Temperature storage device based on chemical reversible thermal effect and control method of temperature storage device

Abstract

The present application relates to a temperature storage device based on chemical reversible thermal effect and a method for controlling the temperature storage device, which solves the problems of low energy storage density, large energy loss during storage and transmission, and difficulty in adapting to complex and changeable energy supply scenarios in traditional sensible heat energy storage devices. The device includes a chemical heat exchange energy storage module, which includes a first valve, a second valve, and a third valve. The chemical heat exchange energy storage module also includes a sealed container, a pressure control valve, an energy fiber, a heating module, a heat exchange coil, and a spray coil. The third valve is used to control the start and stop of the spray coil. The present application has the following effects: utilizing chemical reversible thermal effect to store temperature, combined with intelligent control and model optimization, to achieve efficient temperature storage and performance improvement.

Description

Technical Field

[0001] The present invention relates to the field of energy storage and management, and in particular to a temperature storage device based on a chemically reversible thermal effect and a control method for the temperature storage device. Background Art

[0002] In the urgent situation of global energy transformation and sustainable development, energy storage and management technology has become a key area. In particular, thermal storage technology plays an important role in balancing energy supply and demand and improving energy efficiency. Its importance is even more prominent in areas with concentrated energy consumption, such as buildings. For example, in buildings, the high energy consumption of air-conditioning systems during peak hours puts a heavy burden on the power grid. Therefore, efficient cold and heat storage technology has become an important means to relieve the pressure on the power grid and achieve reasonable energy allocation.

[0003] The technologies involved are rich and varied. Traditional temperature storage devices are mostly based on the principle of sensible heat storage, while in terms of cold storage technology, phase change material cold storage has emerged, and there are also various cold storage designs based on different structural improvements, such as setting up cold storage boxes, parallel cold storage branches, and using modular cooling.

[0004] However, these related technologies have many disadvantages. The energy storage density of traditional sensible heat energy storage devices is low, and the energy loss during storage and transmission is large, making it difficult to adapt to complex and changeable energy supply scenarios. Although phase change material cold storage has the advantage of high energy storage density, its thermal conductivity is low, and the energy exchange speed is difficult to meet peak demand. Existing improved cold storage designs, such as setting up cold storage boxes, will cause uneven refrigerant flow, parallel cold storage branches make the system complicated, modular cooling has the problem of cold loss, and most of them rely on additional units and have poor versatility. Summary of the invention

[0005] In order to utilize the chemical reversible thermal effect to store temperature, combined with intelligent management and control and model optimization, to achieve efficient temperature storage and performance improvement, the present application provides a temperature storage device based on the chemical reversible thermal effect and a control method for the temperature storage device.

[0006] In a first aspect, the present application provides a temperature storage device based on a chemically reversible thermal effect, which adopts the following technical solution:

[0007] A temperature storage device based on chemical reversible thermal effect, comprising a chemical heat exchange energy storage module, the chemical heat exchange energy storage module comprising a first valve and a second valve, and the chemical heat exchange energy storage module further comprising:

[0008] A sealed container for accommodating various components located in the chemical heat exchange energy storage module;

[0009] A pressure control valve connected to an external compressor for adjusting the pressure inside the chemical heat exchange energy storage module;

[0010] Energy fiber, composed of fiber materials with moisture absorption and thermal conductivity, used for the transfer of cold / heat;

[0011] A heating module, which is provided with a heating plate inside and is connected to an electric heater to provide required heat to the energy fiber;

[0012] The heat exchange coil is made of thermally conductive material and is used to absorb and take away the heat / cold released by the energy fiber, and the flow rate of the heat exchange medium and the start / stop state of the heat exchange process are controlled by adjusting the opening of the first valve and the second valve;

[0013] The spray coil is evenly distributed in the chemical heat exchange energy storage module and is equipped with a spray port. The medium water flows through the spray port and evenly sprinkles on the energy fiber to cause the energy fiber to absorb water and undergo a hydrolysis reaction, thereby absorbing or releasing heat;

[0014] The third valve is used to control the start and stop of the sprinkler coil.

[0015] By adopting the above technical solution, the temperature storage device utilizes the chemical reversible thermal effect, contains components through sealed containers, adjusts pressure with pressure control valves, transfers heat with energy fibers, supplies heat with heating modules, controls heat exchange with heat exchange coils, absorbs / releases heat by spray coils to induce hydrolysis reactions, and controls fluid start and stop with valves. All components work together to achieve heat storage and cold storage functions, adjust temperature, and realize energy storage and release.

[0016] Optionally, the chemical heat exchange energy storage module is a pipeline external module located inside the main pipeline, and only branch pipes are configured inside the pipeline external module.

[0017] By adopting the above technical solution, the chemical heat exchange energy storage module is set as an external module located inside the main pipeline and only equipped with branch pipes. This design is easy to install and maintain, and can realize the integration of the chemical heat exchange energy storage module without affecting the normal operation of the main pipeline. The configuration of the branch pipe can make the fluid distribution more flexible, facilitate the heat exchange and reaction control in the module, and improve the overall operation efficiency of the system.

[0018] Optionally, the chemical heat exchange energy storage module is a pipeline built-in module directly connected to the water supply and return system.

[0019] By adopting the above technical solution, the chemical heat exchange energy storage module is directly connected to the water supply and return system to become a pipeline built-in module, which can be more closely integrated into the system. This design can make full use of the water flow of the water supply and return system to achieve more efficient heat exchange and energy storage, and can achieve seamless connection in the existing system, reduce additional connection components, reduce costs and system complexity, and improve the thermal utilization efficiency of the overall system.

[0020] Optionally, the number of chemical heat exchange energy storage modules is at least 2, and different chemical heat exchange energy storage modules can be spliced ​​together.

[0021] By adopting the above technical solution, the number of chemical heat exchange energy storage modules is at least 2 and can be spliced, which can increase the energy storage capacity and meet greater energy storage and exchange needs. Multiple modules can be flexibly combined according to different situations, which improves the scalability and adaptability of the system, and is also conducive to optimizing system performance, ensuring stable storage and release of energy, and improving the reliability and practicality of the entire temperature storage device.

[0022] In a second aspect, the present application provides a control method for a temperature storage device based on a chemically reversible thermal effect, which adopts the following technical solution:

[0023] A control method for a temperature storage device based on a chemically reversible thermal effect includes:

[0024] Obtain the working mode set by the temperature storage device and the current status information of the energy fiber;

[0025] According to the mapping relationship between the working mode and the condition information of the energy fiber, the condition information of the energy fiber under the current working mode is determined;

[0026] Analyze whether the current status information of the energy fiber meets the condition information of the energy fiber in this working mode;

[0027] If not, stop the subsequent steps;

[0028] If yes, the control plan is determined and executed according to the correspondence between the working mode of the temperature storage device and the control plan.

[0029] By adopting the above technical solution, by obtaining the working mode and energy fiber status, the required conditions are clarified according to the mapping relationship to determine whether they are met. If they are met, the control plan can be accurately matched and executed. This ensures that the device operates in the appropriate state and in the best control mode, avoids invalid operations, improves operating efficiency, and ensures that the temperature storage device can stably and efficiently realize functions such as cold storage and heat storage.

[0030] Optionally, according to the correspondence between the working mode of the temperature storage device and the control plan, the control plan is determined, and the execution of the control plan includes:

[0031] Obtain the application scenarios of the temperature storage device, including the scenario of solar energy overcapacity, the scenario of chiller auxiliary, and the scenario of data center heat dissipation;

[0032] Based on the application scenario of the temperature storage device, the correspondence between the working mode of the temperature storage device and the control plan, determine this control plan and implement it.

[0033] By adopting the above technical solutions, the temperature storage device can be accurately controlled according to different application scenarios. In scenarios such as solar overcapacity, chiller assistance, and data room heat dissipation, the corresponding control solution is matched with the working mode. It can make full use of the characteristics of the device, meet the special needs of the scene, improve energy utilization efficiency, provide targeted temperature control strategies for different scenarios, ensure that the temperature storage device is adapted to the scene, and ensure stable operation of each scene.

[0034] Optionally, steps in parallel with the execution of the control plan are also included, as follows:

[0035] Obtain relevant data of the application scenario of the temperature storage device, operation data of the temperature storage device during the execution of the control plan, and relevant data of the execution results, and perform data preprocessing;

[0036] According to the correspondence between the application scenario of the temperature storage device and the pre-built parameter optimization model of the temperature storage device, the temperature storage device parameter optimization model that matches the application scenario of this temperature storage device is analyzed and determined, and the relevant data of the application scenario of this temperature storage device after data preprocessing, the operation data of the temperature storage device during the execution of the control plan, and the relevant data of the execution results are input into the temperature storage device parameter optimization model that matches the application scenario of this temperature storage device, and the optimized temperature storage device parameters are output.

[0037] By adopting the above technical solution, this parallel step can optimize the parameters of the temperature storage device in real time. By acquiring scene and operation data, with the help of a model combining a deep neural network and a recurrent neural network, accurate parameter optimization can be achieved for different scenarios. It can enable the device to always operate with the best parameters in each scenario, improve energy utilization efficiency, reduce resource waste, ensure stable operation of the device, and enhance its adaptability and functionality to complex and changing scenarios.

[0038] Optionally, the acquisition of a pre-built temperature storage device parameter optimization model includes:

[0039] Define the relevant data of the application scenarios of the heat storage device, the operation data of the heat storage device during the execution of the control plan, and the relevant data of the execution results as the comprehensive data of the heat storage device. The comprehensive data of the heat storage device after data processing is divided into a training set and a test set according to the preset proportion. During the training process, the heat storage device selects and executes a series of actions based on the action probability distribution output by the policy network based on the current state. The environment feeds back the reward value according to the preset reward rules based on the execution results of these actions and updates to the next state. The heat storage device stores the relevant experience in the experience recovery buffer, where the relevant experience includes state, action, reward, and next state;

[0040] When the buffer accumulates a preset number of records, batches of data are randomly extracted and the policy network parameters are updated using the proximal policy optimization algorithm;

[0041] After completing the parameter update, the experience replay buffer is cleared again, and then the temperature storage device continues to perform new actions in the environment based on the updated policy network, collects relevant experience of state, action, reward, and next state again, and stores it in the experience replay buffer;

[0042] Continue training for the set number of training rounds and simultaneously monitor the model's prediction accuracy on the training set;

[0043] When the prediction accuracy of the model on the training set exceeds the preset accuracy for three consecutive training rounds, it is determined that the training of the temperature storage device parameter optimization model has been completed.

[0044] By adopting the above technical solutions, dividing the training set and the test set, using the policy network to interact with the environment, accumulating experience and updating parameters, repeatedly training and monitoring accuracy, the model can accurately learn the optimal actions under different conditions, find the best operating parameters for the temperature storage device, improve its overall performance and stability, and achieve efficient temperature storage control.

[0045] Optionally, the optimized storage device parameters may include:

[0046] Analyzing whether the number of chemical heat exchange energy storage modules included in the temperature storage device exceeds one;

[0047] If not, maintain the original setting;

[0048] If yes, then obtain relevant information of different chemical heat exchange energy storage modules, including location distribution, received light, and different heat / cold sources close to them;

[0049] The relevant information of different chemical heat exchange energy storage modules, the relevant data of the application scenario of this temperature storage device after data preprocessing, the operation data of the temperature storage device during the execution of the control plan and the relevant data of the execution results are input into the pre-built multi-agent collaborative optimization model, and the parameters of different chemical heat exchange energy storage modules after optimization are output.

[0050] By adopting the above technical solution, the process can optimize the parameters of multi-module temperature storage devices in a targeted manner. First, determine the number of modules. If it is greater than one, collect information such as the location, illumination, heat source and cold source of each module, combine the scene and operation data, and use the multi-agent collaborative optimization model to output the optimization parameters. This enables each module to achieve differentiated and precise control according to its own conditions and system requirements, improve the overall temperature storage efficiency, reduce energy consumption, and enhance system stability and adaptability.

[0051] Optionally, the multi-agent collaborative optimization model is constructed as follows:

[0052] Define a temperature storage device containing multiple chemical heat exchange energy storage modules as an intelligent agent, collect the actual operation data of the intelligent agent in different application scenarios, different seasons, different time periods and various working conditions, and perform data preprocessing. The actual operation data includes the state attributes of each intelligent agent at each moment, the actions taken and the corresponding overall system performance indicators. The state attributes include the real-time temperature, pressure value, chemical concentration and working mode identification inside the module. The actions taken include the spray water flow rate, heating power value and valve opening and closing status. The overall system performance indicators include overall energy efficiency and temperature uniformity in each area.

[0053] Divide the preprocessed running data into a training set and a validation set according to a preset ratio;

[0054] In the training set, the agent is allowed to select and execute actions based on the current state according to the action probability distribution output by the proximal strategy optimization algorithm. After the agent completes the action, the environment will feedback the reward value according to the pre-set reward rules and update the current state;

[0055] After receiving the reward and state update information, the agent stores the relevant experience generated by this interaction, i.e. the current state, the action performed, the reward obtained, and the next state after the update, into the buffer. After accumulating a preset amount, it extracts batch data to update the neural network parameters and repeats the training continuously.

[0056] Regularly evaluate the model performance on the validation set, using a comprehensive quantitative evaluation method with validation adjustment to calculate the comprehensive quantitative score;

[0057] When the comprehensive quantization score exceeds the preset comprehensive quantization score or the number of training rounds reaches the preset number of training rounds, it is determined that the multi-agent collaborative optimization model has completed training.

[0058] By adopting the above technical solutions, the multi-scenario and multi-condition operation data of the multi-module temperature storage device are comprehensively collected. After preprocessing and dividing the training and verification sets, the agent is trained by the proximal strategy optimization algorithm to learn to optimize the decision. The model performance is guaranteed through comprehensive quantitative evaluation. The model can accurately grasp the complex operation rules of each module, optimize the multi-module coordination strategy, and improve the overall energy efficiency and temperature uniformity of the temperature storage device. BRIEF DESCRIPTION OF THE DRAWINGS

[0059] Figure 1 It is a schematic diagram of the overall structure of a chemical heat exchange energy storage module according to an embodiment of the present application.

[0060] Figure 2 It is a structural schematic diagram of the components included in the chemical heat exchange energy storage module of the embodiment of the present application.

[0061] Figure 3It is a schematic diagram of the overall structure of a chemical heat exchange energy storage module according to another embodiment of the present application.

[0062] Figure 4 It is a schematic diagram of the overall process of a control method of a temperature storage device based on a chemically reversible thermal effect according to another embodiment of the present application.

[0063] Figure 5 It is a flowchart of another embodiment of the present application that determines the control plan according to the correspondence between the working mode of the temperature storage device and the control plan, and executes the control plan.

[0064] In the figure, 1. pipeline external module; 2. sealed container; 3. pressure control valve; 4. energy fiber; 5. heating module; 6. heat exchange coil; 7. first valve; 8. second valve; 9. spray coil; 10. third valve; 11. pipeline internal module. DETAILED DESCRIPTION

[0065] The present application is further described in detail below in conjunction with the accompanying drawings.

[0066] Reference Figure 1 and Figure 2 , is a temperature storage device based on chemical reversible thermal effect disclosed in the present application, including a chemical heat exchange energy storage module, the chemical heat exchange energy storage module is a pipeline external module 1 located inside the main pipeline, and only a branch pipe is configured inside the pipeline external module 1.

[0067] The chemical heat exchange energy storage module consists of a first valve 7 and a second valve 8. The chemical heat exchange energy storage module also includes a sealed container 2, a pressure control valve 3, an energy fiber 4, a heating module 5, a heat exchange coil 6, a spray coil 9 and a third valve 10.

[0068] Specifically, the sealed container 2 is made of corrosion-resistant materials, including but not limited to stainless steel, aluminum alloy or high-density polymer materials, and is used to accommodate various components located in the chemical heat exchange energy storage module, ensuring that the module is highly sealed during operation and preventing the external environment from interfering with the internal chemical reaction.

[0069] The pressure control valve 3 is connected to an external compressor and is made of high-strength alloy materials, including but not limited to carbon steel, stainless steel, etc., and is used to adjust the pressure inside the chemical heat exchange energy storage module. When chemical regeneration is required, the pressure control valve 3 can adjust the inside of the chemical heat exchange energy storage module to a negative pressure state to promote the regeneration process of the chemical substances in the energy fiber 4.

[0070] The energy fiber 4 is made of fiber materials with high hygroscopicity and high thermal conductivity, including but not limited to polyamide fiber, polyester fiber, modified aramid fiber, ceramic fiber, polyimide fiber and mixed fiber, coated with specific chemical substances, and has excellent cold / heat transfer capabilities.

[0071] The heating module 5 is provided with a silica gel heating plate inside and is externally connected to a high-efficiency electric heater, which is used to provide necessary heat to the energy fiber 4, drive the regeneration reaction of the chemical substances, and ensure the efficient conduction of the heat storage process.

[0072] The heat exchange coil 6 is made of a high thermal conductivity material and is used to absorb and carry away the heat / cold released by the energy fiber 4. By adjusting the opening of the first valve 7 and the second valve 8, the flow rate of the heat exchange medium and the start and stop state of the heat exchange process are controlled.

[0073] The spray coil 9 is evenly distributed in the chemical heat exchange energy storage module, and is made of corrosion-resistant materials, including but not limited to stainless steel, PTFE-coated pipes, etc., and is equipped with a spray port. The medium water flows through the spray port and evenly sprinkles on the energy fiber 4, prompting the energy fiber 4 to undergo a hydrolysis reaction after absorbing water, thereby absorbing or releasing heat; the third valve 10 is used to control the start and stop of the spray coil 9 to ensure accurate regulation of the spray process. The diameter of the spray port is determined according to the specific application scenario and design requirements.

[0074] Reference Figure 3 In another embodiment, for the modification of the existing system, due to the limitation of the existing pipelines, the chemical heat exchange energy storage module is a pipeline built-in module 11 directly connected to the water supply and return system, and the rest of the components contained in the chemical heat exchange energy storage module remain unchanged.

[0075] In order to meet the high energy demand per unit time in certain usage scenarios, a single module is difficult to provide sufficient energy output. The number of chemical heat exchange energy storage modules is at least 2, and different chemical heat exchange energy storage modules can be spliced ​​together. For example, the system is composed of 6 chemical heat exchange energy storage modules connected to form an overall cold storage / heat storage solution. The modular structure not only simplifies the installation and maintenance of the system, but also allows the number and size of modules included in the system to be flexibly adjusted according to the specific design cooling requirements, thereby achieving customized energy storage capabilities.

[0076] Reference Figure 4 The present application also discloses a control method of a temperature storage device based on a chemically reversible thermal effect, which specifically comprises the following steps:

[0077] Step S100, obtaining the working mode set by the temperature storage device and the current status information of the energy fiber 4.

[0078] Due to the different chemical substances contained in the energy fiber 4, the working modes that can be set in the temperature storage device will also be different. The working mode refers to the operating mode set by the temperature storage device to achieve different functions, such as cold storage mode, heat storage mode, regeneration mode, cooling mode, heat release mode, etc. In different modes, the working states and mutual cooperation modes of the components of the device are different to achieve the corresponding temperature storage or release functions.

[0079] Taking the cold storage mode as an example, the energy fiber 4 corresponding to the cold storage mode includes but is not limited to ammonium nitrate (NH4NO3), ammonium chloride (NH4Cl), and potassium chloride (KCl).

[0080] The corresponding chemical reactions are as follows: Ammonium nitrate (NH4NO3): NH4NO3(s)+H2O(l)→NH4+(aq)+NO3−(aq); Ammonium chloride (NH4Cl): NH4Cl(s)+H2O(l)→NH4+(aq)+Cl−(aq); Potassium chloride (KCl): KCl(s)+H2O(l)→K+(aq)+Cl−(aq); When these substances are dissolved in water, they absorb a large amount of heat and lower the solution temperature, thereby achieving an efficient cold storage effect.

[0081] Taking the heat storage mode as an example, the energy fiber 4 corresponding to the heat storage mode includes but is not limited to sodium hydroxide (NaOH), potassium hydroxide (KOH), and lithium chloride (LiCl).

[0082] The corresponding chemical reactions are as follows: sodium hydroxide (NaOH): NaOH(s)+H2O(l)→Na+(aq)+OH−(aq); potassium hydroxide (KOH): KOH(s)+H2O(l)→K+(aq)+OH−(aq); lithium chloride (LiCl): LiCl(s)+H2O(l)→Li+(aq)+Cl−(aq). When these substances are dissolved in water, they release a large amount of heat, raising the temperature of the solution, thereby achieving an efficient heat storage effect.

[0083] Taking the regeneration mode as an example, the reaction chemical formula of the regeneration mode is as follows: NH4NO3(aq)→NH4NO3(s)+H2O(g); NH4Cl(aq)→NH4Cl(s)+H2O(g); KCl(aq)→KCl(s)+H2O(g); NaOH(aq)→NaOH(s)+H2O(g); KOH(aq)→KOH(s)+H2O(g); LiCl(aq)→LiCl(s)+H2O(g).

[0084] The temperature storage device can be set with an operation interface, and the operator selects the working mode through the interface. The device records the selected mode information and transmits it to the control unit through the internal communication line for acquisition. For example, in an industrial control scenario, the staff can select "cold storage mode" on the central console, and the system sends this mode information to the control method execution module.

[0085] The status information of the energy fiber 4 includes the current physical state and chemical state of the energy fiber 4. For example, the moisture absorption degree of the fiber, the reaction progress of the chemical substance, whether there is damage, and other factors that affect its heat transfer and chemical reaction efficiency.

[0086] Step S200 , determining the condition information of the energy fiber 4 in the current working mode according to the mapping relationship between the working mode and the condition information of the energy fiber 4 .

[0087] Mapping relationship: can be understood as a pre-set corresponding rule, which associates different working modes with specific condition information that the energy fiber 4 needs to meet.

[0088] The mapping relationship may be established in a manner involving experimental data accumulation or theoretical analysis and simulation. Regarding experimental data accumulation, various states of the energy fiber 4 are tested and recorded in different working modes through a large number of experimental tests.

[0089] Regarding theoretical analysis and simulation, with the help of theories of chemical kinetics, heat transfer and other related disciplines, the reaction and heat transfer process of energy fiber 4 under different working modes are analyzed and simulated. By establishing a mathematical model, the performance of energy fiber 4 under different conditions is predicted, so as to determine the ideal condition information of energy fiber 4 under different working modes and establish a mapping relationship.

[0090] Condition information of energy fiber 4 in this working mode: refers to the various state conditions that energy fiber 4 needs to have in order to efficiently and stably cooperate to realize the function of this mode in the currently selected working mode. For example, in the cold storage mode, energy fiber 4 may need specific initial water content, specific concentration of chemical substances and other conditions to ensure that the hydrolysis reaction can fully absorb heat and achieve the purpose of efficient cold storage.

[0091] Step S300, analyzing whether the current status information of the energy fiber 4 meets the condition information of the energy fiber 4 in this working mode. If not, execute step S400; if yes, execute step S500.

[0092] This step is to compare the actual status information of the current energy fiber 4 obtained in step S100 with the condition information that the energy fiber 4 should meet in the current working mode determined in step S200. Through this comparison, it is determined whether the energy fiber 4 can effectively support the temperature storage device to operate in the set working mode in the current state.

[0093] The specific analysis content and methods are as follows:

[0094] Analysis of moisture absorption: If the current working mode is the cold storage mode, the initial moisture content of the energy fiber 4 is required to be between 10% and 15%. The actual moisture content of the energy fiber 4 obtained by the humidity sensor is 12%, which is within the specified range, and the condition is met. If the actual moisture content is 8%, which is lower than the required range, the condition is not met.

[0095] Chemical substance state analysis: Assuming that in the heat storage mode, the concentration of the sodium hydroxide chemical substance coated on the surface of the energy fiber 4 needs to be maintained above 90%. The actual concentration is 92% through a chemical analysis sensor or related detection means, which meets the condition; if the test result is 85%, it does not meet the condition.

[0096] Fiber physical state analysis: For example, regardless of the working mode, the energy fiber 4 is required to be intact. The energy fiber 4 is scanned and imaged using an image sensor. If the image shows that the fiber is not broken or damaged, the condition is met; if obvious cracks are found, the condition is not met.

[0097] The result judgment example is as follows: If in a certain working mode, all key status information such as the moisture absorption degree, chemical substance state and physical state of the energy fiber 4 meet the corresponding condition requirements, then it is judged as "yes", that is, the current status information of the energy fiber 4 meets the condition information of the energy fiber 4 in this working mode, and the temperature storage device can continue the subsequent operation steps. If any of the key information does not meet the conditions, such as the chemical substance concentration does not meet the standard in the above example, it is judged as "no", and the subsequent steps need to be stopped at this time to avoid the device from operating under unsuitable conditions, which may cause performance degradation, failure or even safety problems.

[0098] Step S400, stop the subsequent steps.

[0099] Step S500, determining the current control plan according to the correspondence between the working mode of the temperature storage device and the control plan, and executing the current control plan.

[0100] The correspondence between the working mode of the temperature storage device and the control scheme is a pre-set relationship, which links different working modes of the temperature storage device (such as cold storage mode, heat storage mode, regeneration mode, cooling mode, heat release mode, etc.) with the corresponding control scheme. The control scheme includes the operation and adjustment of various components in the device (such as the opening and closing of the valve, the power adjustment of the heating module 5, the pressure adjustment of the pressure control valve 3, etc.) to ensure that the temperature storage device can operate stably according to the expected function.

[0101] Management and control plan: It is a collection of specific operations and control instructions, which aims to make the temperature storage device achieve optimal performance in the corresponding working mode. It involves the coordinated operation of various components in the device to complete functions such as energy storage, release or conversion.

[0102] Taking the cold storage mode as an example, the corresponding control scheme in the cold storage mode is as follows: the system first starts the spray coil 9 by opening the third valve 10. At this time, the cooling water is evenly sprinkled on the surface of the energy fiber 4 through the spray port at a precisely controlled flow rate. The energy fiber 4 is made of a fiber material with high hygroscopicity and high thermal conductivity, and the surface is coated with specific chemical substances, which undergo a diffusion endothermic reaction during the hydrolysis process. When water comes into contact with the chemical substances, the chemical reaction absorbs a large amount of ambient heat, thereby significantly reducing the temperature around the module and achieving efficient cold storage. In order to ensure that the hydrolysis reaction is carried out under optimal conditions, the system adjusts the internal pressure of the module in real time through the pressure control valve 3. The pressure control valve 3 adopts a high-precision sensor and an automatic adjustment mechanism, which can dynamically adjust the internal pressure according to the heat changes generated during the reaction process to maintain the stability and efficiency of the reaction system. In addition, the entire cold storage process is monitored by an intelligent control system, and temperature and pressure data are fed back in real time to ensure that the reaction is carried out within a predetermined parameter range to avoid overcooling or abnormal system pressure. To improve the reliability of the system, the cold storage mode can be equipped with additional multiple safety protection measures, including overpressure protection, automatic water shut-off device for over-temperature and emergency shutdown valve to prevent system failure or safety hazards caused by unexpected situations.

[0103] Taking the heat storage mode as an example, the corresponding control scheme in the heat storage mode is as follows: In the heat storage mode, the system first starts the spray coil 9 by opening the third valve 10. At this time, the medium water is evenly sprinkled on the surface of the energy fiber 4 through the spray port at a precisely controlled flow rate. The energy fiber 4 is made of fiber materials with high hygroscopicity and high thermal conductivity, and the surface is coated with specific chemical substances. These chemical substances will react exothermically after contacting with water. Specifically, when water comes into contact with chemical substances, the chemical reaction releases a large amount of heat, which significantly increases the temperature around the module and realizes effective storage of heat.

[0104] Taking the regeneration mode as an example, the control scheme corresponding to the regeneration mode is as follows: In the regeneration mode, ensure that the third valve 10 is closed to ensure the airtightness of the module and maintain the internal pressure. Subsequently, the heating module 5 is started to quickly increase the temperature of the energy fiber 4 through the high-efficiency electric heater, so that the chemical substances on its surface undergo a reverse reaction, separate from the water molecules, and return to their original state. In order to ensure that the reaction is carried out under optimal conditions, the pressure control valve 3 adjusts the internal pressure of the module in real time to maintain the stability and efficiency of the reaction. The regeneration mode not only improves the energy efficiency ratio of the system and extends the service life of the energy fiber 4, but is also suitable for application scenarios that require frequent energy cycles, such as renewable energy systems and industrial environmental control, to ensure the long-term stable operation of the module.

[0105] Taking the cooling mode as an example, the corresponding control scheme in the cooling mode is as follows: by opening the first valve 7 and the second valve 8, the heat exchange medium water at room temperature is allowed to enter from the first valve 7, flow through the heat exchange coil 6, and absorb the cold released by the energy fiber 4. The water temperature in the coil is reduced and finally flows out from the second valve 8, thereby realizing the release of cold.

[0106] Taking the heat release mode as an example, the corresponding control scheme in the heat release mode is as follows: in the heat release mode, the system opens the first valve 7 and the second valve 8, allowing the heat exchange medium water at room temperature to enter from the first valve 7, flow through the heat exchange coil 6, and absorb the cold released by the energy fiber 4. The water temperature in the coil rises and finally flows out from the second valve 8, thereby realizing the release of heat.

[0107] Reference Figure 5 According to the corresponding relationship between the working mode of the temperature storage device and the control plan, the control plan is determined and implemented, including:

[0108] Step S510, obtaining scenarios in which the temperature storage device is applied, wherein the scenarios include solar energy overcapacity scenarios, chiller unit auxiliary scenarios, and data center heat dissipation scenarios.

[0109] Solar energy overcapacity scenario: When the energy collected by the solar energy collection device exceeds the current usage demand, the excess energy needs to be stored. The thermal storage device plays the role of storing excess solar energy in this scenario. For example, when the sun is abundant during the day, if the electricity generated by the solar panels cannot be completely consumed, the thermal storage device can be used to store this excess energy in the form of heat or cold for subsequent use, such as releasing the stored energy at night or on cloudy days.

[0110] Chiller auxiliary scenario: In this scenario, the temperature storage device assists the chiller. The chiller is mainly used for refrigeration. The temperature storage device can store the cold energy generated by the chiller when the chiller has a high refrigeration efficiency and release the cold energy when needed, thus playing a role in balancing the refrigeration demand and improving energy efficiency. For example, in the air-conditioning system of a commercial building, the chiller can be used for efficient refrigeration at night, and the cold energy can be stored in the temperature storage device. The cold energy can be released during peak hours during the day, reducing the workload of the chiller.

[0111] Data room cooling scenario: Data rooms generate a lot of heat due to the operation of servers and other equipment, which requires heat dissipation. The temperature storage device can store the excess heat generated by servers, etc., or provide cooling function when needed, helping to maintain the temperature of the room within an appropriate range and ensure the normal operation of the equipment.

[0112] The application scenario of the temperature storage device can be obtained by using a variety of sensors and monitoring equipment to determine the scenario in which the temperature storage device is located.

[0113] In a solar energy overcapacity scenario, a power monitor can be used to monitor the power output of the solar energy collection device. If the output power exceeds the load consumption power for a long time, it can be determined as a solar energy overcapacity scenario.

[0114] For the chiller auxiliary scenario, the operating status of the chiller can be monitored, such as the cooling capacity, the temperature and flow rate of the cooling medium and other parameters. When the cooling capacity reaches a certain threshold and the demand is not high, this scenario is determined.

[0115] In the data room heat dissipation scenario management and control, temperature sensors can be used to monitor the temperature inside the data room. When the room temperature is lower than a certain set value, a temperature storage device may be needed to store heat; when the temperature is higher than the set value, a temperature storage device may be needed to assist in heat dissipation. At the same time, the server load and the working status of the cooling system can be monitored to comprehensively judge the scenario.

[0116] Step S520, determine the current control plan according to the application scenario of the temperature storage device, the correspondence between the working mode of the temperature storage device and the control plan, and execute the current control plan.

[0117] First, the application scenario of the heat storage device obtained in step S510 and the currently set working mode are combined. For example, if the solar energy production capacity is in excess and the working mode is heat storage.

[0118] Then, a search is performed in the pre-built correspondence library. This correspondence library may be a table or algorithm model stored in the controller. Taking the solar energy overcapacity and heat storage mode as an example, the following control scheme may be found: turn on the heating module 5 and adjust the heating power to 80% of the rated power; open the first valve 7 to 50% of the opening, and close the second valve 8; open the third valve 10 and control the spray flow of the spray coil 9 to 5L / min.

[0119] This correspondence is established based on a large number of tests and simulations of device performance in different scenarios and working modes. For example, in the heat storage mode of the solar energy overcapacity scenario, it was found through experiments that when the power of the heating module 5 is 80% of the rated power, the excess solar energy can be effectively used for efficient heat storage, and energy waste or equipment overheating will not be caused by excessive power.

[0120] In the chiller auxiliary scenario and when the working mode is cold storage, it is assumed that the determined control plan is: turn off the heating module 5; open the first valve 7 to 30% opening, open the second valve 8 to 70% opening; open the third valve 10, and set the spray flow rate to 3L / min.

[0121] During the execution process, the controller will immediately cut off the power supply of the heating module 5 to stop heating; by controlling the valve actuator, the opening of the first valve 7 is adjusted to 30%, and the opening of the second valve 8 is adjusted to 70% to ensure that the flow of the heat exchange medium is appropriate; at the same time, the third valve 10 is opened, and the spray flow is stabilized at 3L / min through the flow regulating device, so that the energy fiber 4 can fully absorb heat to achieve efficient cold storage.

[0122] A control method for a temperature storage device based on a chemically reversible thermal effect also includes steps in parallel with executing the control scheme, specifically as follows:

[0123] Step SA00, obtaining relevant data of the application scenario of the temperature storage device, the operation data of the temperature storage device during the execution of the control plan and the relevant data of the execution result, and performing data preprocessing.

[0124] Data related to the application scenarios of the thermal storage device: refers to data that can reflect the characteristics of the actual environment in which the thermal storage device is located. For example, in the scenario of overcapacity of solar energy, data such as the real-time power of solar panels, sunshine intensity, and ambient temperature can reflect the characteristics of energy supply in this scenario. In the data room cooling scenario, the equipment power, room space size, initial temperature, etc. of the data room are all scene-related data, which determine the heat conditions that the thermal storage device needs to deal with in this scenario.

[0125] The acquisition of scene-related data can be obtained through various sensors. In the scenario of solar energy overcapacity, power sensors are used to measure the output power of solar panels, light sensors to obtain sunlight intensity, and temperature sensors to monitor ambient temperature. In the data center heat dissipation scenario, the power data of the server is obtained through the equipment management system, and the temperature of different locations in the room is measured using space temperature sensors.

[0126] Operation data of the temperature storage device during the execution of the control plan: during the execution of the control plan, the real-time working status data of each component of the temperature storage device itself. For example, the actual heating power of the heating module 5, the real-time opening of the valve, the temperature change of the energy fiber 4, etc. These data can intuitively show the operation status of each part of the temperature storage device when executing the control plan.

[0127] The operation data can be obtained by the sensors and monitoring equipment of the temperature storage device itself, such as the power sensor of the heating module 5 feedbacks its heating power, the valve position sensor monitors the valve opening in real time, and the temperature sensor measures the temperature of the energy fiber 4.

[0128] The execution result data can be obtained through calculation and measurement. For example, by measuring the temperature difference and flow rate at the inlet and outlet of the heat exchange coil 6, the energy storage or release amount can be calculated in combination with time; by comparing the set temperature range with the actual temperature range, it can be judged whether the expected temperature control effect is achieved.

[0129] Data related to execution results: data used to evaluate the effectiveness of the implementation of the control plan. For example, the amount of energy stored in the temperature storage device over a period of time, the efficiency of releasing cold / heat, whether the expected temperature control range is achieved, etc. These data reflect whether the control plan has achieved the expected goals.

[0130] Data preprocessing: Clean and convert the acquired data to make it meet the requirements of subsequent data analysis and model input. For example, remove noise and outliers in the data, standardize or normalize the data, and unify data in different formats.

[0131] Step SB00, according to the correspondence between the application scenario of the temperature storage device and the pre-built temperature storage device parameter optimization model, analyze and determine the temperature storage device parameter optimization model that matches the application scenario of this temperature storage device, and input the relevant data of the application scenario of this temperature storage device after data preprocessing, the operation data of the temperature storage device during the execution of the control plan and the relevant data of the execution result into the temperature storage device parameter optimization model that matches the application scenario of this temperature storage device, and output the optimized temperature storage device parameters, wherein the architecture of the temperature storage device parameter optimization model adopts an architecture that combines a deep neural network with a recurrent neural network.

[0132] Correspondence between the application scenarios of thermal storage devices and the pre-built thermal storage device parameter optimization model: This is a mapping relationship based on a large number of experiments and theoretical studies, which shows that different application scenarios (such as solar energy overcapacity, chiller assistance, data center cooling, etc.) correspond to specific thermal storage device parameter optimization models. Each scenario has its own unique energy characteristics and requirements for thermal storage devices. Through this correspondence, the most suitable optimization model can be matched for each scenario.

[0133] Architecture combining deep neural network and recurrent neural network: Deep neural network consists of multiple hidden layers, has powerful feature extraction capabilities, and can learn complex patterns in data. Recurrent neural network is particularly suitable for processing time series data. It can remember previous input information to better analyze and predict data that changes over time. The architecture combining the two can make full use of the spatial and temporal information in the data, conduct a comprehensive and in-depth analysis of the operating data of the temperature storage device, and obtain more accurate optimization parameters.

[0134] The process of data input and parameter output is as follows: the relevant data after data preprocessing in step SA00, i.e., the relevant data of the application scenario of the temperature storage device (such as the equipment power and space temperature of the data room), the operation data of the temperature storage device during the execution of the control plan (such as the actual power of the heating module 5, the valve opening, etc.) and the relevant data of the execution result (such as whether the room temperature is effectively controlled within the specified range), are input into the determined temperature storage device parameter optimization model (such as model A) in the format and order required by the model. The deep neural network and the recurrent neural network inside the model begin to work together. The deep neural network extracts various features from the data, and the recurrent neural network analyzes and processes the features related to the time series. Through a series of complex calculations and iterations, the model finally outputs a set of optimized temperature storage device parameters. For example, for the data room cooling scenario, the model may output that the optimal power of the heating module 5 is X watts, the opening of a certain valve should be adjusted to Y%, and the spray flow rate of the spray coil 9 is Z liters / minute, etc. These parameters will help improve the operating efficiency and effect of the temperature storage device in the data room cooling scenario.

[0135] The acquisition of the pre-built temperature storage device parameter optimization model includes:

[0136] Step Sa00, defines the relevant data of the application scenarios of the heat storage device, the operating data of the heat storage device during the execution of the control plan, and the relevant data of the execution results as the comprehensive data of the heat storage device, and divides the comprehensive data of the heat storage device after data processing into a training set and a test set according to a preset proportion. During the training process, the heat storage device selects and executes a series of actions based on the action probability distribution output by the strategy network based on the current state. The environment feeds back the reward value according to the preset reward rules based on the execution results of these actions, and updates to the next state. The heat storage device stores the relevant experience in the experience recovery buffer, where the relevant experience includes state, action, reward, and next state.

[0137] Preset ratio: refers to the pre-set ratio for dividing data into training set and test set. For example, the common 8:2 or 7:3 means that the processed data is divided into training set and test set according to the ratio of 80% to 20% or 70% to 30%, respectively, so as to balance the amount of data for model training and the amount of data for evaluating model performance.

[0138] Policy network: A neural network-based model that calculates the probability distribution of performing various actions based on the current state of the temperature storage device. For example, when the temperature storage device is at a certain temperature and pressure state, the policy network can calculate the probability of opening the spray coil 9, adjusting the valve opening, and other different actions.

[0139] Action probability distribution: represents the probability of the temperature storage device performing each possible action under a given state. For example, under a certain state, the probability of opening the spray coil 9 is 0.6, the probability of closing the heating module 5 is 0.4, and the sum of these probabilities is 1.

[0140] Environment: refers to the external physical environment in which the thermal storage device is located and the systems that interact with it, such as solar energy generation systems, chiller systems, or data center environments. The environment will feedback corresponding results based on the actions performed by the thermal storage device.

[0141] Reward value: It is the feedback value given by the environment based on the result of the action performed by the temperature storage device. If the action makes the temperature storage device closer to the target state, such as in the data center cooling scenario, making the room temperature closer to the appropriate range, the reward value may be positive; conversely, if the action causes deviation from the target state, the reward value is negative.

[0142] Experience recovery buffer: It is a data storage container used to store the relevant experience generated by the temperature storage device during its interaction with the environment, including the four-tuple of state, action, reward, and next state, to provide data support for subsequent model training.

[0143] The data division process is as follows: Assume that we have 1,000 processed data with a preset ratio of 8:2. Then the first 800 data are divided into a training set for training the policy network to let the model learn the best action selection under different states; the last 200 data are divided into a test set to evaluate the performance of the trained model on unseen data and judge the generalization ability of the model.

[0144] The action selection and feedback process is as follows: The heat storage device is in a certain initial state, for example, the temperature is 30°C and the pressure is 1 standard atmosphere. At this time, the policy network calculates the action probability distribution based on this state. Assume that the probability of opening the spray coil 9 is 0.7 and the probability of keeping the power of the heating module 5 unchanged is 0.3. The heat storage device selects to perform an action based on this probability distribution, such as opening the spray coil 9. The environment responds to this action. For example, in the scenario of solar energy overcapacity, after opening the spray coil 9, the heat storage device absorbs heat, the temperature drops, and the pressure also changes. The environment feedbacks a reward value based on this result. Assume that because the temperature is closer to the target heat storage temperature, the reward value is +5. The heat storage device enters a new state, that is, the temperature becomes 28°C and the pressure becomes 0.95 standard atmospheres. This state, the executed action (opening the spray coil 9), the reward value (+5) and the new state (28°C, 0.95 standard atmospheres) are stored in the experience recovery buffer.

[0145] Step Sb00, when the buffer accumulates a preset number of records, batch data is randomly extracted and the policy network parameters are updated using the proximal policy optimization algorithm.

[0146] Preset number of records: This is the number of experience records that need to be accumulated in the experience recovery buffer. For example, if it is set to 1,000, when the four-tuple experience records of state, action, reward, and next state in the experience recovery buffer reach 1,000, the subsequent operation is triggered. This number needs to be set to ensure that there is enough data for model training and updating, but not to cause low training efficiency due to excessive data volume.

[0147] Randomly extract batch data: randomly select a certain amount of data from the experience recovery buffer. For example, from a buffer that has accumulated to a preset number (such as 1,000), randomly extract 32 or 64 data as a batch. This is done to simulate different training scenarios, avoid the model's over-reliance on certain specific experience data, and improve the model's generalization ability and stability.

[0148] Proximal Policy Optimization (PPO): This is an algorithm used to optimize policy networks. It limits the step size of each update during the policy network parameter update process so that the policy network will not be too aggressive when updating parameters, thereby ensuring the stability of the training process. While ensuring policy improvement, the PPO algorithm reduces the risk of poor model performance due to excessive parameter updates. The algorithm is based on the policy gradient method and optimizes the parameters of the policy network by maximizing the cumulative reward.

[0149] Policy network parameters: The policy network consists of a series of weights and biases, which are the parameters of the policy network. For example, in a simple neural network structure, the connection weights between neurons and the bias value of each neuron. These parameters determine how the policy network calculates the action probability distribution based on the input state of the heat storage device. By continuously updating these parameters, the policy network can output a better action probability distribution, allowing the heat storage device to make more appropriate action choices under different states.

[0150] Step Sc00, after completing the parameter update, clear the experience replay buffer again, and then let the temperature storage device continue to perform new actions in the environment based on the updated strategy network, collect relevant experience of state, action, reward, and next state again, and store it in the experience replay buffer.

[0151] Clear the experience replay buffer: Delete all the state, action, reward, and next state quadruple data previously stored in the experience replay buffer to prepare for a new round of data collection. This is to ensure that the experience data used in each training is based on the latest policy network and to avoid interference of old data on model training.

[0152] The operation flow of step Sc00 is as follows:

[0153] Buffer clearing: Once step Sb00 completes the update of the policy network parameters, the system will immediately trigger the clearing mechanism of the experience playback buffer. In the actual programming implementation, if Python language is used and the experience data is stored in the form of a list, the list of stored experience data can be simply cleared through the list_variable.clear() statement, quickly and effectively clearing the historical data in the buffer and preparing for a new round of data collection.

[0154] State perception and action decision: The temperature storage device uses its own various sensors (such as temperature sensors, pressure sensors, humidity sensors, etc.) to perceive the current state information in real time, including but not limited to the temperature and pressure values ​​inside the device, the humidity of the energy fiber 4, and the working status of each component. These perceived state information is passed as input to the updated strategy network. The strategy network conducts a deep analysis of the input state based on its internal calculation logic and updated parameters, and calculates the probability distribution of the temperature storage device performing various possible actions (such as adjusting the valve opening, starting or stopping the heating module 5, controlling the spray flow of the spray coil 9, etc.) in the current state. For example, in a scenario where solar energy production capacity is in excess and heat storage is required, if the current device temperature is low and the humidity of the energy fiber 4 is appropriate, the strategy network may output that the action of turning on the heating module 5 and adjusting the power to 80% has a higher probability of execution after calculation.

[0155] Action execution and environmental feedback: The temperature storage device selects the action with higher probability to execute according to the action probability distribution output by the strategy network. Continuing with the above solar thermal storage scenario as an example, the temperature storage device executes the action of turning on the heating module 5 and setting the power to 80%. At this time, the environment in which the temperature storage device is located (such as the solar energy system, the heat exchange network connected to it, etc.) will respond to this action. As the heating module 5 runs, the temperature inside the temperature storage device gradually rises, and the pressure may also change accordingly, and the heat exchange process with the surrounding environment will also be affected. According to these changes, the environment gives the temperature storage device a corresponding reward value based on a pre-set reward mechanism. If the heating operation makes the device temperature closer to the target heat storage temperature, which effectively promotes the storage efficiency of solar energy, the environment may give a positive reward value, such as +3; conversely, if the operation causes the temperature to be out of control or deviate from the target state, a negative reward value may be given.

[0156] Experience collection and storage: After executing an action and obtaining environmental feedback, the temperature storage device organizes the current state before executing the action, the actual action executed, the reward value obtained, and the new state generated after the action is executed (i.e., the next state) into a four-tuple. For example, (current temperature 32°C, pressure 1.05atm, energy fiber 4 humidity 80%, turn on heating module 5 and set power 80%, reward value +3, next state temperature 34°C, pressure 1.08atm, energy fiber 4 humidity 78%). Subsequently, this four-tuple experience data is stored in the experience playback buffer to provide new data samples for subsequent model training, continuously enrich training data resources, and help the continuous optimization of the policy network.

[0157] Step Sd00, continue training according to the set number of training rounds, and simultaneously monitor the prediction accuracy of the model on the training set.

[0158] Set number of training rounds: This is the predetermined number of cycles for training the temperature storage device parameter optimization model.

[0159] The training process is as follows: Starting from the first round of training, the model takes the state information in the training set as input, calculates the action probability distribution through the policy network, and selects actions accordingly. For example, in the data center cooling scenario, the model predicts actions such as adjusting the flow of the sprinkler coil 9 or adjusting the valve opening based on the current state information such as the room temperature and equipment load.

[0160] Based on these actions, simulate the operation of the temperature storage device in the environment to obtain the corresponding reward value and the next state. Then store these experience data (state, action, reward, next state) in the experience playback buffer. When the buffer reaches the preset number of records, randomly extract batch data in accordance with step Sb00, and use the proximal policy optimization algorithm to update the policy network parameters. After completing one round of training, enter the next round of training, repeat the above process, and continuously update the policy network parameters so that the model can learn more deeply about the data features and rules in the training set. For example, in the 100th round of training, the model may have a better understanding of the common temperature change patterns in the cooling scenarios of the data room, and can more accurately predict the appropriate actions.

[0161] The monitoring method is as follows: During each round of training, the prediction accuracy of the model on the training set is monitored in real time. This can be achieved by adding the corresponding evaluation function to the training code. For example, when using a deep learning framework (such as PyTorch or TensorFlow) in Python, you can write a function that compares the actions predicted by the model with the actions actually recorded in the training set and calculates the matching ratio between the two, that is, the prediction accuracy. After each round of training is completed, the prediction accuracy of that round is recorded to form a curve showing the change of accuracy with the number of training rounds. By observing this curve, you can intuitively understand the training effect of the model. If the prediction accuracy increases steadily with the increase of the number of training rounds, it means that the model is learning effectively; if the accuracy fluctuates or stagnates, it may be necessary to adjust the training parameters or check the data quality.

[0162] Step Se00: When the prediction accuracy of the model on the training set exceeds the preset accuracy for three consecutive training rounds, it is determined that the parameter optimization model of the temperature storage device has completed training.

[0163] The judgment process is as follows: During the process of continuous training and monitoring of prediction accuracy in step Sd00, the system will continuously compare the prediction accuracy after each round of training with the preset accuracy. For example, after the 150th round of training, the model prediction accuracy is 88%, which does not reach the preset 90%, and the next round of training will continue. When the prediction accuracy after a round of training exceeds the preset accuracy for the first time, the system starts counting. Assume that after the 180th round of training, the accuracy reaches 92%, exceeding the preset 90%, and the count is 1 at this time.

[0164] Then the 181st round of training is carried out. If the prediction accuracy of this round still exceeds the preset accuracy, for example, reaching 93%, the count is increased to 2.

[0165] Then conduct the 182nd round of training. If the prediction accuracy still exceeds the preset accuracy, such as 91%, and the prediction accuracy of the model on the training set exceeds the preset accuracy for three consecutive training rounds, the system determines that the temperature storage device parameter optimization model has completed training.

[0166] The parameters of the temperature storage device after output optimization include:

[0167] Step SB10, analyzing whether the number of chemical heat exchange energy storage modules included in the temperature storage device exceeds one. If not, executing step SB20; if yes, executing step SB30.

[0168] If, after analysis and judgment, it is determined that the number of chemical heat exchange energy storage modules included in the temperature storage device does not exceed one (i.e., there is only one module), the system will execute step SB20, i.e., maintain the original settings. This means that in the case of only one module at present, there is no need to perform additional optimization and adjustment on the module parameters, and maintaining the current operating parameters and configuration can meet the needs.

[0169] If it is determined that the number of chemical heat exchange energy storage modules exceeds one, it means that the operation of the temperature storage device is more complicated, and different modules may have mutual influences, and targeted optimization is required according to the specific conditions of each module. At this time, the system will execute step SB30 and enter the process of obtaining relevant information of different chemical heat exchange energy storage modules to prepare for subsequent parameter optimization.

[0170] Step SB20, maintain the original settings.

[0171] Step SB30, obtaining relevant information of different chemical heat exchange energy storage modules, including location distribution, received light, and different heat sources / cold sources nearby.

[0172] Position distribution: refers to the specific layout of multiple chemical heat exchange energy storage modules in the overall structure of the temperature storage device. For example, in a large temperature storage system, different chemical heat exchange energy storage modules may be distributed on different floors, in different areas, or arranged in a specific geometric shape. Position distribution affects the heat conduction and fluid flow between modules, and thus affects the performance of the entire temperature storage device.

[0173] The theoretical position information of each module can be obtained through the design drawings or three-dimensional models of the temperature storage device. After the actual installation is completed, the actual position of each module can be accurately measured and calibrated using positioning sensors (such as indoor positioning systems, RFID-based positioning technology, etc.). In addition, the position distribution of each module can also be clearly determined by visually scanning the internal structure of the temperature storage device (such as industrial CT scanning).

[0174] Received light: For temperature storage devices used in solar energy-related scenarios, the intensity and duration of sunlight received by chemical heat exchange energy storage modules at different locations will vary. Light conditions directly affect the module's ability to use solar energy for heat storage. Modules that receive more light may be able to convert solar energy into heat energy more efficiently and store it.

[0175] The received light can be obtained by installing a light sensor on or near the surface of each chemical heat exchange energy storage module to measure the light intensity received by the module in real time. At the same time, by combining the time information and the geographical location information of the device and calculating the trajectory of the sun, the length of time each module receives light in different time periods can be calculated. In addition, the use of satellite remote sensing data or geographic information system (GIS) data can also assist in analyzing the light receiving conditions of different modules.

[0176] Close heat source / cold source: The heat source can be waste heat discharge equipment in the industrial production process, heating furnace, etc., and the cold source can be a chiller, natural cold source (such as groundwater, cold air, etc.). The relative distance and position relationship between different chemical heat exchange energy storage modules and the heat source or cold source determines the heat exchange efficiency and method between the module and them. Modules close to the heat source are more likely to absorb heat for heat storage, while modules close to the cold source are more conducive to storing cold or releasing heat.

[0177] For heat sources, the ambient temperature can be measured by temperature sensors, and the location and distance of the heat source can be determined based on the temperature gradient. For cold sources, if they are equipment such as chillers, their location information can be obtained through the control system connected to the cold source equipment; if they are natural cold sources, such as groundwater, their relative position relationship with each module of the temperature storage device can be determined through geological exploration data and water flow monitoring equipment.

[0178] Step SB40, input the relevant information of different chemical heat exchange energy storage modules, the relevant data of the application scenario of this temperature storage device after data preprocessing, the operation data of the temperature storage device during the execution of the control plan and the relevant data of the execution results into the pre-built multi-agent collaborative optimization model, and output the parameters of different chemical heat exchange energy storage modules after optimization.

[0179] Multi-agent collaborative optimization model: an advanced intelligent algorithm model that simulates the collaboration and interaction between multiple agents. In this scenario, each chemical heat exchange energy storage module can be regarded as an agent. By learning and analyzing a large amount of data, the model can find the optimal strategy for the collaborative work between multiple modules, thereby optimizing the operating parameters of each module to achieve efficient operation of the entire temperature storage device. For example, the model will consider factors such as the location, light, heat source / cold source of different modules, coordinate the heating, heat exchange and other operations of each module, so that the temperature storage device can achieve optimal performance in different scenarios.

[0180] After receiving the data, the multi-agent collaborative optimization model uses its internal complex algorithms and structures to perform calculations. The agents in the model simulate the behavior and decision-making process of different chemical heat exchange energy storage modules, and find the optimal parameter combination through information interaction and collaboration. For example, the model may adjust the power, valve opening and other parameters of each module heating module 5 through multiple iterative calculations to achieve the goals of maximizing the overall storage efficiency and minimizing energy consumption.

[0181] After the model is run, the parameters of different chemical heat exchange energy storage modules are finally output after optimization. These parameters are obtained based on a comprehensive analysis of the input data, which enables each module to work in coordination with other modules in its specific working environment to achieve performance optimization of the entire temperature storage device. For example, for modules that are close to the heat source and receive more light, the model may output optimization parameters such as increasing the power of the heating module 5 and increasing the flow rate of the heat exchange coil 6 to make full use of the heat source and solar energy to improve the temperature storage effect.

[0182] The construction of the multi-agent collaborative optimization model is as follows:

[0183] Step SB41, defines a temperature storage device including multiple chemical heat exchange energy storage modules as an intelligent entity, collects actual operation data of the intelligent entity in different application scenarios, different seasons, different time periods and various working conditions, and performs data preprocessing, wherein the actual operation data includes the state attributes of each intelligent entity at each moment, the actions taken and the corresponding overall system performance indicators, wherein the state attributes include the real-time temperature, pressure value, chemical substance concentration and working mode identification inside the module, the actions taken include the spray water flow rate, heating power value and valve opening and closing status, and the overall system performance indicators include overall energy efficiency and temperature uniformity of each area.

[0184] Among them, application scenarios include: the use environment of the temperature storage device, such as the scenario of excess solar energy production capacity, which requires the storage of excess solar energy; the scenario of auxiliary chillers, which assists the chillers in balancing the cooling capacity; the scenario of heat dissipation in data centers, which controls the temperature of the data centers. Different scenarios have different demands and operation requirements for temperature storage devices.

[0185] Working conditions: The working conditions of the temperature storage device during operation, including full load operation, partial load operation, startup phase, stable operation phase, etc. Under different working conditions, the working status and performance of each component of the device are different.

[0186] Agent: In this scenario, each chemical heat exchange energy storage module is regarded as an agent that can perceive its own state, receive environmental information, and make decisions and perform actions according to the strategy.

[0187] State attributes: Parameters used to describe the real-time state of the intelligent body (chemical heat exchange energy storage module). The real-time temperature inside the module reflects its thermal energy storage; the pressure value affects the chemical reaction and energy transfer; the concentration of chemical substances determines the rate and efficiency of chemical reactions; the working mode indicator indicates which working mode the module is currently in, such as heat storage, cold storage, heat release, or cold release.

[0188] Actions taken: The actions performed by the agent to achieve the temperature storage target. The spray water flow rate controls the hydrolysis reaction rate and heat absorption / release; the heating power value adjusts the temperature in the module; the valve opening and closing state determines the flow path and flow rate of the heat exchange medium, affecting the heat exchange.

[0189] Overall system performance index: a parameter that measures the operating effect of the entire multi-module temperature storage device. The overall energy efficiency reflects the efficiency of the device in utilizing energy; the temperature uniformity of each area reflects the degree of balance of temperature distribution in different areas when the device adjusts the temperature, which is very important for some scenarios with high requirements for temperature consistency (such as data centers).

[0190] Step SB42, dividing the preprocessed operating data into a training set and a validation set according to a preset ratio.

[0191] Preprocessed operation data: refers to the actual operation data of the multi-module temperature storage device in different scenes, seasons, time periods and working conditions after data cleaning, standardization, format conversion and other processing in step SB41. These data have removed noise, unified scale and format, and are more suitable for model training and evaluation.

[0192] Preset ratio: It is the ratio relationship for dividing the training set and the validation set that is preset before building the multi-agent collaborative optimization model. Common ratios include 7:3, 8:2, etc. For example, if the ratio is 8:2, it means that 80% of the preprocessed data is divided into the training set and 20% of the data is divided into the validation set. The setting of this ratio needs to comprehensively consider factors such as the amount of data, model complexity, and training effect to ensure that the model has enough data for training and learning, and has suitable data for verifying model performance.

[0193] Training set: This part of data is used to train the multi-agent collaborative optimization model. During the training process, the model continuously adjusts its own parameters and strategies by learning data such as state attributes, actions taken, and overall system performance indicators in the training set to optimize the decision-making ability of the multi-module temperature storage device operation. For example, the model learns through the training set how different modules can select the best spray water flow, heating power, and valve opening and closing state according to their own state and environmental information in the scenario of overcapacity of solar energy to improve overall energy efficiency.

[0194] Validation set: used to evaluate the performance of the model during the training process. During the model training process, the validation set is used to test the model every certain number of training rounds. The data in the validation set does not participate in the parameter update process of the model, and it can independently test the generalization ability of the model on new data. For example, after the model has been trained on the training set for multiple rounds, the validation set is used to verify whether the model can accurately predict the operating status and performance indicators of the multi-module temperature storage device under different working conditions, and whether the model has problems such as overfitting.

[0195] Step SB43, in the training set, let the intelligent agent select and execute actions based on the current state according to the action probability distribution output by the proximal strategy optimization algorithm. After the intelligent agent completes the action, the environment will feedback the reward value according to the pre-set reward rules and update the current state. After receiving the reward and state update information, the intelligent agent stores the relevant experience generated by this interaction, that is, the current state, the action performed, the reward obtained, and the next state after the update, into the buffer. After accumulating a preset amount, batch data is extracted to update the neural network parameters, and training is repeated continuously.

[0196] Proximal Policy Optimization (PPO): It is an optimization algorithm based on policy gradients, which is used to adjust the policy network of the agent. It guides the agent to learn the optimal strategy by maximizing the cumulative reward. In the scenario of a multi-module temperature storage device, the PPO algorithm calculates the probability distribution of each possible action (such as the flow rate of spray water, the value of heating power, and the opening and closing state of the valve) based on the current state of the agent (i.e., the chemical heat exchange energy storage module), such as the real-time temperature and pressure value inside the module, the concentration of chemical substances, etc. In this way, the agent can choose an action to execute based on this probability distribution, rather than blindly trying all actions, thereby improving learning efficiency.

[0197] The reward rules can be set as follows: a reward mechanism based on system performance indicators: The pre-set reward rules are the key to guiding the agent to learn the optimal strategy. The reward value is mainly determined based on the overall performance indicators of the system, including overall energy efficiency, temperature uniformity of each area, etc. For example, if the overall energy efficiency of the system is significantly improved after the agent performs an action, such as consuming the same amount of electricity but storing more available heat during a heat storage process, the environment gives the agent a positive reward, assuming the reward value is +5. This indicates that the agent's action has a positive contribution to the system performance. On the contrary, if the overall energy efficiency decreases after the action is performed, such as excessive energy consumption during the heating process but the heat storage effect is not effectively improved, the environment will give a negative reward, such as -3. For the temperature uniformity of each area, if the standard deviation of the temperature in different areas becomes smaller after the action is performed, it means that the temperature uniformity has improved, and the agent will also receive a corresponding positive reward. For example, when cooling a large building, the temperature difference between different floors is large. After the action is performed, the temperature of each floor is closer, and a reward of +4 can be given at this time. On the contrary, if the temperature uniformity deteriorates, a negative reward is given. In addition, the reward rules can also take other factors into consideration, such as equipment loss, operational stability, etc. If the equipment runs more stably after the action is executed, and unnecessary switching times or pressure fluctuations are reduced, appropriate rewards can also be given.

[0198] The impact of the environment on the state of the agent: After the agent completes the execution of the action, the environment will update the state of the agent according to the action. For example, after the agent performs the action of increasing the flow rate of the spray water, the temperature inside the chemical heat exchange energy storage module will decrease due to the heat absorption of the spray water, the pressure value may change due to the injection of the liquid and the change of the chemical reaction, and the concentration of the chemical substance will also change as the reaction proceeds. At the same time, the working mode identification may be adjusted according to the overall operation of the system. Assuming that it was originally in the cold storage mode, when the temperature reaches a certain target value, the working mode identification may be updated to the standby mode. These state update information will be fed back to the agent, and the agent will select the next action through the policy network again based on the updated state and continue to interact with the environment. This continuous state update and interaction process enables the agent to continuously adapt to environmental changes and learn the optimal behavior strategy.

[0199] The role of the experience replay buffer: After receiving the reward and state update information, the agent will store the relevant experience generated by this interaction, that is, the current state, the action performed, the reward obtained, and the next state after the update, into the buffer, which is also called the experience replay buffer. Its role is to store the experience of the agent during the training process so that data can be extracted later for neural network parameter update. For example, the state of the agent at a certain moment is (temperature: 30℃, pressure: 1.2MPa, chemical concentration: 0.8mol / L, working mode identification: cold storage mode), and the action performed is to increase the spray water flow (flow rate increases from 5L / min to 8L / min), and obtain a reward value of +3. The next state after the update is (temperature: 28℃, pressure: 1.22MPa, chemical concentration: 0.78mol / L, working mode identification: cold storage mode), and this information will be stored in the buffer. When the buffer accumulates to a preset amount, such as 1,000 experiences, a batch of data, such as 64 experiences, will be randomly extracted from the buffer. The benefit of this is that it breaks the time correlation between data, allowing the neural network to more effectively learn the relationship between different states and actions. Through these extracted data, the neural network parameters are updated using technologies such as the back propagation algorithm. This training process is repeated continuously. As the number of training rounds increases, the neural network parameters are continuously optimized, and the strategy of the intelligent agent will become closer and closer to the optimal strategy, thereby achieving efficient control of the temperature storage device.

[0200] Step SB44, regularly evaluate the model performance on the validation set, use the validation-adjusted comprehensive quantitative evaluation method to calculate the comprehensive quantitative score.

[0201] The comprehensive quantitative evaluation method uses a comprehensive consideration of multiple indicators. This method comprehensively considers multiple indicators related to the performance of the multi-module temperature storage device. For example, overall energy efficiency is a key indicator that reflects the energy utilization efficiency of the temperature storage device in the process of storing and releasing energy. It is measured by calculating the ratio of the input energy of the device to the output effective energy. The higher the ratio, the higher the energy efficiency. The temperature uniformity of each area is also an important indicator. For some application scenarios with high requirements for temperature consistency (such as data centers), temperature uniformity directly affects the normal operation and service life of the equipment. The temperature uniformity can be quantified by calculating the standard deviation or coefficient of variation of the temperature in different areas. The smaller the value, the more uniform the temperature.

[0202] Indicator weight allocation: In order to more accurately reflect the effect of the model on the overall performance improvement of the multi-module temperature storage device, different weights need to be assigned to different indicators. The determination of weights is usually based on actual application requirements and the relative importance of each indicator. For example, in the data center cooling scenario, due to the extremely high requirements for temperature uniformity, a higher weight, such as 0.6, may be assigned to the temperature uniformity indicator; while overall energy efficiency is also important, but the weight can be set to 0.4 relatively speaking. In this way, when calculating the comprehensive quantitative score, the impact of temperature uniformity on the final result will be greater.

[0203] Quantitative calculation method: Use a certain mathematical formula to comprehensively calculate each indicator and its weight. Assume that the comprehensive quantitative score is represented by S, the overall energy efficiency index is E, the weight is w1; the temperature uniformity index of each area is T, and the weight is w2. Then the calculation formula for the comprehensive quantitative score can be S=w1×E+w2×T. Through this calculation method, multiple performance indicators are integrated into a comprehensive quantitative score, which is convenient and intuitive to evaluate the model performance.

[0204] Step SB45, when the comprehensive quantization score exceeds the preset comprehensive quantization score or the number of training rounds reaches the preset number of training rounds, it is determined that the multi-agent collaborative optimization model has completed training.

[0205] Preset comprehensive quantitative score: Before model training begins, a comprehensive quantitative score threshold that is expected to be achieved is pre-set based on the actual application requirements and performance goals of the multi-module thermal storage device. This threshold is the standard for judging whether the model has achieved the ideal performance state. For example, if the goal is to achieve a high energy utilization efficiency and highly uniform temperature in each area of ​​the data center in the cooling scenario of the multi-module thermal storage device, the preset comprehensive quantitative score set according to these requirements may be 0.9.

[0206] During the model training process, evaluation is performed regularly on the validation set according to step SB44, and a comprehensive quantitative score is obtained for each evaluation. For example, after the 50th round of training, the comprehensive quantitative score S1 = 0.85 is calculated by evaluation on the validation set; after the 60th round of training, the comprehensive quantitative score S2 = 0.88 is obtained.

[0207] Compare the comprehensive quantization score obtained each time with the preset comprehensive quantization score. Assuming that the preset comprehensive quantization score is set to 0.9, when the calculated comprehensive quantization score exceeds 0.9 for the first time, for example, after the 70th round of training, the comprehensive quantization score reaches 0.92, then the model has met the conditions for completing training.

[0208] Furthermore, when the scenario is a data center heat dissipation scenario, a control method based on chemical reversible thermal effect may have the following steps:

[0209] Step Sf00, collects server load, temperature and operating status data information of the temperature storage device, and performs data preprocessing.

[0210] The server-related data involves the following: 1. Load: With the help of professional management software, the CPU and memory usage rates are collected to two decimal places per second; the disk controller is monitored and the disk I / O rate is collected with an accuracy of 0.01MB / s. 2. Temperature: Micro high-precision sensors are installed in key heat-generating parts and in the chassis, collecting data 5 times per second with an accuracy of ±0.1℃.

[0211] The data of the temperature storage device are as follows: 1. Energy fiber 4: collect temperature and pressure near it 10 times per second, with an accuracy of ±0.1℃ and ±0.01MPa respectively; combine the humidity sensor and the law of conservation of mass to judge the state of the hydrolysis reaction 5 times per second. 2. Equipment parameters: use an electromagnetic flowmeter to measure the spray water flow 5 times per second, with an accuracy of ±0.1L / min; the power sensor collects the power of the heating module 5 5 times per second, with an accuracy of ±0.1kW; the magnetostrictive valve position sensor obtains the valve opening 10 times per second.

[0212] Step Sg00, inputting the real-time collected operation data of the temperature storage device into a pre-built working mode classification model based on a decision tree, and outputting the corresponding working mode code.

[0213] The construction of the decision tree model involves the following processes: 1. Data preparation: Collect a large amount of operating data of the temperature storage device under various working conditions in the past data room cooling scene, covering the temperature, pressure, hydrolysis reaction state of the energy fiber 4, the water flow of the spray coil 9, the power of the heating module 5, the valve opening and other information. At the same time, clarify the working mode corresponding to each working condition, such as full cooling mode, auxiliary cooling mode, etc. 2. Feature selection: Use methods such as information gain ratio to screen out features that are closely related to the working mode. For example, it is found that features such as the temperature change rate of the energy fiber 4 and the working frequency of the spray coil 9 are of high value for judging the working mode. 3. Model training: Use classic decision tree algorithms such as ID3 and C4.5, and use the prepared data as the training set to build a decision tree model. During the training process, the structure of the tree is continuously optimized to improve the classification accuracy of the model for the working mode.

[0214] Step Sh00, integrates the working mode code with the server load, temperature, and operating status of the temperature storage device, and uses a fusion network based on a multi-head self-attention mechanism to fuse multi-source data.

[0215] The fusion of multi-source data involves the following processes: 1. Construction of a fusion network based on a multi-head self-attention mechanism: A fusion network based on a multi-head self-attention mechanism is constructed, which contains multiple self-attention heads (such as 8). Each self-attention head is responsible for paying attention to and analyzing the association of each element in the comprehensive data vector from different angles. 2. Attention calculation and data fusion: Taking one of the self-attention heads as an example, for each element in the comprehensive data vector (such as CPU usage), the similarity score between it and all other elements is calculated (such as using dot product operations), and normalized by the Softmax function to obtain the attention weight of the element and other elements. For example, the attention weights of the CPU usage and the temperature of the energy fiber 4 are calculated to be high, indicating that the two data are closely related in the current state. The attention weights of all elements are multiplied and accumulated by their corresponding values ​​to obtain the output of the self-attention head. The outputs of multiple self-attention heads are spliced ​​together to complete the fusion of multi-source data, providing a more in-depth and relevant data expression for subsequent analysis, and helping to make accurate decisions.

[0216] Step Si00, analyzing whether the situation information of the data center meets the preset complex situation condition information.

[0217] The preset complex condition conditions can be as follows: 1. Server load dimension: Set the server CPU usage rate to exceed 80% for 5 minutes, or the memory usage rate to be higher than 90% for 10 minutes, and the disk I / O rate frequently encounters read and write bottlenecks (read and write rates are lower than 50% of the normal average) in a short period of time (such as 3 minutes), as a complex condition condition in terms of load. 2. Server temperature dimension: If the temperature of key parts of the server (such as CPU core, GPU chip) rises by more than 5°C within 10 minutes, and the ambient temperature of multiple areas of the computer room (such as more than 3 different monitoring points) rises to more than 30°C at the same time, it is considered a temperature-related complex condition. 3. Temperature storage device dimension: When the temperature of the energy fiber 4 of the temperature storage device is close to its safe working upper limit (such as reaching 90% of the upper limit), and the pressure fluctuates abnormally (the pressure changes by more than ±0.05MPa within 1 minute), and the spray coil 9 fails (such as unstable flow and deviates from the set value by more than 20%), it is listed as a complex condition of the temperature storage device.

[0218] Data extraction and comparison: Extract relevant data on server load, temperature and operating status of the heat storage device from the integrated data vector fused in step Sh00. Compare these real-time data with the preset complex conditions one by one.

[0219] Logical judgment output: Through the written judgment logic program (such as conditional judgment statements using Python language), comprehensive judgment is made on the data of each dimension. If any one or more complex condition conditions in the above multiple dimensions are met, it is determined that the situation information of the data center meets the preset complex conditions, providing a basis for the subsequent initiation of more complex decision-making exploration processes.

[0220] Step Sj00, if yes, then the A algorithm is combined with the Monte Carlo tree search. For each temperature storage device control action node and server-related action node, the heuristic function value is calculated.

[0221] Specific combination process: Taking the control decision of the heat storage device and the server as the search target, first use the current state of the data center as the root node of the Monte Carlo tree. Algorithm A calculates the heuristic value of each possible action (such as adjusting the heating power of the heat storage device, scheduling server tasks, etc.) starting from the root node according to the pre-set heuristic function, sorts the actions according to the size of the heuristic value, and prioritizes the action with a high heuristic value as the expansion direction of the Monte Carlo tree search, and expands the node expansion and simulation evaluation of the Monte Carlo tree in this direction.

[0222] The calculation of the heuristic function value involves the following process: 1. Temperature storage device control action node: For temperature storage device control actions, such as adjusting the power of the heating module 5, the heuristic function comprehensively considers the expected impact of power adjustment on the server temperature, the change in energy consumption, and the effect on the stability of the temperature storage device itself. For example, increasing the heating power may reduce the server temperature, but it will increase energy consumption and affect the pressure stability of the temperature storage device. By setting weights for each factor (such as the server temperature impact weight is set to 0.5, the energy consumption weight is set to 0.3, and the temperature storage device stability weight is set to 0.2), the heuristic function value of the action node is calculated. Assuming that after a certain power increase, the expected server temperature reduction effect score is 8 points (out of 10 points), the energy consumption increase score is 4 points, and the temperature storage device stability score is 6 points, then the heuristic function value is. 2. Server-related action nodes: For server-related actions, such as server task scheduling, the heuristic function considers the improvement of server load balancing after task redistribution, the impact on the overall service response time, and the effect on energy consumption. Similarly, weights are assigned to each factor (such as the load balancing weight is set to 0.4, the service response time weight is set to 0.4, and the energy consumption weight is set to 0.2), and the heuristic function value is calculated. For example, if a task scheduling scheme gives a load balancing score of 7 points, a service response time score of 6 points, and an energy consumption score of 5 points, then the heuristic function value is. By calculating the heuristic function value of each node, a basis is provided for the subsequent node expansion and decision evaluation of the Monte Carlo tree search.

[0223] Step Sk00, using the finite element analysis method to establish a heat transfer model between the energy fiber 4 and the server, dividing the energy fiber 4 and the server into small units, calculating the heat transfer rate between units based on the law of conservation of energy and Fourier's law of heat conduction, and simulating the temperature changes of key parts of the server. At the same time, the energy consumption of the temperature storage device is calculated based on the equipment power and operating time.

[0224] Unit division: Finite element analysis is used to divide the energy fiber 4 and the server into tiny units, and the geometric shape, size, thermal conductivity, specific heat capacity and other physical properties of each unit are clarified. For example, the energy fiber 4 is divided into regular rectangular units according to its shape to ensure that the unit size is small to accurately simulate heat transfer.

[0225] Heat transfer calculation: Based on the law of conservation of energy and Fourier's law of heat conduction, the heat transfer rate of adjacent units is calculated. The heat transfer rate of adjacent units is proportional to their temperature difference, contact area, and thermal conductivity. For example, for units A and B, the temperatures are (T_A) and (T_B), respectively, the contact area is (S), the thermal conductivity is (k), and the distance between the centers of mass is (\Deltax), then the heat transfer rate is (q=kS\frac{T_A-T_B}{\Deltax}). By iteratively calculating the heat transfer rate of each adjacent unit, the temperature changes of key parts of the server (such as CPU cores and GPU chips) are simulated.

[0226] Energy consumption of temperature storage device: Calculate the energy consumption of temperature storage device according to the power and operation time of the equipment. The energy consumption of heating module 5 is the product of its power and operation time. For example, if the power is 5kW and it runs for half an hour, the energy consumption is 2.5kWh. The energy consumption of spray coil 9 should take into account the power of water pump, operation time, and specific heat capacity of water. Assuming that the water pump power is 1kW and runs for 20 minutes, the specific heat capacity of water is (4.2×10^3J / (kg·℃)), the mass of spray water is 100kg, and the temperature changes by 10℃, first calculate the heat of water heating according to (Q=mc\DeltaT), then convert it into electrical energy, add the power consumption of water pump, and get the total energy consumption.

[0227] Comprehensive energy consumption consideration: Summarize the energy consumption of each part of the temperature storage device to obtain the overall energy consumption under different control actions. Combined with the server temperature change data of the heat transfer simulation, it provides a key basis for evaluating the comprehensive impact of decision-making actions on the data center.

[0228] Step S100 uses a deep learning algorithm combining a long short-term memory network (LSTM) and a gated recurrent unit (GRU) to take historical server load, temperature, and thermal storage device operation data as input, learn the complex nonlinear relationship between the data, and predict the future trends of server load distribution, temperature changes, and energy consumption under different action combinations (including thermal storage device control actions and server-related actions). Based on the simulation results, a score is evaluated for each action node.

[0229] Action combination prediction: Combine different temperature storage device control actions (such as adjusting heating power, valve opening, etc.) and server-related actions (such as task scheduling, adjusting power supply strategy), together with the current data center status data (obtained from step Sh00) as input, and input them into the trained model. The model predicts the future trends of server load distribution, temperature changes, and energy consumption under these action combinations. For example, it predicts the changes in server CPU usage, key component temperatures, and overall energy consumption in the next 10 minutes after increasing the heating power of the temperature storage device and adjusting the server task allocation.

[0230] Action node evaluation: Based on the prediction results of the model, each action node is evaluated and scored. The score takes into account multiple factors, such as server temperature stability (the smaller the temperature fluctuation, the higher the score), energy consumption rationality (the closer the energy consumption is to the expected target, the higher the score), server load balance (the more balanced the load, the higher the score). Set weights for each factor (such as temperature stability weight 0.4, energy consumption weight 0.3, load balance weight 0.3), and obtain the final score of each action node through weighted calculation, providing a quantitative basis for subsequent selection of the optimal decision.

[0231] Step Sm00, after the Monte Carlo tree search is completed, the action combination with the highest score is selected as the optimal decision. Through the Modbus protocol, the intelligent control interface is used to send instructions to the temperature storage device and the server.

[0232] After the Monte Carlo tree search is completed, a series of action nodes will be generated, each of which corresponds to an action combination and its simulation evaluation results. From these nodes, the action combination with the highest score is selected, which is the optimal decision for the complex situation of the current data center. For example, after searching and evaluating, it is found that the action combination of "increasing the heating power of the temperature storage device by 20% and migrating some high-load tasks to the backup server group" performs best in ensuring stable server temperature, reasonable energy consumption, and load balancing, and has the highest score, so it is determined as the optimal decision.

[0233] The embodiments of this specific implementation method are all preferred embodiments of the present application, and are not intended to limit the protection scope of the present application. Therefore, all equivalent changes made based on the structure, shape, and principle of the present application should be included in the protection scope of the present application.

Claims

1. A control method for a temperature storage device based on a chemically reversible thermal effect, characterized in that: The temperature storage device comprises a chemical heat exchange energy storage module, the chemical heat exchange energy storage module comprises a first valve (7), a second valve (8), and the chemical heat exchange energy storage module further comprises: A sealed container (2) for accommodating various components in the chemical heat exchange energy storage module; A pressure control valve (3), connected to an external compressor, for adjusting the pressure inside the chemical heat exchange energy storage module; Energy fiber (4), composed of fiber material with moisture absorption and thermal conductivity, used for transferring cold / heat; A heating module (5) having a heating plate disposed therein and connected to an electric heater for providing required heat to the energy fiber (4); The heat exchange coil (6) is made of a heat conductive material and is used to absorb and carry away the heat / cold released by the energy fiber (4), and to control the flow rate of the heat exchange medium and the start / stop state of the heat exchange process by adjusting the opening of the first valve (7) and the second valve (8); The spray coil (9) is evenly distributed in the chemical heat exchange energy storage module and is equipped with a spray port. The medium water flows through the spray port and evenly falls on the energy fiber (4), prompting the energy fiber (4) to undergo a hydrolysis reaction after absorbing water, thereby absorbing or releasing heat; A third valve (10) is used to control the start and stop of the spray coil (9); Control methods include: Obtaining the working mode set by the temperature storage device and the current status information of the energy fiber (4); Determining the condition information of the energy fiber (4) under the current working mode according to the mapping relationship between the working mode and the condition information of the energy fiber (4); Analyzing whether the current status information of the energy fiber (4) meets the condition information of the energy fiber (4) in the current working mode; If not, stop the subsequent steps; If yes, determine the current control plan according to the correspondence between the working mode of the temperature storage device and the control plan, and execute the current control plan; According to the corresponding relationship between the working mode of the temperature storage device and the control plan, the control plan is determined, and the implementation of the control plan includes: Obtain the application scenarios of the temperature storage device, including the scenario of solar energy overcapacity, the scenario of chiller auxiliary, and the scenario of data center heat dissipation; Determine the current control plan based on the application scenario of the temperature storage device, the corresponding relationship between the working mode of the temperature storage device and the control plan, and implement the current control plan; It also includes steps that run in parallel with the implementation of the control plan, as follows: Obtain relevant data of the application scenario of the temperature storage device, operation data of the temperature storage device during the execution of the control plan, and relevant data of the execution results, and perform data preprocessing; According to the correspondence between the application scenario of the temperature storage device and the pre-built parameter optimization model of the temperature storage device, the temperature storage device parameter optimization model that matches the application scenario of this temperature storage device is analyzed and determined, and the relevant data of the application scenario of this temperature storage device after data preprocessing, the operation data of the temperature storage device during the execution of the control plan, and the relevant data of the execution results are input into the temperature storage device parameter optimization model that matches the application scenario of this temperature storage device, and the optimized temperature storage device parameters are output.

2. A control method for a temperature storage device based on a chemically reversible thermal effect according to claim 1, characterized in that: The chemical heat exchange energy storage module is a pipeline external module (1) located inside the main pipeline, and only branch pipes are arranged inside the pipeline external module (1).

3. The control method of a temperature storage device based on chemical reversible thermal effect according to claim 1, characterized in that: The chemical heat exchange energy storage module is a pipeline built-in module (11) directly connected to the water supply and return system.

4. A control method for a temperature storage device based on a chemically reversible thermal effect according to claim 2 or 3, characterized in that: The number of chemical heat exchange energy storage modules is at least 2, and different chemical heat exchange energy storage modules can be spliced ​​together.

5. A control method for a temperature storage device based on a chemically reversible thermal effect according to claim 4, characterized in that: The acquisition of the pre-built temperature storage device parameter optimization model includes: Define the relevant data of the application scenarios of the heat storage device, the operation data of the heat storage device during the execution of the control plan, and the relevant data of the execution results as the comprehensive data of the heat storage device. The comprehensive data of the heat storage device after data processing is divided into a training set and a test set according to the preset proportion. During the training process, the heat storage device selects and executes a series of actions based on the action probability distribution output by the policy network based on the current state. The environment feeds back the reward value according to the preset reward rules based on the execution results of these actions and updates to the next state. The heat storage device stores the relevant experience in the experience recovery buffer, where the relevant experience includes state, action, reward, and next state; When the buffer accumulates a preset number of records, batches of data are randomly extracted and the policy network parameters are updated using the proximal policy optimization algorithm; After completing the parameter update, the experience replay buffer is cleared again, and then the temperature storage device continues to perform new actions in the environment based on the updated policy network, collects relevant experience of state, action, reward, and next state again, and stores it in the experience replay buffer; Continue training according to the set number of training rounds, and simultaneously monitor the prediction accuracy of the model on the training set; When the prediction accuracy of the model on the training set exceeds the preset accuracy for three consecutive training rounds, it is determined that the training of the temperature storage device parameter optimization model has been completed.

6. A control method for a temperature storage device based on a chemically reversible thermal effect according to claim 5, characterized in that: The parameters of the temperature storage device after output optimization include: Analyzing whether the number of chemical heat exchange energy storage modules included in the temperature storage device exceeds one; If not, maintain the original setting; If yes, then obtain relevant information of different chemical heat exchange energy storage modules, including location distribution, received light, and different heat / cold sources close to them; The relevant information of different chemical heat exchange energy storage modules, the relevant data of the application scenario of this temperature storage device after data preprocessing, the operation data of the temperature storage device during the execution of the control plan and the relevant data of the execution results are input into the pre-built multi-agent collaborative optimization model, and the parameters of different chemical heat exchange energy storage modules after optimization are output.

7. A control method for a temperature storage device based on a chemically reversible thermal effect according to claim 6, characterized in that: The construction of the multi-agent collaborative optimization model is as follows: Define a temperature storage device containing multiple chemical heat exchange energy storage modules as an intelligent agent, collect the actual operation data of the intelligent agent in different application scenarios, different seasons, different time periods and various working conditions, and perform data preprocessing. The actual operation data includes the state attributes of each intelligent agent at each moment, the actions taken and the corresponding overall system performance indicators. The state attributes include the real-time temperature, pressure value, chemical concentration and working mode identification inside the module. The actions taken include the spray water flow rate, heating power value and valve opening and closing status. The overall system performance indicators include overall energy efficiency and temperature uniformity in each area. Divide the preprocessed running data into a training set and a validation set according to a preset ratio; In the training set, the agent is allowed to select and execute actions based on the current state according to the action probability distribution output by the proximal strategy optimization algorithm. After the agent completes the action, the environment will feedback the reward value according to the pre-set reward rules and update the current state; After receiving the reward and state update information, the agent stores the relevant experience generated by this interaction, i.e. the current state, the action performed, the reward obtained, and the next state after the update, into the buffer. After accumulating a preset amount, it extracts batch data to update the neural network parameters and repeats the training continuously. Regularly evaluate the model performance on the validation set, using a comprehensive quantitative evaluation method with validation adjustment to calculate the comprehensive quantitative score; When the comprehensive quantization score exceeds the preset comprehensive quantization score or the number of training rounds reaches the preset number of training rounds, it is determined that the multi-agent collaborative optimization model has completed training.

Images