A partition type horizontal regenerator and a control method thereof

Abstract

The application discloses a kind of partition plate type horizontal cold storage tank and its control method, belong to cold storage tank technical field, the cold storage tank includes horizontal tank body, its opposite ends are respectively set as water inlet cavity and water outlet cavity, inside is divided into multiple transition cavities by several cavity partitions, adjacent cavities are communicated by water conveying passage, and spoiler is staggered in each cavity It is installed with flow uniforming plate near the inlet of water conveying passage;Its control method is constructed by obtaining historical load data multi-time scale load prediction model, combined with each cavity temperature and differential pressure state information, generates filling and discharging cold scheduling strategy by multi-objective optimization and is decomposed into valve execution instruction, simultaneously with flow state instability monitoring recovery and filling and discharging cold economic scheduling function.The application is suppressed by partition, spoiler and flow uniforming plate and destroys inclined temperature layer, cooperates with valve grading control and multi-objective optimization scheduling, improves the effective cold storage capacity of cold storage tank while reducing operating electricity fee and prolonging equipment life.

Description

Technical Field

[0001] This invention relates to the field of cold storage tank technology, specifically to a diaphragm-type horizontal cold storage tank and its control method. Background Technology

[0002] Data centers, as the core infrastructure of the digital economy, generate a large amount of heat during the operation of their IT equipment, requiring continuous cooling systems to maintain normal operating temperatures. In data center cooling systems, cold storage tanks serve as key equipment for backup cold sources and load shaving, providing uninterrupted cooling in the event of chiller failures or power outages, while also utilizing peak-valley electricity price differences for economical operation. Horizontal cold storage tanks are widely used in data center cooling systems due to their advantages such as small footprint, flexible installation, and stable liquid levels. During the charging process, the chilled water and return water in the cold storage tank form a thermal stratification due to temperature differences. During the cooling process, this stratification structure dynamically evolves as cold energy is released, with the transition region between the thermal strata being the thermocline. The thickness of the thermocline directly determines the effective cold storage capacity of the cold storage tank; a thinner thermocline results in a larger effective cold storage capacity and stronger cooling capacity. Therefore, the stability control of the thermocline is a core indicator for evaluating the operational quality of the cold storage tank.

[0003] However, existing horizontal chilled water storage tanks have significant shortcomings in controlling the temperature gradient. The significance of the temperature gradient lies in its thickness, which determines the upper limit of the available cooling capacity of the storage tank. A thin temperature gradient means clear stratification of hot and cold water, allowing for a stable output of chilled water close to the design temperature during cooling until the cooling capacity is exhausted. Conversely, a thicker temperature gradient means an expanded mixing zone of hot and cold water, resulting in a decrease in effective cooling capacity, a premature rise in the cooling outlet temperature, and a significant reduction in the actual backup time of the storage tank. In severe cases, it may not be able to sustain operation until the backup cooling source is activated or power is restored. Currently, existing technologies have extremely limited means of intervening in the temperature gradient, mostly relying on passive structures such as installing flow equalization plates at the inlet of the horizontal tank or adding baffles inside the tank, attempting to reduce the impact on temperature stratification by physically blocking the water flow to evenly disperse in the inlet area. However, such passive structures can only be optimized for specific flow rates and operating conditions during the design phase. Once the charging and discharging flow rates fluctuate or the operating conditions change, their flow equalization effect will significantly decrease. More importantly, the existing control strategies for cold storage tanks rely entirely on temperature thresholds, determining the start and stop of charging and discharging based solely on whether the outlet water temperature reaches the set value. They do not take the thickness of the thermocline as an active control variable, nor do they establish a quantitative relationship between flow distribution and thermocline evolution. They cannot actively adjust the flow field distribution based on real-time operating conditions to suppress the growth of the thermocline. Therefore, there is an urgent need for a cold storage tank control scheme that can incorporate the thickness of the thermocline into the control closed loop and actively intervene based on operating conditions to maintain the stability of the thermocline. Summary of the Invention

[0004] To address the problems mentioned in the background section, this invention provides a diaphragm-type horizontal cold storage tank and its control method. The aforementioned objective of this application is achieved through the following technical solution:

[0005] A partition-type horizontal cold storage tank includes a horizontal tank body, with an inlet chamber and an outlet chamber at opposite ends. The tank body has several interconnected transition chambers arranged within it based on partitions. Adjacent transition chambers, the inlet chamber, and the outlet chamber are connected by water conveyance channels. Several flow-disrupting plates are alternately arranged within the inlet chamber, the outlet chamber, and the transition chambers. A flow-equalizing plate is also fixedly installed at one end of each transition chamber near the inlet of the water conveyance channel.

[0006] By adopting the above scheme, a horizontal tank is used as the main structure. The two opposite ends of the horizontal tank are respectively set as inlet and outlet chambers, allowing chilled water to enter from one end and exit from the other, forming a unidirectional flow path throughout the entire tank. The interior of the horizontal tank is divided into several interconnected transition chambers by several cavity partitions. Adjacent transition chambers, inlet chambers, and outlet chambers are connected by water conveyance channels, allowing water to flow sequentially through each chamber along a serpentine path within the tank. This extends the flow path length within the tank and increases the heat exchange time between the chilled water and the cold storage medium. Several interfering flow patterns are staggered within the inlet chamber, outlet chamber, and several transition chambers. The baffles and baffles alternately obstruct and guide the water flow within the cavity, forcing it to flow in a meandering pattern. This effectively disrupts the straight-line inertia of the water flow, resulting in a more uniform distribution of the water flow across the cavity's cross-section and preventing localized high-speed water flows from directly impacting the thermocline interface. Several transition cavities also have flow equalization plates fixedly installed at one end near the water inlet. These plates perform secondary rectification of the water flow entering the cavity, eliminating localized high-speed jets and eddies, ensuring that the water flows into each cavity at a uniform surface velocity. Through the combined action of the cavity baffles, baffles, and flow equalization plates, the water flow within the tank is continuously redistributed and homogenized during the flow process, effectively suppressing the destructive growth of the thermocline.

[0007] In a preferred embodiment, this application may be further configured such that: a water conveying valve that controls the water conveying efficiency based on the opening degree is fixedly installed inside the water conveying channel, and the water conveying valve is fixedly installed at the entrance of the water conveying channel.

[0008] By adopting the above scheme, water conveying valves that control water conveying efficiency based on their opening degree are fixedly installed inside the water conveying channel, and the water conveying valves are fixedly installed at the entrance of the water conveying channel, so that the flow area of ​​the water conveying channel between each cavity can be independently adjusted; by adjusting the opening degree of each water conveying valve, the local resistance coefficient of each channel can be changed, thereby controlling the water flow distribution ratio through each cavity; the design of installing the water conveying valves at the entrance of the channel allows the water flow to be throttled and regulated before entering the downstream cavity. The regulated water flow directly enters the flow equalization plate area of ​​the downstream cavity after passing through the channel, and is further rectified by the flow equalization plate before entering the main cavity area; by actively adjusting the opening degree of each valve to change the flow distribution of each cavity, when the thickness of the thermocline layer in a certain cavity increases, the flow rate of that cavity can be reduced to reduce the disturbance to the thermocline layer, while the flow rate of other cavities can be increased to maintain the total flow rate unchanged, so that the flow distribution of each cavity matches the current thermocline layer state of each cavity.

[0009] In a preferred embodiment, this application may be further configured such that: a plurality of flow equalization holes for buffering water flow are equidistantly opened on the flow equalization plate; a main inlet valve is installed on the side of the inlet chamber away from the water conveying channel, and a main outlet valve is installed on the side of the outlet chamber away from the water conveying channel.

[0010] By adopting the above scheme, several flow equalization holes for buffering water flow are equidistantly opened on the flow equalization plate. The flow equalization holes disperse the water flow that is concentrated through the passage into multiple fine jets. After the fine jets mix with each other downstream of the flow equalization plate, they form a uniform velocity distribution, eliminating the local high-speed zone and low-speed dead zone at the passage outlet. A main inlet valve is installed on the side of the inlet chamber away from the water supply passage to control the total flow rate of chilled water entering the cold storage tank. A main outlet valve is installed on the side of the outlet chamber away from the water supply passage to control the total flow rate of chilled water leaving the cold storage tank. The main inlet valve and the main outlet valve work together to determine the overall charging and discharging rate of the cold storage tank. The main inlet valve, the main outlet valve and each water supply valve form a hierarchical control structure. The main valve controls the overall on / off state and the total flow rate, while each water supply valve controls the flow distribution of each chamber. The cooperation between the two ensures both precise control of the total flow rate and differentiated distribution of the flow rate in each chamber.

[0011] The second objective of this invention is achieved through the following technical solution:

[0012] A control method for a diaphragm-type horizontal cold storage tank, applied to a diaphragm-type horizontal cold storage tank, includes the following steps:

[0013] S10: Obtain historical load information, historical chilled water supply and return temperature information, and historical ambient temperature and humidity information of the data center, and build a multi-time-scale load prediction model;

[0014] S20: Obtain the cavity status information of each independent cavity of the horizontal cold storage tank. The cavity status information includes the top temperature information, bottom temperature information, multi-point temperature information of the cavity sidewall, pressure difference information of the water conveying channel, and pressure difference information before and after the flow equalization plate.

[0015] S30: Based on the cold load prediction information output by the multi-timescale load prediction model and combined with the cavity status information, generate a cold storage tank charging and discharging scheduling strategy based on a pre-set multi-objective optimization algorithm.

[0016] S40: Decompose the charging and discharging cooling scheduling strategy into execution instruction information, which includes the opening and closing timing instructions for the main inlet valve and the main outlet valve of the cold storage tank, as well as the opening adjustment instructions for the water supply valves between each chamber.

[0017] By adopting the above scheme, historical load information, historical chilled water supply and return temperature information, and historical environmental temperature and humidity information of the data center are acquired, and a multi-timescale load prediction model is constructed to provide forward prediction capability for cooling capacity scheduling. By acquiring the top temperature information, bottom temperature information, multi-point temperature information of the sidewalls of each independent cavity of the horizontal cold storage tank, pressure difference information of the water conveying channel, and pressure difference information before and after the flow equalization plate, real-time cavity operation status perception is provided for optimized scheduling. Based on the cold load prediction information and cavity status information, a multi-objective optimization algorithm is used to generate a cold storage tank charging and discharging scheduling strategy, so that the scheduling scheme can meet the cold load demand while taking into account the stability of the thermocline, equipment life and operating economy. The charging and discharging scheduling strategy is decomposed into the opening and closing timing commands of the main inlet valve and the main outlet valve, as well as the opening adjustment commands of the water conveying valves between each cavity, so that the theoretically optimal strategy can be transformed into specific operation commands that can be responded to by each execution component.

[0018] In a preferred embodiment, this application can be further configured such that step S10 includes the following steps:

[0019] S101: Obtain real-time power consumption information of data center IT equipment and aggregate it into equipment load time series according to a preset time window; obtain supply and return water temperature difference and flow information of chilled water system, and generate historical cooling load time series by combining the equipment load time series.

[0020] S102: Decompose the historical cooling load time series into minute-level ultra-short-term components, hour-level short-term components, and day-level medium-term components.

[0021] S103: Based on the Long Short-Term Memory network, corresponding prediction sub-models are constructed for the ultra-short-term component, short-term component and medium-term component respectively, and the outputs of the three sub-models are weighted and fused through the attention mechanism to obtain multi-timescale cold load prediction information.

[0022] By adopting the above scheme, real-time power consumption information of data center IT equipment is acquired and aggregated into equipment load time series according to a preset time window. At the same time, the supply and return water temperature difference and flow information of the chilled water system are acquired and combined with the equipment load time series to generate historical cooling load time series, providing an accurate input data foundation for the prediction model. The historical cooling load time series is decomposed into minute-level ultra-short-term components, hour-level short-term components, and day-level medium-term components, enabling the model to capture transient fluctuations, hourly variation patterns, and daily periodic trends of the load at different time scales. Based on the Long Short-Term Memory network, corresponding prediction sub-models are constructed for each component, and the outputs of the three sub-models are weighted and fused through an attention mechanism. The attention mechanism is used to adaptively adjust the weight of the prediction results at each time scale, so that the final prediction results are dominated by the most reliable sub-model in different time periods.

[0023] In a preferred embodiment, this application can be further configured such that step S20 includes the following steps:

[0024] S201: Temperature sensors placed at the top and bottom of each cavity are used to obtain the top and bottom temperature information of each cavity and calculate the real-time temperature gradient value of each cavity; a distributed fiber optic temperature measurement system is used to obtain the multi-point temperature distribution along the axial and radial directions of the sidewalls of each cavity, generate multi-point temperature information of the sidewalls of the cavities, and construct a temperature field cloud map inside each cavity.

[0025] S202: The differential pressure information of the water conveyance channel is obtained in real time by differential pressure sensors arranged at both ends of the water conveyance channel between each cavity. The real-time flow information of each channel is calculated based on the differential pressure information of the water conveyance channel and the pre-stored channel flow resistance characteristic curve. The differential pressure information of the flow equalization plate before and after each flow equalization plate is obtained in real time by differential pressure sensors arranged before and after each flow equalization plate. The real-time blockage degree of each flow equalization plate is evaluated based on the differential pressure information of the flow equalization plate before and after the flow equalization plate and the pre-stored flow equalization plate orifice characteristic curve.

[0026] By adopting the above scheme, temperature sensors placed at the top and bottom of each cavity acquire top and bottom temperature information and calculate real-time temperature gradient values, providing basic data for calculating the effective cold storage capacity and thermocline thickness of each cavity; a distributed fiber optic temperature measurement system acquires multi-point temperature distribution along the axial and radial directions of each cavity sidewall and generates multi-point temperature information of the cavity sidewall, constructing a temperature field cloud map inside each cavity, enabling the system to accurately determine the position and shape of the thermocline in three-dimensional space; differential pressure sensors placed at both ends of the water conveyance channels between each cavity acquire real-time differential pressure information of the water conveyance channels, and calculate the real-time flow information of each channel in combination with the channel flow resistance characteristic curve, providing feedback basis for valve opening adjustment; differential pressure sensors placed before and after each flow equalization plate acquire real-time differential pressure information before and after the flow equalization plate, and evaluate the real-time blockage degree of each flow equalization plate in combination with the flow equalization plate orifice characteristic curve, providing data support for online monitoring and predictive maintenance of the flow equalization plate blockage status.

[0027] In a preferred embodiment, this application can be further configured such that step S30 includes the following steps:

[0028] S301: Calculate the real-time effective cold storage capacity of each cavity based on the top temperature information and bottom temperature information, and calculate the slope temperature layer thickness value of each cavity based on the temperature field cloud map of each cavity. Compare the slope temperature layer thickness value of each cavity with the preset slope temperature layer stability threshold to determine the flow stability level of each cavity.

[0029] S302: Calculate the predicted total cooling capacity required by the cold storage tank based on the cold load forecast information, and calculate the current effective cold storage capacity based on the temperature field cloud map of each cavity; when the current effective cold storage capacity is less than the predicted total cooling capacity required, it is determined that the cold storage tank cannot independently meet the cold load demand, and the refrigeration unit is triggered to supplement cooling to compensate for the cooling capacity gap; when the current effective cold storage capacity is less than the difference between the upper limit of the total cold storage capacity of the cold storage tank and the preset safety margin, and the current period is a low electricity price period, it is determined that the cold storage tank enters the charging mode to supplement the cold capacity reserve; when the current effective cold storage capacity is greater than the predicted total cooling capacity required and the current period is a high electricity price period, it is determined that the cold storage tank enters the releasing mode to release cold capacity to replace the operation of the refrigeration unit; when the current period is a flat electricity price period, with the goal of maintaining the current cold storage capacity not lower than the preset lower limit safety value, the mode switching is not triggered, and the current operating mode is maintained;

[0030] S303: Based on the cold load prediction information and the thickness of the inclined temperature layer of each cavity, the target total inflow and outflow of the cold storage tank is determined by a multi-objective optimization algorithm, and the target opening of the water conveyance channel between each cavity is allocated in combination with the flow stability level of each cavity, so that the flow path of water in the cold storage tank is distributed according to the temperature stratification state of each cavity.

[0031] The multi-objective optimization algorithm includes the following constraints: the target valve opening of the water conveyance channel between each cavity is not less than the minimum allowable opening corresponding to the current blockage level of the channel, and the minimum allowable opening is positively correlated with the blockage level of the flow equalization plate; and under the premise of meeting the cold load distribution requirements, the opening adjustment priority of the channel corresponding to the cavity with a high flow stability level is higher than that of the channel corresponding to the cavity with a low flow stability level.

[0032] The multi-objective optimization algorithm takes minimizing the weighted sum of squares of the deviation of the oblique temperature layer thickness in each cavity as the first objective function, minimizing the sum of valve action frequencies in the water conveyance channels between all cavities as the second objective function, and minimizing the sum of flow direction reversal impact energy during the switching of the cooling mode in all cavities as the third objective function. It solves the Pareto optimal solution set through a non-dominated sorting genetic algorithm with an elite strategy, and selects an optimal compromise solution from the Pareto optimal solution set as the final scheduling scheme based on the current operating scenario.

[0033] By adopting the above scheme, the real-time effective cold storage capacity of each cavity is calculated using top and bottom temperature information. The thickness of the slope temperature layer in each cavity is calculated using temperature field cloud maps, and the flow stability level of each cavity is determined, providing a thermodynamic feasibility assessment basis for load allocation. Based on cooling load forecast information and temperature field cloud maps of each cavity, the refrigeration unit is triggered to supplement cooling to prevent insufficient cooling when the current effective cold storage capacity is less than the predicted total required cooling release. Cooling supplementation is determined when the current effective cold storage capacity is sufficient and electricity prices are low; cooling release is determined during peak electricity prices; and the current mode is maintained during periods of flat electricity prices, thus ensuring the system... The switching decision integrates three dimensions: load demand, cold storage status, and economy. A multi-objective optimization algorithm is used to determine the target total inflow and outflow water flow and allocate the target opening of the water conveyance channels between each cavity. The first objective function is to minimize the weighted sum of the squares of the deviation of the oblique temperature layer thickness in each cavity, the second objective function is to minimize the sum of the valve action frequencies, and the third objective function is to minimize the sum of the flow direction reversal impact energy. A non-dominated sorting genetic algorithm with an elite strategy is used to solve the Pareto optimal solution set and select the optimal compromise solution, thus achieving the optimal trade-off among multiple competing optimization objectives.

[0034] In a preferred embodiment, this application can be further configured such that step S40 is followed by step S50, wherein step S50 includes the following steps:

[0035] S501: Real-time monitoring of the temperature change rate at the top and bottom of each cavity. When the temperature change rate of any cavity exceeds a preset rate threshold, it is determined that a flow instability event has occurred in that cavity. In response to the flow instability event, the main flow path of the cavity is cut off by closing the upstream water supply valve and reducing the opening of the downstream water supply valve, while maintaining only the bypass micro-circulation flow rate. The flow recovery program of the cavity is then initiated. The flow recovery program maintains the water circulation in the cavity at a preset low flow rate until the thickness of the inclined temperature layer in the cavity recovers to a preset stable range.

[0036] S502: Based on the real-time pressure difference value of the water conveyance channel between each cavity and the target valve opening command, generate the expected opening time information and expected closing time information of the valves in each channel, and generate the valve action sequence table corresponding to each channel according to the expected opening time information and expected closing time information of the valves in each channel.

[0037] S503: Obtain the action characteristic parameter information of the water supply valves between each chamber. The action characteristic parameter information includes the full stroke time information of the valve, the pressure drop change curve during the valve opening process, and the water hammer pressure peak information during the valve closing process. Based on the pressure drop change curve and water hammer pressure peak information of each valve, calculate the superimposed pressure pulse peak information generated when any two valves operate simultaneously. With minimizing the maximum value of the superimposed pressure pulse peak information within the action time window of all valves as the optimization objective, and with the expected opening time information and expected closing time information of each valve as adjustable variables, perform peak-shifting optimization processing on the valve action timing table through a genetic algorithm to stagger the action times of each passage valve, thereby avoiding the superposition of pressure pulses generated by the simultaneous opening or closing of multiple passage valves.

[0038] By adopting the above scheme, the system monitors the temperature change rate at the top and bottom of each cavity in real time. When the temperature change rate of any cavity exceeds the preset rate threshold, it is determined that a flow instability event has occurred in that cavity. This allows the system to detect abnormalities in the early stages of the disruption of the thermocline. In response to the flow instability event, the system closes the upstream water valve of the cavity and reduces the opening of the downstream water valve, cutting off the main flow path and maintaining only the bypass micro-circulation velocity. This allows the cavity to recover its thermocline structure under low disturbance conditions, while other cavities continue to provide normal cooling during the flow recovery period. Based on the real-time differential pressure value of each passage and the target valve opening command, the system generates the expected opening and closing time information of each passage valve and generates a valve action sequence table. The optimization objective is to minimize the maximum value of the superimposed pressure pulse peak within the time window of all valve actions. The system uses a genetic algorithm to perform peak-shifting optimization on the valve action sequence table, ensuring that the action times of each passage valve are staggered. This effectively avoids the impact damage to the pipeline system and cavity structure caused by the superposition of pressure pulses generated by the simultaneous opening or closing of multiple passage valves.

[0039] In a preferred embodiment, this application may be further configured such that step S01 is included before step S30, wherein step S01 includes the following steps:

[0040] S011: Obtain real-time electricity price information and future time-of-use electricity price forecast information for grid connection;

[0041] S012: Based on the cooling load forecast information and future time-of-use electricity price forecast information output by the multi-time-scale cooling load forecast model, calculate the charging cost and cooling revenue in each time window. With the goal of minimizing the total operating electricity cost of the cold storage tank, and combining the upper limit of the total cold storage capacity of the cold storage tank and the upper limit of the independent cold storage capacity and charging / discharging rate of each cavity, generate a charging / discharging economic scheduling plan across time windows through a dynamic programming algorithm. The charging / discharging economic scheduling plan includes the decision of whether the cold storage tank should be in charging mode or cooling mode in each time window.

[0042] S013: Based on the historical operating data of each cavity, establish a charging and discharging cold efficiency decay model for each cavity. The charging and discharging cold efficiency decay model characterizes the degradation law of the equivalent cold storage capacity of the cavity after experiencing different flow rates and temperature cycles.

[0043] S014: When generating the cold storage economic scheduling plan through dynamic programming algorithm, the cold storage efficiency decay model of each cavity is introduced as a penalty term into the objective function to balance the water flow frequency of each cavity and extend the overall service life of the cold storage tank.

[0044] By adopting the above scheme, and by acquiring real-time electricity price information and future time-of-use electricity price prediction information from the grid, the cold storage tank can actively charge and store energy during off-peak hours and release cold energy to replace the operation of the refrigeration unit during peak hours, thereby reducing operating costs by utilizing the peak-valley price difference. A dynamic programming algorithm is used to generate a cross-time window economic scheduling plan for charging and discharging, achieving optimal global cost while meeting cooling load demands. A charging and discharging efficiency decay model is established based on historical operating data of each chamber, characterizing the degradation of the equivalent cold storage capacity after the chambers undergo different flow rates and temperature cycles, enabling the system to quantify the aging degree of each chamber. The efficiency decay model is introduced as a penalty term into the objective function, allowing the optimization algorithm to actively balance the water flow frequency of each chamber while pursuing optimal economic efficiency, preventing some chambers from aging faster due to frequent use, and effectively extending the overall service life of the cold storage tank.

[0045] The above-mentioned objective of this application is achieved through the following technical solution:

[0046] A control system for a diaphragm-type horizontal cold storage tank, used to execute a control method for a diaphragm-type horizontal cold storage tank, comprising:

[0047] The load forecasting module is used to acquire historical load information, historical chilled water supply and return temperature information, and historical ambient temperature and humidity information of the data center, and to build a multi-time-scale load forecasting model.

[0048] The status sensing module is used to acquire the cavity status information of each independent cavity of the horizontal cold storage tank. The cavity status information includes the top temperature information, bottom temperature information, multi-point temperature information of the cavity sidewall, pressure difference information of the water conveying channel, and pressure difference information before and after the flow equalization plate.

[0049] The scheduling optimization module is used to generate a cold storage tank charging and discharging scheduling strategy based on the cold load forecast information output by the multi-time scale load forecast model, combined with the cavity status information, and based on a pre-set multi-objective optimization algorithm.

[0050] The execution decomposition module decomposes the charging and discharging cooling scheduling strategy into execution instruction information, which includes the opening and closing timing instructions for the main inlet valve and the main outlet valve of the cold storage tank, as well as the opening adjustment instructions for the water supply valves between each chamber.

[0051] The beneficial effects of the diaphragm-type horizontal cold storage tank and its control method of the present invention are as follows:

[0052] 1. This technical solution divides the interior of a horizontal tank into multiple independent cavities using partitions. These cavities are interconnected via serpentine water channels, allowing water to flow sequentially through each cavity along a meandering path within the tank. This forcibly extends the water flow path length and heat exchange time. Simultaneously, baffles are interspersed within each cavity to disrupt the straight-flow inertia of the water, and flow equalization plates are installed at the inlets of the water channels to eliminate localized jets. This combination of structures continuously redirects, deflects, and homogenizes the water flow within the tank, fundamentally suppressing the destructive growth of the thermocline. Furthermore, independently installed water valves at the inlets of each water channel allow for differentiated and precise adjustment of the flow rate in each cavity based on the current thermocline state. This upgrade from passive flow equalization to active flow distribution actively incorporates the thermocline thickness into the control loop, significantly improving the effective cold storage capacity and continuous cold release capability of the cold storage tank during dynamic operation.

[0053] 2. This technical solution combines a multi-timescale load prediction model with multi-dimensional state perception information such as the top temperature, bottom temperature, multi-point temperature on the sidewalls, and temperature field cloud map of each cavity. It uses three synergistic optimization objective functions: minimizing the weighted sum of squared deviations in the thermocline layer thickness of each cavity, minimizing the sum of valve action frequencies, and minimizing the sum of flow direction reversal impact energy. A non-dominated sorting genetic algorithm with an elite strategy is used to solve for the Pareto optimal solution set, and the optimal compromise solution is selected as the final scheduling scheme. This ensures that the load allocation decision simultaneously considers thermodynamic stability, equipment lifespan, and operational stability. Furthermore, by real-time monitoring of the temperature change rate at the top and bottom of each cavity, a flow recovery procedure is triggered in the early stages of thermocline layer instability. This involves closing the upstream and downstream valves of the affected cavity to cut off the main flow path and maintain only a bypass microcirculation. This allows the disturbed cavity to recover its thermocline layer structure under low-disturbance conditions while other cavities continue to provide normal cooling. This achieves a closed-loop management upgrade from passively enduring thermocline layer damage to actively sensing and repairing flow instability, ensuring the cooling reliability and thermal stratification stability of the cold storage tank under all operating conditions.

[0054] 3. This technical solution introduces a dynamic programming algorithm to generate a cross-time window charging and discharging economic scheduling plan before scheduling optimization. It combines real-time electricity price information and future time-of-use electricity price forecasts to calculate the charging cost and discharging revenue within each time window. This allows the cold storage tank to actively charge and store energy during off-peak electricity price periods and release cooling capacity to replace the operation of the refrigeration unit during peak electricity price periods, effectively utilizing the peak-valley electricity price difference to reduce the total operating cost. At the same time, a charging and discharging efficiency decay model is established based on the historical operating data of each chamber, and the decay law of each chamber is introduced as a penalty term into the optimization objective function. This allows the scheduling decision to actively balance the water flow frequency of each chamber while pursuing optimal economy, avoiding the uneven loss of some chambers aging faster due to frequent use while other chambers are idle for a long time. This allows the cold storage tank to achieve a longer service life as a whole, realizing the synergistic optimization of operating economy and equipment durability. Attached Figure Description

[0055] Figure 1 This is a first three-dimensional structural diagram of the horizontal tank body of the horizontal cold storage tank and its control method according to the present application.

[0056] Figure 2 This is a second three-dimensional structural diagram of the horizontal tank body of the horizontal cold storage tank and its control method according to the present application;

[0057] Figure 3 This is a schematic flowchart illustrating an embodiment of the control method of a diaphragm-type horizontal cold storage tank and its control method according to this application.

[0058] Figure 4This is a flowchart illustrating step S10 of an embodiment of the control method of a partition-type horizontal cold storage tank and its control method according to this application.

[0059] Among them, 1. Horizontal tank; 2. Main inlet valve; 3. Inlet cavity; 4. Water conveying passage; 5. Transition cavity; 6. Outlet cavity; 7. Main outlet valve; 401. Cavity partition; 402. Water conveying valve; 501. Baffle; 502. Flow equalization plate; 521. Flow equalization hole. Detailed Implementation

[0060] The technical solution of the present invention will now be clearly and completely described with reference to the accompanying drawings. Obviously, the described embodiments are only some, not all, of the embodiments of the present invention. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0061] In one embodiment, such as Figure 1-2 As shown, this application discloses a partition-type horizontal cold storage tank, including a horizontal tank body 1. The two opposite ends of the horizontal tank body 1 are respectively configured as an inlet cavity 3 and an outlet cavity 6. The interior of the horizontal tank body 1 is provided with a plurality of interconnected transition cavities 5 based on a plurality of cavity partitions 401. Adjacent transition cavities 5, inlet cavities 3 and outlet cavities 6 are connected by a water conveying channel 4. A plurality of interference flow plates 501 are staggered in the inlet cavity 3, outlet cavity 6 and the plurality of transition cavities 5. A flow equalization plate 502 is also fixedly installed at one end of the plurality of transition cavities near the inlet of the water conveying channel 4.

[0062] Specifically, the horizontal tank 1 has an inlet chamber 3 and an outlet chamber 6 at opposite ends. Each inlet chamber 3 has several distributed optical fibers on its sidewall, allowing chilled water to enter from one end and exit from the other, forming a unidirectional overall flow throughout the tank. The interior of the horizontal tank 1 is divided into several interconnected transition chambers 5 by several chamber partitions 401. Adjacent transition chambers 5, inlet chambers 3, and outlet chambers 6 are connected by water conveyance channels 4, allowing water to flow sequentially through each chamber along a serpentine path within the tank, effectively extending the flow path length and residence time. Several interference flow plates 501 are staggered within the inlet chambers 3, outlet chambers 6, and transition chambers 5. During its flow, the water is alternately blocked and guided by the baffles 501, forcing it to flow back and forth, thus ensuring a uniform distribution of the water flow across the cross-section of the cavity and preventing local high-speed water flow from directly impacting the thermocline interface. Several transition cavities 5 are also fixedly installed with flow equalization plates 502 near the inlet of the water conveyance channel 4. The flow equalization plates 502 perform secondary rectification on the water flow entering the cavity, eliminating local high-speed jets and eddies. Through the combined action of the cavity baffles 401, baffles 501, and flow equalization plates 502, the water flow is continuously redirected, reversed, and homogenized throughout the entire process, effectively suppressing the destructive growth of the thermocline and ensuring the stability of the thermal stratification and the effective thermal storage capacity of the cold storage tank during the charging and discharging process.

[0063] In one embodiment, a water conveying valve 402 that controls the water conveying efficiency based on the opening degree is fixedly installed inside the water conveying channel 4, and the water conveying valve 402 is fixedly installed at the entrance of the water conveying channel 4.

[0064] Specifically, a water conveying valve 402 is fixedly installed inside the water conveying channel 4, which controls the water conveying efficiency based on the opening and closing degree. The water conveying valve 402 is fixedly installed at the inlet of the water conveying channel 4, so that the flow area of ​​the water conveying channel 4 between each cavity can be independently adjusted. By adjusting the opening and closing degree of each water conveying valve 402, the local resistance coefficient of each channel is changed, thereby controlling the water flow distribution ratio through each cavity, so that the flow distribution of each cavity matches the current slope temperature state of each cavity. The design of installing the water conveying valve 402 at the inlet of the channel allows the water flow to be throttled and regulated before entering the downstream cavity. After the regulated water flow passes through the channel, it directly enters the flow equalization plate 502 area of ​​the downstream cavity, is further rectified by the flow equalization plate 502, and then enters the main cavity area. This realizes the series coordination of flow distribution regulation and flow equalization rectification, so that the flow rate of each cavity can be adjusted in a differentiated and fine manner according to its own slope temperature state, while avoiding direct disturbance of the downstream flow field by valve regulation.

[0065] In one embodiment, the flow equalization plate 502 is provided with a plurality of flow equalization holes 521 for buffering water flow at equal intervals; a main water inlet valve 2 is installed on the side of the water inlet cavity 3 away from the water conveying channel 4, and a main water outlet valve 7 is installed on the side of the water outlet cavity 6 away from the water conveying channel 4.

[0066] Specifically, the flow equalization plate 502 has several flow equalization holes 521 evenly spaced to buffer the water flow. The flow equalization holes 521 disperse the water flow that is concentrated in the passage into multiple fine jets. After the fine jets mix with each other downstream of the flow equalization plate 502, they form a uniform velocity distribution, eliminating the local high-speed zone and low-speed dead zone at the passage outlet. This allows the water flow entering each cavity to be uniformly propelled by surface velocity rather than flowing through the cavity cross-section by jet impact. The inlet cavity 3 is equipped with a main inlet valve 2 on the side away from the water conveyance passage 4, and the outlet cavity... A main outlet valve 7 is installed on the side of body 6 away from the water supply passage 4. The main inlet valve 2 and the main outlet valve 7 work together to determine the overall charging and discharging rate of the cold storage tank. The main inlet valve 2, the main outlet valve 7 and each water supply valve 402 form a hierarchical control structure. The main valve controls the overall on / off state and the total flow rate, while each water supply valve 402 controls the flow distribution of each cavity. The two work together to ensure precise control of the total flow rate and realize differentiated flow distribution of each cavity, so that the overall control and refined internal distribution of the cold storage tank do not interfere with each other.

[0067] In one embodiment, such as Figure 3 As shown, this application also discloses a control method for a diaphragm-type horizontal cold storage tank, applied to a diaphragm-type horizontal cold storage tank, including the following steps:

[0068] S10: Obtain historical load information, historical chilled water supply and return temperature information, and historical ambient temperature and humidity information of the data center, and build a multi-time-scale load prediction model;

[0069] S20: Obtain the cavity status information of each independent cavity of the horizontal cold storage tank. The cavity status information includes the top temperature information, bottom temperature information, multi-point temperature information of the cavity sidewall, pressure difference information of the water conveying channel, and pressure difference information before and after the flow equalization plate.

[0070] S30: Based on the cold load prediction information output by the multi-timescale load prediction model and combined with the cavity status information, generate a cold storage tank charging and discharging scheduling strategy based on a pre-set multi-objective optimization algorithm.

[0071] S40: Decompose the charging and discharging cooling scheduling strategy into execution instruction information, which includes the opening and closing timing instructions for the main inlet valve and the main outlet valve of the cold storage tank, as well as the opening adjustment instructions for the water supply valves between each chamber.

[0072] In this embodiment, historical load information refers to the recorded operating load data of IT equipment in the data center over a past period; historical chilled water supply and return temperature information refers to the recorded supply and return water temperatures of the chilled water system over the past period; historical ambient temperature and humidity information refers to the recorded ambient temperature and humidity data of the location where the cold storage tank is located over a past period; the multi-timescale load prediction model is a mathematical model that can simultaneously output the predicted cooling load results at different time granularities in the future; cavity status information is a data set reflecting the current operating status of each independent cavity; top temperature information is the temperature measurement value of the upper region of each cavity; bottom temperature information is the temperature measurement value of the lower region of each cavity; and multi-point temperature information of the cavity sidewalls consists of multiple measuring points along the axial and radial directions of each cavity sidewall. Temperature distribution data; pressure difference information of the water conveyance channel refers to the fluid pressure difference between the two ends of the water conveyance channel; pressure difference information before and after the flow equalization plate refers to the fluid pressure difference on both sides of the flow equalization plate; cooling load forecast information refers to the future cooling load demand forecast value output by the multi-timescale load forecasting model; multi-objective optimization algorithm refers to the mathematical optimization method that considers multiple optimization objectives simultaneously; charging and discharging cold scheduling strategy refers to the decision scheme for determining the operation mode and flow distribution of the cold storage tank in each time period; execution instruction information refers to the data set that converts the scheduling strategy into specific control signals; the opening and closing sequence instruction information of the main inlet valve and the main outlet valve refers to the control signal that controls the opening and closing sequence of the main inlet valve and the main outlet valve; the opening degree adjustment instruction information refers to the control signal that controls the opening degree of each water conveyance valve.

[0073] Specifically, historical load information, historical chilled water supply and return temperature information, and historical environmental temperature and humidity information of the data center are acquired, and a multi-timescale load prediction model is constructed to provide a forward-looking basis for subsequent cooling load scheduling. Top temperature information, bottom temperature information, multi-point temperature information of the sidewalls of the chambers, pressure difference information of the water conveying channel, and pressure difference information before and after the flow equalization plate of each independent chamber of the horizontal cold storage tank are acquired to provide real-time chamber operation status awareness for optimized scheduling. Based on the cold load prediction information and chamber status information, a multi-objective optimization algorithm is used to generate a cold storage tank charging and discharging scheduling strategy, so that the scheduling scheme can meet the cold load demand while taking into account the stability of the plenum and the economic efficiency of operation. The charging and discharging scheduling strategy is decomposed into the opening and closing timing instructions of the main inlet valve and the main outlet valve, as well as the opening adjustment instructions of the water conveying valves between each chamber, so that the theoretically optimal strategy can be transformed into specific operation instructions that can be responded to by each execution component, completing the entire link from global prediction, status awareness, optimization decision-making to instruction decomposition.

[0074] In one embodiment, such as Figure 4 As shown, step S10 includes the following steps:

[0075] S101: Obtain real-time power consumption information of data center IT equipment and aggregate it into equipment load time series according to a preset time window; obtain supply and return water temperature difference and flow information of chilled water system, and generate historical cooling load time series by combining the equipment load time series.

[0076] S102: Decompose the historical cooling load time series into minute-level ultra-short-term components, hour-level short-term components, and day-level medium-term components.

[0077] S103: Based on the Long Short-Term Memory network, corresponding prediction sub-models are constructed for the ultra-short-term component, short-term component and medium-term component respectively, and the outputs of the three sub-models are weighted and fused through the attention mechanism to obtain multi-timescale cold load prediction information.

[0078] In this embodiment, real-time power consumption information refers to the current power consumption data of data center IT equipment; equipment load time series refers to the power consumption data sequence of IT equipment arranged in chronological order; supply and return water temperature difference refers to the difference between the supply and return water temperatures of chilled water; historical cooling load time series refers to the past cooling load demand data sequence arranged in chronological order; ultra-short-term component refers to the load fluctuation component at the minute-level time scale; short-term component refers to the load change component at the hour-level time scale; medium-term component refers to the load trend component at the day-level time scale; long short-term memory network refers to a recurrent neural network structure that can effectively handle long-term dependencies in time series; and attention mechanism refers to a weighted fusion method that can adaptively adjust the importance of each input feature.

[0079] Specifically, real-time power consumption information of data center IT equipment is acquired and aggregated into equipment load time series according to a preset time window. Simultaneously, the supply and return water temperature difference and flow rate information of the chilled water system are acquired and combined with the equipment load time series to generate a historical cooling load time series. This provides an accurate input data foundation for the prediction model, enabling the prediction results to truly reflect the dynamic relationship between the heat dissipation demand of IT equipment and the cooling capacity of the chilled water system. The historical cooling load time series is decomposed into minute-level ultra-short-term components, hour-level short-term components, and day-level medium-term components, allowing the model to capture transient fluctuations, hourly variation patterns, and daily periodic trends of the load at different time scales. Based on a long short-term memory network, corresponding prediction sub-models are constructed for each component, and the outputs of the three sub-models are weighted and fused using an attention mechanism. The attention mechanism adaptively adjusts the weights of the prediction results at each time scale, ensuring that the final prediction results are dominated by the most reliable sub-model at different time periods. This yields multi-time-scale cooling load prediction information that balances rapid response capability and long-term trend accuracy.

[0080] In one embodiment, step S20 includes the following steps:

[0081] S201: Temperature sensors placed at the top and bottom of each cavity are used to obtain the top and bottom temperature information of each cavity and calculate the real-time temperature gradient value of each cavity; a distributed fiber optic temperature measurement system is used to obtain the multi-point temperature distribution along the axial and radial directions of the sidewalls of each cavity, generate multi-point temperature information of the sidewalls of the cavities, and construct a temperature field cloud map inside each cavity.

[0082] S202: The differential pressure information of the water conveyance channel is obtained in real time by differential pressure sensors arranged at both ends of the water conveyance channel between each cavity. The real-time flow information of each channel is calculated based on the differential pressure information of the water conveyance channel and the pre-stored channel flow resistance characteristic curve. The differential pressure information of the flow equalization plate before and after each flow equalization plate is obtained in real time by differential pressure sensors arranged before and after each flow equalization plate. The real-time blockage degree of each flow equalization plate is evaluated based on the differential pressure information of the flow equalization plate before and after the flow equalization plate and the pre-stored flow equalization plate orifice characteristic curve.

[0083] In this embodiment, the top temperature information is the temperature value measured by a temperature sensor located at the top of the cavity; the bottom temperature information is the temperature value measured by a temperature sensor located at the bottom of the cavity; the real-time temperature gradient value is the difference between the top and bottom temperatures, reflecting the rate of temperature change in the vertical direction of the cavity; the distributed fiber optic temperature measurement system is a temperature measurement system that uses optical fibers as distributed temperature sensors; the temperature field cloud map is a two-dimensional or three-dimensional visualization graphic reflecting the temperature distribution inside the cavity; the channel flow resistance characteristic curve is a characteristic curve describing the relationship between the pressure difference at both ends of the water conveying channel and the flow rate; the real-time flow rate information is the volumetric flow rate value of each water conveying channel at the current moment; the flow equalization plate orifice characteristic curve is a characteristic curve describing the relationship between the pressure difference before and after the flow equalization plate and the flow rate; and the real-time blockage degree is the degree of decrease in the flow capacity of the flow equalization plate due to scaling or blockage.

[0084] Specifically, temperature sensors placed at the top and bottom of each cavity acquire top and bottom temperature information and calculate the real-time temperature gradient value of each cavity, providing basic data for calculating the effective cold storage capacity of each cavity and assessing the thickness of the thermocline layer. The temperature difference between the top and bottom directly reflects the intensity of the thermal stratification. A distributed fiber optic temperature measurement system acquires the multi-point temperature distribution along the axial and radial directions of each cavity sidewall and generates multi-point temperature information of the cavity sidewall, constructing a temperature field cloud map inside each cavity, enabling the system to accurately determine the position, thickness, and shape of the thermocline layer in three-dimensional space. Differential pressure sensors placed at both ends of the water conveyance channels between each cavity acquire real-time differential pressure information of the water conveyance channels, and calculate the real-time flow information of each channel in combination with the flow resistance characteristic curve of the channel, providing a flow feedback basis for valve opening adjustment. Differential pressure sensors placed before and after each flow equalization plate acquire real-time differential pressure information before and after the flow equalization plate, and assess the real-time blockage degree of each flow equalization plate in combination with the orifice characteristic curve of the flow equalization plate, enabling the system to monitor the scaling status of the flow equalization plate online without disassembling the equipment, providing a quantitative basis for predictive maintenance.

[0085] In one embodiment, step S30 includes the following steps:

[0086] S301: Calculate the real-time effective cold storage capacity of each cavity based on the top temperature information and bottom temperature information, and calculate the slope temperature layer thickness value of each cavity based on the temperature field cloud map of each cavity. Compare the slope temperature layer thickness value of each cavity with the preset slope temperature layer stability threshold to determine the flow stability level of each cavity.

[0087] S302: Calculate the predicted total cooling capacity required by the cold storage tank based on the cold load forecast information, and calculate the current effective cold storage capacity based on the temperature field cloud map of each cavity; when the current effective cold storage capacity is less than the predicted total cooling capacity required, it is determined that the cold storage tank cannot independently meet the cold load demand, and the refrigeration unit is triggered to supplement cooling to compensate for the cooling capacity gap; when the current effective cold storage capacity is less than the difference between the upper limit of the total cold storage capacity of the cold storage tank and the preset safety margin, and the current period is a low electricity price period, it is determined that the cold storage tank enters the charging mode to supplement the cold capacity reserve; when the current effective cold storage capacity is greater than the predicted total cooling capacity required and the current period is a high electricity price period, it is determined that the cold storage tank enters the releasing mode to release cold capacity to replace the operation of the refrigeration unit; when the current period is a flat electricity price period, with the goal of maintaining the current cold storage capacity not lower than the preset lower limit safety value, the mode switching is not triggered, and the current operating mode is maintained;

[0088] S303: Based on the cold load prediction information and the thickness of the inclined temperature layer of each cavity, the target total inflow and outflow of the cold storage tank is determined by a multi-objective optimization algorithm, and the target opening of the water conveyance channel between each cavity is allocated in combination with the flow stability level of each cavity, so that the flow path of water in the cold storage tank is distributed according to the temperature stratification state of each cavity.

[0089] The multi-objective optimization algorithm includes the following constraints: the target valve opening of the water conveyance channel between each cavity is not less than the minimum allowable opening corresponding to the current blockage level of the channel, and the minimum allowable opening is positively correlated with the blockage level of the flow equalization plate; and under the premise of meeting the cold load distribution requirements, the opening adjustment priority of the channel corresponding to the cavity with a high flow stability level is higher than that of the channel corresponding to the cavity with a low flow stability level.

[0090] The multi-objective optimization algorithm takes minimizing the weighted sum of squares of the deviation of the oblique temperature layer thickness in each cavity as the first objective function, minimizing the sum of valve action frequencies in the water conveyance channels between all cavities as the second objective function, and minimizing the sum of flow direction reversal impact energy during the switching of the cooling mode in all cavities as the third objective function. It solves the Pareto optimal solution set through a non-dominated sorting genetic algorithm with an elite strategy, and selects an optimal compromise solution from the Pareto optimal solution set as the final scheduling scheme based on the current operating scenario.

[0091] In this embodiment, the real-time effective cold storage capacity is the value of the cold energy currently stored and available for release in each cavity; the slope temperature layer thickness is the vertical thickness value of the transition region between the cold and hot stratification interfaces; the preset slope temperature layer stability threshold is the critical thickness value for determining whether the slope temperature layer is stable; the flow stability level is the classification of the stability of the flow state of each cavity based on the slope temperature layer thickness; the predicted total cold energy required for release is the total cold energy that needs to be released from the cold storage tank according to the cold load prediction calculation; the current effective cold storage capacity is the sum of the available cold energy currently stored in all cavities; the upper limit of the total cold storage capacity is the maximum cold energy value that the cold storage tank can store; the preset safety margin is the safety margin reserved to prevent overcharging; the target total inflow and outflow water flow rate is the overall cooling rate of the cold storage tank determined by optimization; The standard opening degree is the expected opening value of the valve in the water conveyance channel between each cavity; the minimum allowable opening degree is the minimum opening value that ensures the valve is not blocked and can be stably adjusted; the weighted sum of squares is the sum of the squares of the deviation of the oblique temperature layer thickness of each cavity multiplied by its weight; the valve action frequency is the number of times the water conveyance valve is opened or closed per unit time; the flow direction reversal impact energy is the impact energy value generated by the sudden change in water flow direction when switching between charging and discharging cooling modes; the non-dominated sorting genetic algorithm with elite strategy is a multi-objective optimization evolutionary algorithm that retains the best solution set; the Pareto optimal solution set is the optimal solution set among multiple optimization objectives that cannot be further improved by trade-offs; the optimal compromise solution is the single solution with the best overall performance in the current operating scenario selected from the Pareto optimal solution set.

[0092] Specifically, the system calculates the real-time effective cold storage capacity of each cavity based on top and bottom temperature information, calculates the slope layer thickness of each cavity based on the temperature field cloud map of each cavity, and compares the slope layer thickness of each cavity with a preset slope layer stability threshold to determine the flow stability level of each cavity. This provides a thermodynamic feasibility judgment basis for load allocation, enabling the system to identify which cavities are suitable for load bearing and which cavities require protective operation. Based on cold load forecast information, the system calculates the predicted total cold release capacity of the cold storage tank and calculates the current effective cold storage capacity based on the temperature field cloud map of each cavity. When the effective cold storage capacity is insufficient, the refrigeration unit is triggered to supplement cooling to ensure cooling supply safety. When the effective cold storage capacity is sufficient and the electricity price is low... The system determines cooling during off-peak hours, cooling during peak electricity price periods, and maintains the current mode during flat electricity price periods, making mode switching decisions integrate three dimensions: load demand, cooling storage status, and economic efficiency. A multi-objective optimization algorithm is used to determine the target total inflow and outflow water flow and allocate the target opening of the water conveyance channels between each cavity. The first objective function is to minimize the weighted sum of the squared deviations of the oblique temperature layer thickness in each cavity, the second objective function is to minimize the sum of the valve action frequencies, and the third objective function is to minimize the sum of the flow direction reversal impact energy. A non-dominated sorting genetic algorithm with an elite strategy is used to solve the Pareto optimal solution set and select the optimal compromise solution based on the current operating scenario, achieving the optimal trade-off among multiple competing optimization objectives.

[0093] In one embodiment, step S40 is followed by step S50, which includes the following steps:

[0094] S501: Real-time monitoring of the temperature change rate at the top and bottom of each cavity. When the temperature change rate of any cavity exceeds a preset rate threshold, it is determined that a flow instability event has occurred in that cavity. In response to the flow instability event, the main flow path of the cavity is cut off by closing the upstream water supply valve and reducing the opening of the downstream water supply valve, while maintaining only the bypass micro-circulation flow rate. The flow recovery program of the cavity is then initiated. The flow recovery program maintains the water circulation in the cavity at a preset low flow rate until the thickness of the inclined temperature layer in the cavity recovers to a preset stable range.

[0095] S502: Based on the real-time pressure difference value of the water conveyance channel between each cavity and the target valve opening command, generate the expected opening time information and expected closing time information of the valves in each channel, and generate the valve action sequence table corresponding to each channel according to the expected opening time information and expected closing time information of the valves in each channel.

[0096] S503: Obtain the action characteristic parameter information of the water supply valves between each chamber. The action characteristic parameter information includes the full stroke time information of the valve, the pressure drop change curve during the valve opening process, and the water hammer pressure peak information during the valve closing process. Based on the pressure drop change curve and water hammer pressure peak information of each valve, calculate the superimposed pressure pulse peak information generated when any two valves operate simultaneously. With minimizing the maximum value of the superimposed pressure pulse peak information within the action time window of all valves as the optimization objective, and with the expected opening time information and expected closing time information of each valve as adjustable variables, perform peak-shifting optimization processing on the valve action timing table through a genetic algorithm to stagger the action times of each passage valve, thereby avoiding the superposition of pressure pulses generated by the simultaneous opening or closing of multiple passage valves.

[0097] In this embodiment, the top temperature change rate is the rate of change of the temperature at the top of the cavity over time; the bottom temperature change rate is the rate of change of the temperature at the bottom of the cavity over time; the preset rate threshold is the critical temperature change rate value for determining flow instability; the flow instability event is an abnormal operating state event in which the thermocline is severely damaged; the bypass microcirculation flow rate is an extremely low flow rate that maintains a small amount of water flow to sense the temperature without causing disturbance; the flow recovery procedure is a special operating sequence for restoring the stability of the thermocline; the expected opening time information is the predicted time for the valve to reach full opening from its current state; the expected closing time information is the predicted time for the valve to reach full closing from its current state; the valve movement... The timing table is an action plan table for each valve in the passage arranged in chronological order; the action characteristic parameter information is a set of parameters describing the physical characteristics of the valve movement process; the full stroke time information is the total time required for the valve to go from fully closed to fully open; the pressure drop change curve is the curve of the pressure difference between the two ends of the valve changing with time during the valve opening process; the water hammer pressure peak information is the maximum impact pressure value generated at the moment the valve closes; the superimposed pressure pulse peak information is the maximum value of the pressure pulses superimposed when multiple valves operate simultaneously; the genetic algorithm is a global optimization search algorithm that simulates the natural evolution process; the peak staggering optimization process is an optimization process that staggers the pressure peaks in time by adjusting the action time of each valve.

[0098] Specifically, the system monitors the temperature change rates at the top and bottom of each cavity in real time. When the temperature change rate of any cavity exceeds a preset threshold, it determines that a flow instability event has occurred in that cavity. This allows the system to detect anomalies in the early stages of thermocline disruption, preventing the instability from worsening. In response to the flow instability event, the system closes the upstream water valve and reduces the opening of the downstream water valve, cutting off the main flow path of the cavity and maintaining only the bypass microcirculation velocity. This allows the cavity to recover its thermocline structure under low disturbance conditions, while other cavities continue to operate during the flow recovery period. During normal cooling, instability in a single cavity will not affect the cooling continuity of the entire cold storage tank. Based on the real-time differential pressure value of each channel and the target valve opening command, the expected opening and closing time information of each channel valve is generated, and a valve action sequence table is generated. With minimizing the maximum value of the superimposed pressure pulse peak within the time window of all valve actions as the optimization objective, the valve action sequence table is optimized by a genetic algorithm to stagger the peaks, so that the action times of each channel valve are staggered, effectively avoiding the impact damage to the pipeline system and cavity structure caused by the superposition of pressure pulses generated by the simultaneous opening or closing of multiple channel valves.

[0099] In one embodiment, step S01 is included before step S30, and step S01 includes the following steps:

[0100] S011: Obtain real-time electricity price information and future time-of-use electricity price forecast information for grid connection;

[0101] S012: Based on the cooling load forecast information and future time-of-use electricity price forecast information output by the multi-time-scale cooling load forecast model, calculate the charging cost and cooling revenue in each time window. With the goal of minimizing the total operating electricity cost of the cold storage tank, and combining the upper limit of the total cold storage capacity of the cold storage tank and the upper limit of the independent cold storage capacity and charging / discharging rate of each cavity, generate a charging / discharging economic scheduling plan across time windows through a dynamic programming algorithm. The charging / discharging economic scheduling plan includes the decision of whether the cold storage tank should be in charging mode or cooling mode in each time window.

[0102] S013: Based on the historical operating data of each cavity, establish a charging and discharging cold efficiency decay model for each cavity. The charging and discharging cold efficiency decay model characterizes the degradation law of the equivalent cold storage capacity of the cavity after experiencing different flow rates and temperature cycles.

[0103] S014: When generating the cold storage economic scheduling plan through dynamic programming algorithm, the cold storage efficiency decay model of each cavity is introduced as a penalty term into the objective function to balance the water flow frequency of each cavity and extend the overall service life of the cold storage tank.

[0104] In this embodiment, real-time electricity price information refers to the electricity price data obtained from the power grid at the current moment; future time-of-use electricity price prediction information refers to the electricity price prediction data for each time period in the future; charging cost is the electricity expenditure required for charging during off-peak hours; cooling benefit is the electricity expenditure saved by cooling during peak hours; total cycle operating electricity cost is the total electricity expenditure of the cold storage tank throughout the entire operating cycle; the maximum total cold storage capacity is the maximum amount of cold energy that the entire cold storage tank can store; the maximum independent cold storage capacity is the maximum amount of cold energy that each individual chamber can store; and the maximum charging / discharging rate is... The maximum allowable charging or discharging rate for each cavity; the dynamic programming algorithm is an optimization algorithm that solves a multi-stage decision problem by decomposing it into subproblems; the cross-time window charging and discharging economic scheduling plan is the optimal charging and discharging arrangement across multiple electricity price periods; the charging and discharging efficiency decay model is a mathematical model describing the degradation of cavity performance over time; the equivalent cold storage capacity is the effective cold storage capacity of the cavity after considering efficiency decay; the degradation law is the decreasing trend of the cavity's cold storage capacity with increasing cycle number; the penalty term is the additional cost term in the objective function used to suppress adverse behavior.

[0105] Specifically, the system acquires real-time electricity price information and future time-of-use electricity price forecasts from the grid, enabling the cold storage tank to actively charge and store energy during off-peak hours and release cooling capacity to replace the refrigeration unit during peak hours, thus reducing operating costs by utilizing the peak-valley price difference. Based on cooling load forecasts and future time-of-use electricity price forecasts, the system calculates the charging cost and releasing revenue for each time window. With the goal of minimizing the total operating cost, a dynamic programming algorithm is used to generate a cross-time window charging and releasing economic scheduling plan, taking into account the capacity and rate limits of each chamber, to achieve global cost optimization while meeting cooling load demands. Based on the historical operating data of each chamber, a charging and releasing efficiency decay model is established to characterize the degradation of the equivalent cold storage capacity of the chamber after experiencing different flow rates and temperature cycles, enabling the system to quantify the aging degree of each chamber and predict its remaining effective capacity. The efficiency decay model is introduced as a penalty term into the objective function, allowing the optimization algorithm to actively balance the water flow frequency of each chamber while pursuing optimal economic efficiency, preventing some chambers from aging faster due to frequent use and effectively extending the overall service life of the cold storage tank.

[0106] In one embodiment, a control system for a diaphragm-type horizontal cold storage tank is provided, which corresponds one-to-one with the control method for a diaphragm-type horizontal cold storage tank in the above embodiment.

[0107] The present invention and its embodiments have been described above. This description is not restrictive. The accompanying drawings are only one embodiment of the present invention. The actual structure is not limited to this. In short, if a person skilled in the art is inspired by this description and designs a similar structure and embodiment without departing from the spirit of the present invention, such design should fall within the protection scope of the present invention.

Claims

1. A partition-type horizontal cold storage tank, characterized in that: The system includes a horizontal tank (1), with an inlet chamber (3) and an outlet chamber (6) at opposite ends. The interior of the horizontal tank (1) is provided with several interconnected transition chambers (5) based on several chamber partitions (401). Adjacent transition chambers (5), inlet chambers (3), and outlet chambers (6) are connected by a water conveying channel (4). Several flow-disrupting plates (501) are staggered in the inlet chamber (3), outlet chamber (6), and several transition chambers (5). A flow equalization plate (502) is also fixedly installed at one end of several transition chambers (5) near the inlet of the water conveying channel (4).

2. A partition-type horizontal cold storage tank according to claim 1, characterized in that: The water conveying channel (4) is fixedly installed with a water conveying valve (402) that controls the water conveying efficiency based on the degree of opening and closing. The water conveying valve (402) is fixedly installed at the entrance of the water conveying channel (4).

3. A partition-type horizontal cold storage tank according to claim 1, characterized in that: The flow equalization plate (502) is provided with a plurality of flow equalization holes (521) for buffering water flow at equal intervals; a main water inlet valve (2) is installed on the side of the water inlet chamber (3) away from the water conveying channel (4), and a main water outlet valve (7) is installed on the side of the water outlet chamber (6) away from the water conveying channel (4).

4. A control method for a diaphragm-type horizontal cold storage tank, applied to the diaphragm-type horizontal cold storage tank according to any one of claims 1-3, characterized in that: Including the following steps: S10: Obtain historical load information, historical chilled water supply and return temperature information, and historical ambient temperature and humidity information of the data center, and build a multi-time-scale load prediction model; S20: Obtain the cavity status information of each independent cavity of the horizontal cold storage tank. The cavity status information includes the top temperature information, bottom temperature information, multi-point temperature information of the cavity sidewall, pressure difference information of the water conveying channel, and pressure difference information before and after the flow equalization plate. S30: Based on the cold load prediction information output by the multi-timescale load prediction model and combined with the cavity status information, generate a cold storage tank charging and discharging scheduling strategy based on a pre-set multi-objective optimization algorithm. S40: Decompose the charging and discharging cooling scheduling strategy into execution instruction information, which includes the opening and closing timing instructions for the main inlet valve and the main outlet valve of the cold storage tank, as well as the opening adjustment instructions for the water supply valves between each chamber.

5. The control method for a diaphragm-type horizontal cold storage tank according to claim 4, characterized in that: Step S10 includes the following steps: S101: Obtain real-time power consumption information of data center IT equipment and aggregate it into equipment load time series according to a preset time window; obtain supply and return water temperature difference and flow information of chilled water system, and generate historical cooling load time series by combining the equipment load time series. S102: Decompose the historical cooling load time series into minute-level ultra-short-term components, hour-level short-term components, and day-level medium-term components. S103: Based on the Long Short-Term Memory network, corresponding prediction sub-models are constructed for the ultra-short-term component, short-term component and medium-term component respectively, and the outputs of the three sub-models are weighted and fused through the attention mechanism to obtain multi-timescale cold load prediction information.

6. The control method for a diaphragm-type horizontal cold storage tank according to claim 4, characterized in that: Step S20 includes the following steps: S201: The temperature information of the top and bottom of each cavity is obtained by temperature sensors arranged at the top and bottom of each cavity, and the real-time temperature gradient value of each cavity is calculated. The distributed fiber optic temperature measurement system acquires the temperature distribution at multiple points along the axial and radial directions on the sidewalls of each cavity, generates multi-point temperature information of the sidewalls of the cavities, and constructs a temperature field cloud map inside each cavity. S202: The differential pressure information of the water conveyance channel is obtained in real time by differential pressure sensors arranged at both ends of the water conveyance channel between each cavity. The real-time flow information of each channel is calculated based on the differential pressure information of the water conveyance channel and the pre-stored channel flow resistance characteristic curve. The differential pressure information of the flow equalization plate before and after each flow equalization plate is obtained in real time by differential pressure sensors arranged before and after each flow equalization plate. The real-time blockage degree of each flow equalization plate is evaluated based on the differential pressure information of the flow equalization plate before and after the flow equalization plate and the pre-stored flow equalization plate orifice characteristic curve.

7. The control method for a diaphragm-type horizontal cold storage tank according to claim 4, characterized in that: Step S30 includes the following steps: S301: Calculate the real-time effective cold storage capacity of each cavity based on the top temperature information and bottom temperature information, and calculate the slope temperature layer thickness value of each cavity based on the temperature field cloud map of each cavity. Compare the slope temperature layer thickness value of each cavity with the preset slope temperature layer stability threshold to determine the flow stability level of each cavity. S302: Calculate the predicted total cooling capacity required by the cold storage tank based on the cold load forecast information, and calculate the current effective cold storage capacity based on the temperature field cloud map of each cavity; when the current effective cold storage capacity is less than the predicted total cooling capacity required, it is determined that the cold storage tank cannot independently meet the cold load demand, and the refrigeration unit is triggered to supplement cooling to compensate for the cooling capacity gap; when the current effective cold storage capacity is less than the difference between the upper limit of the total cold storage capacity of the cold storage tank and the preset safety margin, and the current period is a low electricity price period, it is determined that the cold storage tank enters the charging mode to supplement the cold capacity reserve; when the current effective cold storage capacity is greater than the predicted total cooling capacity required and the current period is a high electricity price period, it is determined that the cold storage tank enters the releasing mode to release cold capacity to replace the operation of the refrigeration unit; when the current period is a flat electricity price period, with the goal of maintaining the current cold storage capacity not lower than the preset lower limit safety value, the mode switching is not triggered, and the current operating mode is maintained; S303: Based on the cold load prediction information and the thickness of the inclined temperature layer of each cavity, the target total inflow and outflow of the cold storage tank is determined by a multi-objective optimization algorithm, and the target opening of the water conveyance channel between each cavity is allocated in combination with the flow stability level of each cavity, so that the flow path of water in the cold storage tank is distributed according to the temperature stratification state of each cavity. The multi-objective optimization algorithm includes the following constraints: the target valve opening of the water conveyance channel between each cavity is not less than the minimum allowable opening corresponding to the current blockage level of the channel, and the minimum allowable opening is positively correlated with the blockage level of the flow equalization plate; and under the premise of meeting the cold load distribution requirements, the opening adjustment priority of the channel corresponding to the cavity with a high flow stability level is higher than that of the channel corresponding to the cavity with a low flow stability level. The multi-objective optimization algorithm takes minimizing the weighted sum of squares of the deviation of the oblique temperature layer thickness in each cavity as the first objective function, minimizing the sum of valve action frequencies in the water conveyance channels between all cavities as the second objective function, and minimizing the sum of flow direction reversal impact energy during the switching of the cooling mode in all cavities as the third objective function. It solves the Pareto optimal solution set through a non-dominated sorting genetic algorithm with an elite strategy, and selects an optimal compromise solution from the Pareto optimal solution set as the final scheduling scheme based on the current operating scenario.

8. The control method for a diaphragm-type horizontal cold storage tank according to claim 4, characterized in that: Step S40 is followed by step S50, which includes the following steps: S501: Real-time monitoring of the temperature change rate at the top and bottom of each cavity. When the temperature change rate of any cavity exceeds the preset rate threshold, it is determined that a flow instability event has occurred in that cavity. In response to a flow instability event, the main flow path of the cavity is cut off by closing the upstream water supply valve and reducing the opening of the downstream water supply valve, while only maintaining the bypass microcirculation flow rate. The flow recovery program of the cavity is then initiated. The flow recovery program maintains the water circulation in the cavity at a preset low flow rate until the thickness of the inclined thermosphere in the cavity recovers to a preset stable range. S502: Based on the real-time pressure difference value of the water conveyance channel between each cavity and the target valve opening command, generate the expected opening time information and expected closing time information of the valves in each channel, and generate the valve action sequence table corresponding to each channel according to the expected opening time information and expected closing time information of the valves in each channel. S503: Obtain the action characteristic parameter information of the water supply valves between each chamber. The action characteristic parameter information includes the full stroke time information of the valve, the pressure drop change curve during the valve opening process, and the water hammer pressure peak information during the valve closing process. Based on the pressure drop change curve and water hammer pressure peak information of each valve, calculate the superimposed pressure pulse peak information generated when any two valves operate simultaneously. With minimizing the maximum value of the superimposed pressure pulse peak information within the action time window of all valves as the optimization objective, and with the expected opening time information and expected closing time information of each valve as adjustable variables, perform peak-shifting optimization processing on the valve action timing table through a genetic algorithm to stagger the action times of each passage valve, thereby avoiding the superposition of pressure pulses generated by the simultaneous opening or closing of multiple passage valves.

9. A control method for a diaphragm-type horizontal cold storage tank according to claim 4, characterized in that: Before step S30, there is also a step S01, which includes the following steps: S011: Obtain real-time electricity price information and future time-of-use electricity price forecast information for grid connection; S012: Based on the cooling load forecast information and future time-of-use electricity price forecast information output by the multi-time-scale cooling load forecast model, calculate the charging cost and cooling revenue in each time window. With the goal of minimizing the total operating electricity cost of the cold storage tank, and combining the upper limit of the total cold storage capacity of the cold storage tank and the upper limit of the independent cold storage capacity and charging / discharging rate of each cavity, generate a charging / discharging economic scheduling plan across time windows through a dynamic programming algorithm. The charging / discharging economic scheduling plan includes the decision of whether the cold storage tank should be in charging mode or cooling mode in each time window. S013: Based on the historical operating data of each cavity, establish a charging and discharging cold efficiency decay model for each cavity. The charging and discharging cold efficiency decay model characterizes the degradation law of the equivalent cold storage capacity of the cavity after experiencing different flow rates and temperature cycles. S014: When generating the cold storage economic scheduling plan through dynamic programming algorithm, the cold storage efficiency decay model of each cavity is introduced as a penalty term into the objective function to balance the water flow frequency of each cavity and extend the overall service life of the cold storage tank.

10. A control system for a diaphragm-type horizontal cold storage tank, used to execute the control method for a diaphragm-type horizontal cold storage tank as described in any one of claims 4-9, characterized in that, include: The load forecasting module is used to acquire historical load information, historical chilled water supply and return temperature information, and historical ambient temperature and humidity information of the data center, and to build a multi-time-scale load forecasting model. The status sensing module is used to acquire the cavity status information of each independent cavity of the horizontal cold storage tank. The cavity status information includes the top temperature information, bottom temperature information, multi-point temperature information of the cavity sidewall, pressure difference information of the water conveying channel, and pressure difference information before and after the flow equalization plate. The scheduling optimization module is used to generate a cold storage tank charging and discharging scheduling strategy based on the cold load forecast information output by the multi-time scale load forecast model, combined with the cavity status information, and based on a pre-set multi-objective optimization algorithm. The execution decomposition module decomposes the charging and discharging cooling scheduling strategy into execution instruction information, which includes the opening and closing timing instructions for the main inlet valve and the main outlet valve of the cold storage tank, as well as the opening adjustment instructions for the water supply valves between each chamber.

Images