodiments, which will not be repeated here.

[0084] Those skilled in the art will clearly understand that, for the sake of convenience and brevity, the specific working process of the system and / or device described above can be referred to the corresponding process in the foregoing method embodiments, and will not be repeated here.

[0085] Furthermore, in the description of the embodiments of the present invention, unless otherwise explicitly specified and limited, the terms "installation," "connection," and "linking" should be interpreted broadly. For example, they can refer to a fixed connection, a detachable connection, or an integral connection; they can refer to a mechanical connection or an electrical connection; they can refer to a direct connection or an indirect connection through an intermediate medium; and they can refer to the internal connection of two components. Those skilled in the art can understand the specific meaning of the above terms in the present invention based on the specific circumstances.

[0086] If the aforementioned functions are implemented as software functional units and sold or used as independent products, they can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of this invention, essentially, or the part that contributes to the prior art, or a portion of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods described in the various embodiments of this invention. The aforementioned storage medium includes various media capable of storing program code, such as USB flash drives, portable hard drives, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical disks.

[0087] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, and not to limit them; although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some or all of the technical features; and these modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the scope of the technical solutions of the embodiments of the present invention.

Claims

1. A two-stage optimization control method for an air conditioning cold station in a low-load operation condition, characterized in that, include: Periodically determine whether the operating conditions of the chiller plant meet preset conditions based on the number of chillers in operation and the load rate of the chillers. If so, switch to the preset first-stage low-load operation control mode and shut down the target chiller and its supporting equipment; wherein, the target chiller is the chiller with the smallest capacity currently in operation. In the first-stage low-load operation control mode, the system determines whether to switch to the preset second-stage low-load operation control mode based on the total cooling load demand during the preset future cooling period and the cold storage capacity of the pipeline network. If so, in the second-stage low-load operation control mode, the pipeline network is used for cold storage, and after the cold storage is completed, all operating chillers and their supporting equipment are shut down, so that the low-temperature chilled water in the pipeline network is driven by the chilled water secondary pump to supply cooling to the terminal. The water supply temperature in the pipeline network is continuously monitored. If the water supply temperature continues to be higher than the preset temperature detection value within the preset monitoring period, the last shut-down chiller and the corresponding supporting equipment are started.

2. The method of claim 1, wherein, The self-cooling time period is 1 hour; the determination of whether to switch to the preset two-stage low-load operation control mode based on the total cooling load demand within the preset future self-cooling time period and the cold storage capacity of the pipeline network includes: The total cooling load demand for the next hour is predicted based on a pre-established ultra-short-term building load forecasting model, and the cooling storage capacity of the pipe network is determined based on the parameters of the pipe network. If the total cooling load demand in the next hour is less than the network's cooling storage capacity, then switch to the second-stage low-load operation control mode; otherwise, maintain the first-stage low-load operation control mode.

3. The method of claim 2, wherein, The total cooling load demand predicted for the next hour based on a pre-established ultra-short-term building load forecasting model includes: The hourly cooling load of the past 24 hours and the cooling load of every 5 minutes of the past hour are used as inputs to the model, and the cooling load value of the next 5 minutes is used as the predicted output value of the model. By using a rolling forecasting method, the previous forecast value is used as the input variable for the new round of calculations, and the calculations are repeated until the total cooling load demand for the self-supplying period is obtained; wherein, the total cooling load demand is determined by the following formula: ； wherein, is the total cooling load demand for the next 1 hour; is the cooling load for the i-th 5 minutes within the next 1 hour.

4. The method of claim 2, wherein, The cold storage capacity of the pipeline network is determined by the following formula: ； in, The cold storage capacity of the pipeline network; This is the preset temperature rise of the cold network; Cold network pipe section The diameter, Cold network pipe section Length, The density of water, The specific heat capacity of water at constant pressure. Cold network pipe section The capacity factor.

5. The method according to claim 1, characterized in that, The cold storage of the pipeline network includes: The chiller's chilled water outlet temperature setpoint is calculated by a reinforcement learning controller, and the chiller is controlled to operate according to the setpoint to achieve the chilled water storage. The reinforcement learning controller is built using a deep Q-learning algorithm, including a Q-value network, a target network, and an experience replay pool. Both the Q-value network and the target network are constructed using deep neural networks with consistent structures. The input is a state variable vector, and the output is a Q-value vector for each possible action.

6. The method according to claim 5, characterized in that, Determining the setpoint for the chilled water outlet temperature includes: The state variables of this round are input into the Q-value network to obtain the Q-value of each possible action, and the action corresponding to the largest Q-value is taken as the optimal setpoint for the chilled water outlet temperature, as shown in the following formula: ； ； ； in, Set the outlet temperature of the chilled water to a specified value; For the i-th action Q value; Let k be the state variable in round k. For state variables; The chilled water supply temperature for the pipeline network; This refers to the return temperature of the chilled water in the pipe network. This is the difference between the cold storage capacity of the pipeline network and the total cooling load demand in the future period. This represents the current operating power of the chiller.

7. The method according to claim 5, characterized in that, The criteria for determining whether cold storage is complete include: The actual chilled water supply temperature reaches the chiller chilled water supply temperature set value, and the actual chilled water supply and return temperature difference is less than the preset threshold.

8. The method according to claim 5, characterized in that, The method further includes updating the parameters of the reinforcement learning controller through the following steps: The current round's state variables, action variables, reward function value, and the new round's state variables are stored in the experience replay pool as a four-tuple; wherein, the reward function value is determined by the following formula: ； Where R is the value of the reward function; The time required for the cold network to complete cold storage. The preset weighting coefficients, During the cold storage period The chiller's operating power at any given time; For time intervals; Every preset number of rounds, a preset number of samples are randomly selected from the experience replay pool, and the parameters of the Q-value network are updated using the following formula: ； ； in, The loss function; For preset quantity; For Q-value networks, the current state and actions The predicted Q value; Discount factor; For the target network to determine the next round state The maximum Q value that can be output, For learning rate, The loss function with respect to network parameters The gradient; After the Q-value network completes a preset number of updates, the parameters of the target network are updated to the current parameters of the Q-value network.

9. The method according to claim 1, characterized in that, In the two-stage low-load operation control mode, after storing cold energy in the pipeline network and shutting down all operating chillers and their auxiliary equipment after the cold energy storage is completed, the method further includes: If the duration for which the water supply temperature is greater than the preset temperature detection value does not reach the preset monitoring time period, the two-stage low-load operation control mode will be maintained until the normal operation stage of the next day is reached, at which point the control mode under normal operation conditions will be restored.

10. A two-stage optimized control device for an air conditioning chiller plant under low-load operating conditions, characterized in that, include: The first-stage control module is used to periodically determine whether the operating conditions of the chiller station meet the preset conditions based on the number of chillers turned on and the load rate of the chillers. If so, it switches to the preset first-stage low-load operation control mode and shuts down the target chiller and the supporting equipment of the target chiller. The target chiller is the chiller with the smallest capacity currently in operation. The cooling data processing module is used to determine whether to switch to the preset second-stage low-load operation control mode based on the total cooling load demand and the cold storage capacity of the pipeline network during the preset future cooling time period under the first-stage low-load operation control mode. The two-stage control module is used to, if so, store cold in the pipeline network under the two-stage low-load operation control mode, and after the cold storage is completed, shut down all operating chillers and their supporting equipment, so that the low-temperature chilled water in the pipeline network can be driven by the chilled water secondary pump to supply cold to the terminal. The cooling data monitoring module is used to continuously monitor the water supply temperature in the pipeline network. If the water supply temperature continues to be higher than the preset temperature detection value within the preset monitoring time period, the last shut-down chiller and the corresponding supporting equipment will be started.

Images