A lithium bromide heat pump regulation method coupling solar light heat and computing power waste heat

Abstract

The present application relates to the technical field of polytropic complementation and waste heat recovery, in particular to a lithium bromide heat pump regulation method coupled with solar light heat and computing power waste heat; comprising data acquisition, model calculation, dynamic regulation and safety monitoring steps; the method collects multi-source operation data, inputs a thermodynamic matching model; the core is to calculate the coupling heat source exergy efficiency and heat response time constant, generate flow regulation and heat source switching instructions, and maintain the generator in the high efficiency temperature range; at the same time, the crystallization safety margin is calculated in real time, and the anti-crystallization process is triggered when it is lower than the threshold value; the present application realizes the transformation from static threshold dependence to dynamic thermodynamic matching, solves the problem of high-grade heat energy waste caused by source-load asynchrony, and maximizes the system thermodynamic perfection.

Description

Technical Field

[0001] This invention relates to the field of multi-energy complementarity and waste heat recovery technology, specifically a lithium bromide heat pump control method that couples solar thermal energy with computing power waste heat. Background Technology

[0002] In existing multi-energy coupled heating systems, solar thermal energy and waste heat from data center computing power are often used together as the driving heat source for lithium bromide absorption heat pumps. These two types of heat sources have significant time asynchronicity and volatility. Solar irradiance is affected by meteorological cloud cover, which generates high-frequency noise, while computing load changes suddenly with business throughput.

[0003] Existing control schemes generally adopt simple logic control based on static temperature thresholds, lacking in-depth evaluation of heat source energy quality and system dynamic thermal response characteristics. When facing source load fluctuations, this lagging control method is difficult to achieve precise matching of flow and heat, which can easily cause violent fluctuations in generator inlet temperature, leading to frequent deviations of the unit from the high-efficiency operating range and resulting in energy efficiency degradation. In addition, due to the lack of an internal state mapping mechanism based on externally measurable parameters, existing technologies cannot quantify the crystallization safety margin of lithium bromide solution in real time, resulting in the inability to trigger anti-crystallization protection in time under low temperature difference or low load conditions, which seriously threatens the continuous safe operation of the unit.

[0004] Therefore, how to achieve efficient thermodynamic matching of heterogeneous heat sources and improve system stability under complex operating conditions has become an urgent technical problem to be solved. Summary of the Invention

[0005] To address the aforementioned technical problems, this invention provides a lithium bromide heat pump control method that couples solar thermal energy with waste heat from computing power. Specifically, the technical solution of this invention includes:

[0006] Data acquisition steps: Collect solar irradiance data, computing load data, and heat pump unit operating parameters through a multi-source data sensing layer. The heat pump unit operating parameters include dilute solution concentration, concentrated solution concentration, and generator inlet temperature.

[0007] Model solution steps: Input the collected data into the thermodynamic matching model and solve for the exergy efficiency and thermal response time constant of the coupled heat source;

[0008] Dynamic control steps: Based on the exergy efficiency and thermal response time constant of the coupled heat source, generate flow regulation commands and heat source switching commands to maintain the generator inlet temperature within the preset high-efficiency temperature range;

[0009] Safety monitoring steps: Calculate the crystallization safety margin in real time, and trigger the anti-crystallization dilution process when the crystallization safety margin is lower than the preset safety threshold.

[0010] Preferably, the calculation of the exergy efficiency of the coupled heat source includes:

[0011] Analysis of source-load asynchronous characteristics: Based on the time series characteristics of solar irradiance data and computing load data, the time deviation between peak energy supply and peak heat pump demand is identified;

[0012] Calculate the weighted exergy efficiency: Based on the time deviation, determine the weighting coefficients of solar thermal energy and computing waste heat, calculate the Carnot factors corresponding to the solar thermal energy temperature and computing waste heat temperature respectively, and use the weighting coefficients to perform a weighted sum of the Carnot factors of the two to obtain the exergy efficiency of the coupled heat source; the Carnot factor characterizes the work potential of the heat source temperature relative to the ambient temperature.

[0013] Preferably, the generation of flow regulation instructions and heat source switching instructions includes:

[0014] If the exergy efficiency of the coupled heat source is greater than the preset efficiency threshold, the first adjustment command is generated to control the solar thermal circuit and the computing power waste heat circuit to be directly connected in series for heating.

[0015] If the exergy efficiency of the coupled heat source is less than or equal to the preset efficiency threshold and greater than or equal to the minimum operating threshold, a second adjustment command is generated to control the intervention of the auxiliary heat source and adjust the mixing ratio.

[0016] If the exergy efficiency of the coupled heat source is less than the minimum operating threshold, a third adjustment command is generated to cut off the heat pump unit's heating supply and switch to standby cycle mode.

[0017] Preferably, the method further includes:

[0018] Before inputting the data into the multi-source data perception layer, the data is preprocessed to obtain preprocessed data;

[0019] The preprocessing includes:

[0020] The Kalman filter algorithm is used to smooth the solar irradiance data and eliminate high-frequency noise caused by cloud cover.

[0021] The moving average method is used to extract trends from computing load data, which helps to mitigate the impact of sudden fluctuations in computing power on heat source flow.

[0022] Preferably, the safety monitoring steps specifically include:

[0023] Based on the generator inlet temperature and the concentration of the concentrated solution, the saturated crystallization temperature under the current operating conditions is calculated using the Durin equation.

[0024] Calculate the difference between the current solution temperature and the saturation crystallization temperature to obtain the crystallization safety margin;

[0025] The crystallization safety margin is compared with the preset safety threshold: if the crystallization safety margin is less than the preset safety threshold, it is determined that there is a crystallization risk; if the crystallization safety margin is greater than or equal to the preset safety threshold, it is determined that the system is safe to operate.

[0026] Preferably, the anti-crystallization dilution process includes:

[0027] In response to the determination of a crystallization risk, a melting and heating command is generated to increase the inlet temperature of the generator;

[0028] Simultaneously, a bypass adjustment command is generated to open the refrigerant water bypass valve, directly mixing the refrigerant water into the concentrated solution circuit to reduce the concentration of the concentrated solution until the crystallization safety margin is restored to above the preset recovery threshold.

[0029] Preferably, the step of generating a flow regulation command based on the exergy efficiency of the coupled heat source and the thermal response time constant includes:

[0030] Calculate the thermal response time constant, which is defined as the time required for the system to reach 63.2% of the steady-state value from the time it receives the signal of a step change in the temperature of the heat source;

[0031] The execution step size of the flow regulation command is corrected based on the thermal response time constant.

[0032] If the thermal response time constant is less than the preset response threshold, the first step of the long adjustment strategy is adopted to quickly respond to fluctuations.

[0033] If the thermal response time constant is greater than or equal to the preset response threshold, a gradual adjustment strategy with a second step length less than the first step length is adopted to prevent system oscillation.

[0034] Preferably, the method further includes:

[0035] Establish a computing power heat dissipation guarantee mechanism to forcibly activate the backup heat dissipation bypass when the heat pump unit fails or is shut down for maintenance;

[0036] Monitor the rate of change of computing load data. If the rate of change is greater than the preset surge threshold, prioritize locking the flow of the computing waste heat loop and adjust the flow of the solar thermal loop to balance the total heat input and ensure that the cooling of computing equipment is not interrupted. If the rate of change is less than or equal to the preset surge threshold, maintain the current flow adjustment strategy.

[0037] Preferably, the analysis of the asynchronous characteristics of the source load specifically includes:

[0038] The solar irradiance and computing load within a preset time period are predicted using a long short-term memory network model.

[0039] Based on the prediction results, the supply and demand matching index for a future preset time period is calculated. The supply and demand matching index is the ratio of the predicted solar energy supply to the computing power heat load demand.

[0040] If the supply and demand matching index is less than the preset matching threshold, a command to intervene in the thermal storage device is generated in advance to use the thermal storage device to smooth out asynchronous fluctuations in the source and load.

[0041] Compared with the prior art, the present invention has the following beneficial effects:

[0042] 1. This method proposes a control approach based on a thermodynamic matching model. Instead of relying solely on static temperature thresholds, it controls the system by calculating the exergy efficiency of coupled heat sources and the thermal response time constant. By analyzing the asynchronous characteristics of solar thermal energy and computing power waste heat, the weighted exergy efficiency is calculated, which can accurately quantify the work potential of heat sources at different temperatures and optimize the weight allocation. This effectively solves the problem of high-grade heat energy waste caused by the mismatch between solar energy and computing power load in terms of time and grade. It enables the system to maintain a high-efficiency temperature range under complex operating conditions, avoids unit performance degradation caused by heat source fluctuations, and maximizes the thermodynamic perfection of the system.

[0043] 2. This method employs an adaptive flow regulation strategy based on the thermal response time constant. By calculating the time constant from a step signal to a steady state in real time, it intelligently distinguishes between the fast and slow change characteristics of the system and adjusts the execution step size of the regulation command accordingly: a small step size is used for gradual regulation when the response is fast, and a large step size is used for fast response when the response is slow. Combined with Kalman filtering and moving average method for preprocessing the raw data, high-frequency meteorological noise and sudden fluctuations in computing power are effectively filtered out. This design avoids frequent valve oscillations and temperature overshoot caused by lag or over-adjustment in traditional control, thereby ensuring the stability of flow regulation and extending the service life of hardware such as valves and pumps.

[0044] 3. To address the issue that existing technologies cannot directly assess the risk of internal crystallization, this method establishes a mathematical model for calculating the safety margin of crystallization. Utilizing the Thulin equation and the heat exchange differential model, externally collected temperature and concentration parameters are mapped to the saturation crystallization temperature of the internal solution, achieving real-time quantification of risk. Once the safety margin falls below the threshold, a dual defense process of melting crystal heating and bypass dilution is immediately triggered. By simultaneously increasing the solution temperature and decreasing the solution concentration, crystallization nuclei are eliminated in the shortest possible time. This effectively prevents pipeline blockage failures in lithium bromide units under low temperature differential or low load conditions, ensuring the continuous and safe operation of the equipment.

[0045] 4. This method introduces a deep learning-based source load prediction and computing power heat dissipation guarantee mechanism; it uses a long short-term memory network model to predict the future supply and demand matching degree and schedule the thermal storage device in advance to smooth fluctuations; more importantly, it establishes the principle of prioritizing computing power safety, and by monitoring the computing power load change rate, it automatically locks the waste heat loop flow and forcibly activates the backup heat dissipation bypass when the load surges; this mechanism ensures that the heat dissipation of the data center is not interrupted under extreme load impact, solves the contradiction that pursuing heat recovery efficiency may threaten the operational safety of computing power equipment, and improves the reliability and robustness of the entire energy system. Attached Figure Description

[0046] The present invention will be further explained below with reference to the accompanying drawings and embodiments:

[0047] Figure 1 This is a flowchart of the method of the present invention. Detailed Implementation

[0048] To make the objectives, technical solutions, and advantages of this invention clearer, the invention will be further described in detail below with reference to specific embodiments.

[0049] Example 1:

[0050] Please see Figure 1 A method for regulating a lithium bromide heat pump that couples solar thermal energy with waste heat from computing power, comprising:

[0051] Data acquisition steps: Collect solar irradiance data, computing load data, and heat pump unit operating parameters through a multi-source data sensing layer. Among them, the heat pump unit operating parameters include dilute solution concentration, concentrated solution concentration, and generator inlet temperature.

[0052] Model solution steps: Input the collected data into the thermodynamic matching model and solve for the exergy efficiency and thermal response time constant of the coupled heat source;

[0053] Dynamic control steps: Based on the exergy efficiency and thermal response time constant of the coupled heat source, generate flow regulation commands and heat source switching commands to maintain the generator inlet temperature within the preset high-efficiency temperature range;

[0054] Safety monitoring steps: Calculate the crystallization safety margin in real time, and trigger the anti-crystallization dilution process when the crystallization safety margin is lower than the preset safety threshold.

[0055] This embodiment proposes a lithium bromide heat pump control method that couples solar thermal energy with waste heat from computing power. The method relies on a system architecture that includes a solar collector, a data center liquid cooling circuit, a lithium bromide absorption heat pump unit, and an intelligent control center. The system performs a data acquisition step, and obtains the system's boundary conditions and operating status in real time through a multi-source data sensing layer. This multi-source data sensing layer consists of a hardware system composed of a high-precision sensor network and a data acquisition card (DAQ) deployed in the physical system.

[0056] The system collects solar irradiance data in real time through a photoelectric solar intensity meter installed in the solar collector field. Server power consumption or CPU utilization metrics read through the Data Center Infrastructure Management System (DCIM) interface are used as computing load data. Unit: kilowatts (kW); If the read data is a CPU utilization index, the system converts it into a power value using a preset power consumption-load nonlinear fitting curve and then assigns it to the appropriate power level. To ensure the uniqueness of the dimensions in subsequent energy calculations, this embodiment prioritizes directly reading the server power consumption value; and measures the concentration of the dilute solution using an online refractometer and a PT100 temperature sensor. Concentration of concentrated solution and generator inlet temperature ;

[0057] The system executes the model solution step, inputting the collected data into the thermodynamic matching model. This model is not a simple logical judgment, but a computational kernel built on the second law of thermodynamics, designed to calculate the exergy efficiency of coupled heat sources. With thermal response time constant These two key control indicators; the model treats the solar collector and the computing power waste heat loop as two independent non-isothermal heat sources, and constructs a mapping between input parameters and work potential based on Carnot's theorem; and The solar working fluid and the waste heat fluid from computing power were characterized relative to the environmental reference state. The maximum theoretical work potential, i.e., physical pyrolysis, is determined by the model through weighting coefficients. By associating the physical structural parameters of the two, the following mapping relationship is derived: The system is based on the formula:

[0058]

[0059] Calculate exergy efficiency, where, The ambient thermodynamic absolute temperature (K) is given. These are the thermodynamic absolute temperatures (K) of solar energy and waste heat water supply from computing power, respectively. These are the weighting coefficients; based on the formula:

[0060]

[0061] Calculate the time constant, where, This represents the equivalent heat capacity of the system. The subscript eq indicates equivalent, and its calculation method is as follows: It covers the heat capacity composition of the heat exchanger's metal wall, internal fluid, and external insulation layer; To determine the overall heat transfer coefficient, its value was determined based on the initial experimental identification, although It may drift slowly due to scaling during long-term operation, but it is considered a constant within a single control cycle, and in this embodiment, it is periodically corrected through heat exchange efficiency feedback during subsequent operation; The effective heat exchange area is the design constant, and the estimated scaling reduction factor during long-term operation has been deducted during the calculation.

[0062] In the dynamic control step, the controller relies on the calculated... and The system generates corresponding hardware action instructions, including adjusting the opening of the electric regulating valves on the solar circuit and the computing power waste heat circuit to change the mixed flow entering the generator, and controlling the on / off state of the three-way valve to determine whether to use series, parallel, or single heat source heating mode. The purpose is to maintain the generator inlet temperature. Always within the preset high-efficiency temperature range Within this range, the system ensures that the unit's COP is within the optimal range; during the safety monitoring process, the system calculates the crystallization safety margin in real time. In response to Below the preset safety threshold This immediately triggers the anti-crystallization dilution process to prevent the concentrated solution from solidifying and clogging inside the heat exchanger;

[0063] This embodiment introduces a thermodynamic matching model, which no longer relies solely on temperature for control, but combines the quality of energy, i.e., exergy efficiency, with the dynamic characteristics of the system, i.e., the time constant. This enables the system to predict and smoothly transition in complex scenarios such as sudden changes in solar energy or a surge in computing load, effectively avoiding the sudden cooling and heating of the unit and performance degradation caused by heat source fluctuations, and achieving efficient coupling of the two heat sources.

[0064] Calculating the exergy efficiency of the coupled heat source includes:

[0065] Analysis of source-load asynchronous characteristics: Based on the time series characteristics of solar irradiance data and computing load data, the time deviation between peak energy supply and peak heat pump demand is identified;

[0066] Calculate the weighted exergy efficiency: Based on the time deviation, determine the weighting coefficients of solar thermal energy and computing waste heat, calculate the Carnot factors corresponding to the solar thermal temperature and computing waste heat temperature respectively, and use the weighting coefficients to perform a weighted sum of the Carnot factors of the two to obtain the exergy efficiency of the coupled heat source; the Carnot factor characterizes the work potential of the heat source temperature relative to the ambient temperature.

[0067] This embodiment details the specific process of calculating the exergy efficiency of coupled heat sources. Since solar energy and computing load are often asynchronous in time, direct mixing may lead to a waste of high-grade heat energy. Therefore, it is necessary to calculate weighted exergy efficiency to guide matching. The system analyzes the asynchronous characteristics of the source and load, based on solar irradiance data within a sliding time window. Computing load data To identify the peak times of both data, the system employs a rolling update mechanism, meaning that every 15 minutes, it rescans and updates based on a rolling time window that includes historical data from the past 12 hours and predicted data for the next 12 hours. and ;

[0068] To avoid the subjectivity of manual observation, the system construction length is [length missing]. The corresponding data vectors for a 24-hour window and a 15-minute resolution. and Execute the vector extremum index search algorithm respectively: and And convert the index to physical time: , and define time deviation as follows:

[0069]

[0070] in, The source is historical data statistics or short-term forecasts, and the physical meaning is the moment when solar energy supply reaches its peak. The source is load analysis, and its physical meaning is the moment when the demand for the computational heat pump reaches its peak; the system calculates the weighted exergy efficiency, determines the weights based on time deviation, and combines the Carnot factor to calculate the comprehensive exergy efficiency. The calculation formula is as follows:

[0071]

[0072] in, The source is the calculation result, and the physical meaning is the exergy efficiency of the coupled heat source; The data is sourced from real-time data collected by an outdoor weather station. The specific measurement location is a temperature sensor installed in a Stevenson screen 1.5 meters above the ground in a well-ventilated outdoor location to eliminate interference from direct solar radiation. The physical meaning is the environmental reference thermodynamic absolute temperature (K). The source is sensor data, and the specific measurement location is at the main outlet manifold of the solar collector array, and it is located before any mixing valve or heat loss link, in order to accurately obtain the water supply temperature on the heat source side. The physical meaning is the thermodynamic absolute temperature K of the water supply in the solar thermal circuit. The data source is sensor data, specifically measured on the connecting pipe between the outlet of the secondary heat exchanger of the liquid-cooled CDU in the data center and the inlet of the heat pump unit, located close to the heat pump inlet. The physical meaning is the thermodynamic absolute temperature (K) of the water supply for the computing power waste heat loop. If the measured value is in degrees Celsius, 273.15 needs to be added for conversion. The source is the calculation result, and the physical meaning is the dynamic weighting coefficient of solar thermal energy.

[0073] Weighting coefficient The calculation is based on time deviation. With system thermal inertia factor :

[0074]

[0075] Here The base of the natural logarithm; parameter , here Specifically refers to the adjustment coefficient, which is distinct from the physical unit of the time index. , used to characterize the asynchronous growth rate, is set to It is derived from fitting the system response curve under typical operating conditions; The maximum asynchronous tolerance time, expressed in hours (h), is proportional to the effective regulating capacity of the water tank, and the correlation satisfies the following conditions. ; The source is a preset constant, and its physical meaning is an adjustment sensitivity constant, or adjustment coefficient, to ensure the exponential term. The value is dimensionless; in this formula, the time deviation is... It is necessary to uniformly convert the data to hours (h) as the unit of input; The source is the capacity parameter of the hot water storage tank, and its physical meaning is the maximum allowable asynchronous tolerance time of the system;

[0076] To ensure that the calculation results conform to the actual dynamic characteristics of the system and can be reproduced by those skilled in the art, in this embodiment... Set as This value was determined through on-site debugging and can effectively balance the smoothness and sensitivity of weight changes. The setting is 4 hours, which matches the maintenance time of the 50 cubic meter hot water storage tank under full load in the system configuration;

[0077] This embodiment introduces the Carnot factor to accurately quantify the work potential of heat sources at different temperatures. By utilizing the weighting coefficients determined by the asynchronous characteristics of the source and load, the control strategy can favor the heat source that is more suitable at the current moment when facing asynchronous scenarios such as strong solar energy at noon but low computing power load, thereby maximizing the thermodynamic perfection of the system and reducing irreversible losses.

[0078] Generate flow regulation commands and heat source switching commands, including:

[0079] If the exergy efficiency of the coupled heat source is greater than the preset efficiency threshold, the first adjustment command is generated to control the solar thermal circuit and the computing power waste heat circuit to be directly connected in series for heating.

[0080] If the exergy efficiency of the coupled heat source is less than or equal to the preset efficiency threshold and greater than or equal to the minimum operating threshold, a second adjustment command is generated to control the intervention of the auxiliary heat source and adjust the mixing ratio.

[0081] If the exergy efficiency of the coupled heat source is less than the minimum operating threshold, a third adjustment command is generated to cut off the heat pump unit's heating supply and switch to standby cycle mode.

[0082] This embodiment details the hierarchical control logic for generating flow regulation commands and heat source switching commands, with an efficiency threshold preset within the controller. With minimum operating threshold In the specific parameter configuration of this embodiment, considering the COP decay characteristics of lithium bromide units under low heat source quality, a preset efficiency threshold is used. The value is set to 0.55. This value is determined based on the thermodynamic perfection test data of the lithium bromide unit under rated operating conditions. In the laboratory environment, when the exergy efficiency of the coupled heat source is lower than 0.55, the heating performance coefficient COP of the unit begins to decrease significantly and deviates from the linear high-efficiency zone, indicating that the heat source quality is no longer sufficient to support a high-efficiency thermodynamic cycle.

[0083] Minimum operating threshold The value is set at 0.25. This value is an empirical value determined based on the unit's minimum start-up temperature difference characteristics. During multiple on-site commissionings, it was found that when the exergy efficiency is below 0.25, corresponding to a heat source temperature of approximately 65 degrees Celsius, the absorption capacity of the dilute solution decreases sharply, and the unit's COP approaches 1.0, thus losing its energy-saving significance as a heat pump. This is in response to the exergy efficiency of the coupled heat source. Greater than The system generates the first adjustment command, controlling the solar circuit and the computing power waste heat circuit to be connected in series, so that the computing power waste heat acts as a preheating source to heat the working fluid to a medium temperature, and then enters the solar collector to heat it to a high temperature, utilizing the high matching degree of the two heat sources to achieve cascade heating; in response to In and In between, the system generates a second adjustment command, starts the auxiliary heat source, and adjusts the mixing ratio of the three-way regulating valve; to avoid temperature fluctuations caused by simple open-loop control, the system calculates the specific opening degree of the auxiliary heat source mixing valve based on an incremental PI control algorithm. The calculation logic uses an incremental PI control algorithm:

[0084]

[0085] Among them, the error term , This represents the error from the previous moment; Preset gain, To achieve the target generator inlet temperature, an external high-grade heat source is precisely introduced for afterburner combustion to maintain a stable generator inlet temperature; in response to Less than The system generates a third adjustment command to close the main steam or hot water valve of the heat pump unit, causing the unit to enter the internal solution circulation standby state, and to switch the waste heat of computing power to the standby dry cooler for heat dissipation, in order to prevent crystallization failure under extremely low COP conditions.

[0086] The hierarchical control strategy constructed in this embodiment maximizes system benefits through series cascade utilization when the heat source is sufficient, while clearly defining the mode switching boundary when the heat source quality is insufficient, avoiding ineffective operation of the heat pump in the low-efficiency zone, and ensuring the determinism and stability of system control under complex operating conditions with multiple heat sources.

[0087] Example 2:

[0088] The method also includes:

[0089] Before inputting the data into the multi-source data perception layer, the data is preprocessed to obtain preprocessed data;

[0090] Preprocessing includes:

[0091] The Kalman filter algorithm is used to smooth the solar irradiance data and eliminate high-frequency noise caused by cloud cover.

[0092] The moving average method is used to extract trends from computing load data, which helps to mitigate the impact of sudden fluctuations in computing power on heat source flow.

[0093] This embodiment describes the data preprocessing process before the data is input into the perception layer, which aims to eliminate noise and glitches in the raw data and prevent frequent oscillations of the actuator. Regarding the timing description mentioned in the embodiment before the data is input into the multi-source data perception layer, in the specific execution logic of the physical system, it refers to the preprocessing steps performed by the edge computing unit after the data is acquired by the front-end sensor, but before it is transmitted and written into the core database of the perception layer or input into the upper-layer control model. This ensures that the data flow conforms to the logical closed loop of first acquisition, then processing, and then input.

[0094] For solar irradiance data, the system uses a Kalman filter algorithm to eliminate high-frequency noise caused by cloud cover. In the univariate state estimation model of this embodiment, it is assumed that the irradiance change follows a local constant process; therefore, the state transition matrix is ​​explicitly defined. With measurement matrix All values ​​are scalars of 1.0. The above simplifying assumptions apply to the system operating in a quasi-steady state; where the process noise covariance... Measurement of noise covariance and initial error covariance The selection criteria are based on empirical values ​​determined through statistical analysis of historical operating data and system identification experiments. In addition to the state update equation, time update steps and Kalman gain calculation steps are also defined. The complete algorithm flow is as follows:

[0095] Updated in time:

[0096]

[0097]

[0098] Calculate the Kalman gain:

[0099]

[0100] Status Update:

[0101]

[0102] Covariance update:

[0103]

[0104] in, : The predicted prior estimate of irradiance; Prior estimation error covariance; : Posterior estimation error covariance; : The measured irradiance value of the sensor at time k; The optimal estimate of irradiance at time k;

[0105] For computing load data, the system uses a moving average method to smooth out sudden fluctuations in computing power. The calculation formula is as follows:

[0106]

[0107] in, The source is the calculation result, and the physical meaning is the smoothed load value after processing at time t; The source is data collection, and its physical meaning is the original load value of the i-th sampling point in the past; The source is a preset value, and its physical meaning is the length of the sliding window;

[0108] To ensure the convergence and effectiveness of the algorithm in this physical system, the process noise covariance of the Kalman filter is... Set as Measure noise covariance Set to 0.1, initial covariance Set to 1.0; Sliding window length for the moving average method. Specifically, it is set to 60 sampling points. If the sampling frequency is 5 seconds / time, it corresponds to a 5-minute time span. This parameter setting is intended to cover a typical short-term fluctuation cycle of computing load and prevent the flow valve from malfunctioning due to instantaneous computing task throughput.

[0109] This embodiment effectively filters out the drastic fluctuations in irradiance caused by rapidly passing clouds through Kalman filtering, and uses the moving average method to filter out the spikes caused by the instantaneous start and stop of computing tasks. This makes the preprocessed data more reflective of the true trend of energy changes, thereby making subsequent flow regulation more stable under variable weather conditions and extending the service life of valves and pump sets.

[0110] The safety monitoring steps specifically include:

[0111] Based on the generator inlet temperature and the concentration of the concentrated solution, the saturated crystallization temperature under the current operating conditions is calculated using the Durin equation.

[0112] Calculate the difference between the current solution temperature and the saturation crystallization temperature to obtain the crystallization safety margin;

[0113] The crystallization safety margin is compared with the preset safety threshold: if the crystallization safety margin is less than the preset safety threshold, it is determined that there is a crystallization risk; if the crystallization safety margin is greater than or equal to the preset safety threshold, it is determined that the system is safe to operate.

[0114] This embodiment details the specific algorithm for monitoring crystallization safety. Given that crystallization is the most fatal failure in lithium bromide generator units, this step aims to quantify the risk boundary in real time. Based on the generator inlet temperature and concentrated solution concentration, the system uses the Durin equation to calculate the saturated crystallization temperature under the current operating conditions. The coefficients in the formula... The Durin plot coefficients, derived from experimental fitting data, correspond to the fitting slope and bias term in different temperature ranges, respectively; the approximate fitting formula is as follows:

[0115]

[0116] in, The source is the calculation result, and the physical meaning is the saturation crystallization temperature; This represents the solution concentration. When inputting values ​​for calculations, a percentage format must be used consistently, i.e., the value before the percentage sign should be taken, such as... Substitution required Calculations are performed; in this embodiment, for example, the concentration is... The time value is ; The source is an empirical constant obtained by performing least-squares nonlinear regression fitting on standard solubility data of lithium bromide aqueous solution within the working concentration range, with the following values: , , , The vertical asymptote characterizing the crystallization curve ensures formula convergence within the working concentration range;

[0117] System calculates crystallization safety margin To strictly respond to the generator inlet temperature-based response in Example 1 To limit the control, and considering that the data acquisition steps in Example 1 did not include direct measurement of the solution temperature inside the generator, the system introduces a heat exchange terminal difference model to measure the acquired generator inlet heat source temperature. Mapped to the estimated temperature of the concentrated solution at the generator outlet The calculation formula is as follows:

[0118]

[0119] in, The source is a preset thermodynamic constant, which physically represents the nodal temperature difference of the generator heat exchanger tube bundle. This value is a fixed value determined based on the unit design parameters and does not change with operating conditions. In this embodiment, it is set based on the heat exchanger design parameters. Celsius;

[0120] Considering that the actual crystallization risk in lithium bromide generators typically occurs at the lowest temperature point in the concentrated solution loop, i.e., the outlet of the solution heat exchanger or the inlet of the absorber, rather than the higher temperature generator outlet, the system needs to calculate the temperature of the critical low temperature point to avoid false negatives in safety monitoring. The calculation formula is as follows:

[0121]

[0122] in, The characteristic temperature drop of the concentrated solution in the solution heat exchanger is also treated as a design constant, and in this embodiment, it is set based on the rated operating conditions of the unit. Celsius; based on this, the revised formula for calculating the crystallization safety margin is: The system performs a risk assessment and will With preset safety threshold A comparison is performed, including a preset safety threshold. The method for obtaining this value is based on systematic error analysis, specifically set at 8 degrees Celsius. This value consists of three parts: the measurement error boundary of the temperature sensor, ±0.5°C; the crystallization point drift calculated from the concentration sensor error, ±2.5°C; and the thermal inertia temperature drop margin before the anti-crystallization measures take effect, 5.0°C. The sum of these three values ​​is the safety alarm baseline value. (The last part, "responding to...", appears to be an error and is left untranslated.) If the system is deemed to have a crystallization risk, it is considered safe; otherwise, it is considered safe.

[0123] This embodiment not only uses the Thulin plot to quantify the crystallization point, but also introduces an end-to-end difference model. and heat exchanger temperature drop correction The external heat source temperature of the embodiment was established. The clear mathematical correlation between the internal solution and the critical risk point state of the solution resolves the logical gap that cannot be directly assessed by collecting only external parameters, and enables non-invasive safety monitoring based on existing sensor sets.

[0124] Anti-crystallization dilution process includes:

[0125] In response to the determination of a crystallization risk, a melting and heating command is generated to increase the inlet temperature of the generator;

[0126] Simultaneously, a bypass adjustment command is generated to open the refrigerant water bypass valve, directly mixing the refrigerant water into the concentrated solution circuit to reduce the concentration of the concentrated solution until the crystallization safety margin is restored to above the preset recovery threshold.

[0127] This embodiment details the anti-crystallization dilution process. When a risk is detected, the system executes dual safeguards: the system generates a melting and heating command, controlling the auxiliary heat source to increase its power or reduce the solution circulation rate in the generator to raise the outlet temperature, directly increasing the actual solution temperature. This increases the safety margin; the system generates a bypass control command to open the refrigerant water bypass valve located between the condenser and absorber chambers, directly introducing pure water (refrigerant water) from the condenser into the concentrated solution circuit, aiming to physically reduce the concentration of the concentrated solution. This leads to the saturation crystallization temperature. A significant decrease; the process continues until the crystallization safety margin is reached. Restore to the preset recovery threshold In order to avoid frequent oscillations of the system near the critical point, this embodiment adopts hysteresis control logic and sets a preset recovery threshold. The specific setting is 12 degrees Celsius, which is higher than the alarm threshold. Set the temperature 4 degrees Celsius above the safe level before resetting.

[0128] This embodiment adopts a two-pronged approach of heating and dilution. Compared with the slow reaction of heating alone or the impact of dilution alone on the cooling capacity, this combined control can eliminate crystallization nuclei in the shortest time. It is a key line of defense to ensure continuous operation of the computing power waste heat pump system under the condition of low temperature difference.

[0129] Based on the exergy efficiency and thermal response time constant of the coupled heat source, a flow regulation command is generated, including:

[0130] Calculate the thermal response time constant, which is defined as the time required for the system to reach 63.2% of the steady-state value from the time it receives the signal of a step change in the temperature of the heat source.

[0131] The execution step size of the flow regulation command is corrected based on the thermal response time constant.

[0132] If the thermal response time constant is less than the preset response threshold, the first step of the long adjustment strategy is adopted to quickly respond to fluctuations.

[0133] If the thermal response time constant is greater than or equal to the preset response threshold, a gradual adjustment strategy with a second step length less than the first step length is adopted to prevent system oscillation.

[0134] This embodiment details a flow regulation strategy based on the thermal response time constant; the system calculates the thermal response time constant. To ensure full disclosure of the technical solution, this embodiment clarifies the theoretical calculation formula for the thermal response time constant, which is derived based on the lumped parameter method:

[0135]

[0136] Here Includes total heat capacity of circulating pipes and fluids, unit: Specific heat capacity The heat transfer coefficient is set as a function of concentration and obtained by real-time online monitoring of concentration and retrieval from property tables; It has an online update mechanism, and the system updates every [time period]. The system automatically re-identifies the operating power and temperature difference every hour. value;

[0137] in, The source is the equipment's factory parameters, and the physical meaning is the equivalent thermal mass of the generator components; The source is a physical property database, and the physical meaning is the specific heat capacity of the lithium bromide solution at the current concentration; The source is thermal design parameters, and its physical meaning is the overall heat transfer coefficient of the generator; The source is geometric parameters, and the physical meaning is the effective heat transfer area of ​​the generator;

[0138] In actual operation, scaling and other factors can lead to... Value changes, therefore, given The physical definition requires the system to undergo a complete step response process to obtain accurate values. To resolve the timing contradiction between real-time control and delayed calculation, this embodiment employs an asynchronous update mechanism of background iterative identification and foreground real-time application to correct the theoretical values. A current effective time constant is maintained in the controller's memory. The subscript cur indicates current, and the initial value is set to the theoretical value calculated by the above formula.

[0139] The background recognition process employs an online passive event-triggered recognition mechanism: the system monitors the rate of change of the heat source temperature in real time, and when a change is detected... At that time, mark the current moment as the step start moment. And record the steady-state temperature before that moment. ,in, The threshold for step detection is set to 0.5. This threshold is obtained based on the spectral analysis of the noise substrate of the field temperature sensor. Three times the standard deviation of the noise amplitude is selected as the detection threshold to prevent measurement noise from falsely triggering step detection.

[0140] Subsequently, data is continuously buffered, and the controller utilizes pre-allocated data of length [missing information]. The subscript `buf` indicates `buffer`, meaning the circular buffer stores the historical temperature sequence. This length is based on a 1Hz sampling frequency and is calculated by multiplying the generator's maximum thermal time constant under full load by a safety factor of 1.33. It aims to cover a data window of at least 20 minutes to prevent data loss due to excessively long system thermal response times. Key historical data at any given moment is overwritten or lost, thus ensuring the integrity of the backtracking calculation;

[0141] The process continues until the temperature is detected to have returned to a steady state. The logic for determining when the temperature has returned to a steady state is as follows: the calculation length is... Variance of temperature data within the sliding time window And monitor the rate of temperature change in real time. When satisfied and continued At a certain time, the system is determined to have reached a new steady state; the above thresholds are selected based on empirical parameters set according to the fluctuation characteristics of the system under rated flow; step identification threshold. This is an adjustable parameter, and its value is linearly scaled according to the system's water capacity.

[0142] Then, the new steady-state asymptotic value was locked. The system backtracks the buffer data when the generator inlet temperature... First crossing of target value Record the time. ,in, ; Calculate the value of a single measurement And update the stored parameters using the exponentially weighted average method: The weighting coefficients of 0.8 and 0.2 are determined based on autocorrelation analysis of the system's historical operating data. They aim to balance the smoothness and sensitivity of parameter updates, so as to smooth measurement noise and adapt to characteristic drift caused by equipment fouling.

[0143] The foreground control loop directly reads data from memory. As under the current working conditions Correct the execution step size of the flow regulation command. Among them, the preset response threshold Set as The value, measured in seconds, is determined based on the inherent thermal inertia of the generator components and their connecting pipes, and is used to distinguish between the fast and slow changing characteristics of the system; if The first step is a long adjustment strategy. Set to 5% of the valve's full stroke, this step size is an empirical value determined through step response experiments, designed to ensure that a single adjustment can cause a change of approximately 1.5% in the outlet temperature to quickly eliminate deviations; if A second-step gradual adjustment strategy is adopted. The step size is set to 0.8% of the valve's full stroke. This step size is determined based on the minimum resolution of the electric regulating valve positioner (0.5%) plus a dead zone margin of 0.3% to prevent overshoot.

[0144] In order to adjust the step size This is translated into specific actions to form closed-loop control. The controller executes the following discrete incremental adjustment algorithm:

[0145] Real-time calculation of current temperature deviation ,in, This is the preset center value of the high-efficiency temperature range;

[0146] Using sign function to determine the adjustment direction ;

[0147] Generate the final valve opening control command. ;

[0148] Among them, when When the target temperature is higher than the current actual temperature, The command controls the electric regulating valve to open wider. To increase the flow rate of the heat transfer medium, the flow rate is increased, and vice versa, thus utilizing the dynamic step size. Eliminate steady-state error and suppress oscillations;

[0149] This embodiment solves the causal paradox of traditional methods, which require waiting for the response to finish before parameters can be obtained, by constructing an asynchronous update mechanism and a large-capacity circular buffer queue. At the same time, it combines incremental adjustment logic, so that the control strategy can utilize the latest identification results while ensuring real-time response capability.

[0150] Example 3:

[0151] The method also includes:

[0152] Establish a computing power heat dissipation guarantee mechanism to forcibly activate the backup heat dissipation bypass when the heat pump unit fails or is shut down for maintenance;

[0153] Monitor the rate of change of computing load data. If the rate of change is greater than the preset surge threshold, prioritize locking the flow of the computing waste heat loop and adjust the flow of the solar thermal loop to balance the total heat input and ensure that the cooling of computing equipment is not interrupted. If the rate of change is less than or equal to the preset surge threshold, maintain the current flow adjustment strategy.

[0154] This embodiment establishes a computing power heat dissipation guarantee mechanism. Given that the security of computing equipment outweighs the economic benefits of heat recovery, uninterrupted heat dissipation must be ensured. The system is configured with backup heat dissipation bypass activation logic. In response to a fault code reported by the heat pump unit or a shutdown for maintenance, the controller immediately outputs a signal to open the electric valve leading to the dry cooler or cooling tower, forcibly activating the backup heat dissipation bypass to take over the cooling load of the data center. The system monitors the computing power load change rate in real time. The calculation formula is as follows:

[0155]

[0156] in, The source is the calculation result, and the physical meaning is the rate of change; The source is a constant, and its physical meaning is the sampling interval;

[0157] In response to If the load exceeds the preset surge threshold, the system determines that the computing load has surged and enters the flow lock mode. It locks the flow regulation valve on the computing waste heat loop side to keep it fully open or at the current opening degree. It no longer participates in the fine regulation of the heat pump and only balances the total heat input by adjusting the flow of the solar thermal loop or the auxiliary heat source.

[0158] This embodiment establishes the control principle of prioritizing computing power security. By monitoring the rate of change, it distinguishes between normal fluctuations and sudden surges. In extreme cases, it proactively abandons the pursuit of heat recovery efficiency and prioritizes ensuring heat dissipation from traffic flow, thereby ensuring the inherent safety of data center infrastructure under any load impact.

[0159] The analysis of the asynchronous characteristics of the source load includes:

[0160] The solar irradiance and computing load within a preset time period are predicted using a long short-term memory network model.

[0161] Based on the forecast results, the supply and demand matching index for the future preset time period is calculated. The supply and demand matching index is the ratio of the predicted solar energy supply to the computing power heat load demand.

[0162] If the supply and demand matching index is less than the preset matching threshold, a command to intervene in the thermal storage device is generated in advance to use the thermal storage device to smooth out asynchronous fluctuations in the source and load.

[0163] This embodiment details the prediction of asynchronous source-load characteristics and thermal storage intervention; the system utilizes a Long Short-Term Memory (LSTM) network model for prediction; to ensure the reproducibility of the algorithm, this embodiment specifies the prediction timescale: the input tensor structure of the LSTM model is defined as follows. ,in, Batch size is set to 1 during the online inference phase and 64 during the offline training phase; time step... Set as Feature Dimension Set as Among them, in constructing the input tensor At that time, the feature channel index of the last dimension is clearly defined. Corresponding to the normalized solar irradiance data sequence, index The normalized computing load data sequence is used to ensure feature alignment between the inference and training phases; the model output layer is configured to predict the data sequence for the next 4 hours, with a time resolution of 15 minutes, i.e., the output sequence length is 16 time steps.

[0164] The LSTM model contains two hidden layers. The first hidden layer has 64 nodes and uses the tanh activation function. The second hidden layer has 32 nodes and uses a linear activation function to connect to the output layer. The training loss function is mean squared error. To prevent gradient vanishing or model non-convergence due to large differences in the scale of the input data, the system must perform a normalization preprocessing step before inputting the data into the LSTM. The specific formula is as follows:

[0165]

[0166] in, and To dynamically update values, the system re-collects extreme values ​​every quarter based on load peak and valley changes; the normalization process is performed independently for each input feature;

[0167] Among them, regarding solar irradiance data ,Pick ; Targeting computing load data ,Pick ;

[0168] After the LSTM model outputs predicted values, an inverse normalization operation is performed to restore the physical dimensions:

[0169]

[0170] Calculate the supply and demand matching index based on the restored predicted sequence. To accurately reflect the total energy balance within the future time window, this step integrates and sums all time steps of the future prediction sequence. Furthermore, to prevent computational anomalies where the denominator is zero during periods of extremely low computing load or downtime maintenance, a numerical stability factor is introduced. The calculation formula is as follows:

[0171]

[0172] Before calculation Energy supply items and The energy requirement term is uniformly converted into joules using the energy conversion coefficient. As a unit of measurement; the time step involved in the formula The time is consistently set to seconds throughout the entire logic. ;

[0173] in, Set as In this calculation, in order to convert the predicted physical quantity into an energy value, a single-step energy term is defined:

[0174]

[0175] in, The physical true irradiance value obtained after the inverse normalization step, i.e. corresponding Numerical values ​​are used to ensure the accuracy of energy calculations; To predict the time interval of the sequence, this embodiment uses 0.25 hours. To ensure that the energy calculation results are consistent in units of joules, the time variable is used in the following formulas. The value must be 900; it is strictly forbidden to use hourly units for calculation. The effective light-receiving area of ​​the solar collector is set at 500 square meters. The solar collector's photothermal conversion efficiency is set to 0.6; this is based on the computational heat load demand. Considering the computing load data Unlike the unit of solar irradiance, strict dimensional uniformity is required; this embodiment introduces a unit conversion factor. The calculation formula is as follows:

[0176]

[0177] in, The heat exchange efficiency of the plate heat exchanger is set to 0.95. The value 1000 is used to unify the power unit to watts before combining it with the time unit in seconds. Multiplying ensures that both the numerator and denominator are in Joules; in this embodiment, a preset matching threshold is used. The subscript 't' represents the threshold, which is set to... ; in response to This means that the predicted future solar energy supply will be insufficient to cover [the area]. When the system detects a supply-demand imbalance during the heat recovery demand, it generates an intervention command for the thermal storage device in advance. This command is not a simple on / off signal, but includes specific flow rate setpoints. The calculation logic is as follows:

[0178]

[0179] in, To adjust the gain, the heat release valve of the hot water storage tank is opened for energy compensation; This is the rated flow rate of the thermal storage circuit circulation pump;

[0180] This embodiment clarifies the key preprocessing steps for the engineering implementation of neural network models. By clarifying the reference of physical quantities after denormalization and the specific sequence summation boundary and dimension unification process, it eliminates the logical hidden dangers of physical dimensions in energy matching calculations, ensuring the generalization ability of the prediction model under different seasons and load rates, thereby providing a reliable decision-making basis for the precise intervention of thermal storage devices.

[0181] It should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and are not intended to limit it. Although the present invention has been described in detail with reference to preferred embodiments, those skilled in the art should understand that modifications or equivalent substitutions can be made to the technical solutions of the present invention without departing from the spirit and scope of the technical solutions of the present invention.

Claims

1. A method for regulating a lithium bromide heat pump that couples solar thermal energy with waste heat from computing power, characterized in that, include: Data acquisition steps: Collect solar irradiance data, computing load data, and heat pump unit operating parameters through a multi-source data sensing layer. The heat pump unit operating parameters include dilute solution concentration, concentrated solution concentration, and generator inlet temperature. Model solution steps: Input the collected data into the thermodynamic matching model and solve for the exergy efficiency and thermal response time constant of the coupled heat source; Dynamic control steps: Based on the exergy efficiency and thermal response time constant of the coupled heat source, generate flow regulation commands and heat source switching commands to maintain the generator inlet temperature within the preset high-efficiency temperature range; Safety monitoring steps: Calculate the crystallization safety margin in real time, and trigger the anti-crystallization dilution process when the crystallization safety margin is lower than the preset safety threshold; Thermodynamic matching model through weighting coefficients By associating the physical structural parameters of the two, the following mapping relationship is derived: The system is based on the formula: Calculate exergy efficiency ,in, The ambient thermodynamic absolute temperature (K) is given. These are the thermodynamic absolute temperatures (K) of solar energy and waste heat water supply from computing power, respectively. These are the weighting coefficients; based on the formula: Calculate the time constant ,in, This represents the equivalent heat capacity of the system. The subscript eq indicates equivalent, and its calculation method is as follows: It covers the heat capacity composition of the heat exchanger's metal wall, internal fluid, and external insulation layer; The overall heat transfer coefficient is determined based on the initial experimental identification. The effective heat exchange area is the design constant, and the estimated scaling reduction factor during long-term operation has been deducted during the calculation. Weighting coefficient The calculation is based on time deviation. With system thermal inertia factor : Here The base of the natural logarithm; parameter , here Specifically refers to the adjustment coefficient, which is distinct from the physical unit of the time index. , used to characterize the asynchronous growth rate, is set to It is derived from fitting the system response curve under typical operating conditions; The maximum asynchronous tolerance time, expressed in hours (h), is proportional to the effective regulating capacity of the water tank, and the correlation satisfies the following conditions. ; The source is a preset constant, and its physical meaning is an adjustment sensitivity constant, or adjustment coefficient, to ensure the exponential term. The value is dimensionless; in this formula, the time deviation is... It is necessary to uniformly convert the data to hours (h) as the unit of input; The source is the capacity parameter of the hot water storage tank, and its physical meaning is the maximum allowable asynchronous tolerance time of the system.

2. The lithium bromide heat pump control method coupling solar thermal and computing waste heat according to claim 1, characterized in that, The calculation of the exergy efficiency of the coupled heat source includes: Analysis of source-load asynchronous characteristics: Based on the time series characteristics of solar irradiance data and computing load data, the time deviation between peak energy supply and peak heat pump demand is identified; Calculate the weighted exergy efficiency: Based on the time deviation, determine the weighting coefficients of solar thermal energy and computing waste heat, calculate the Carnot factors corresponding to the solar thermal energy temperature and computing waste heat temperature respectively, and use the weighting coefficients to perform a weighted sum of the Carnot factors of the two to obtain the exergy efficiency of the coupled heat source; the Carnot factor characterizes the work potential of the heat source temperature relative to the ambient temperature.

3. The lithium bromide heat pump control method coupling solar thermal and computing waste heat according to claim 1, characterized in that, The generation of flow regulation commands and heat source switching commands includes: If the exergy efficiency of the coupled heat source is greater than the preset efficiency threshold, the first adjustment command is generated to control the solar thermal circuit and the computing power waste heat circuit to be directly connected in series for heating. If the exergy efficiency of the coupled heat source is less than or equal to the preset efficiency threshold and greater than or equal to the minimum operating threshold, a second adjustment command is generated to control the intervention of the auxiliary heat source and adjust the mixing ratio. If the exergy efficiency of the coupled heat source is less than the minimum operating threshold, a third adjustment command is generated to cut off the heat pump unit's heating supply and switch to standby cycle mode.

4. The lithium bromide heat pump control method coupling solar thermal energy and computing waste heat according to claim 1, characterized in that, The method further includes: Before inputting the data into the multi-source data perception layer, the data is preprocessed to obtain preprocessed data; The preprocessing includes: The Kalman filter algorithm is used to smooth the solar irradiance data and eliminate high-frequency noise caused by cloud cover. The moving average method is used to extract trends from computing load data, which helps to mitigate the impact of sudden fluctuations in computing power on heat source flow.

5. The lithium bromide heat pump control method coupling solar thermal and computing waste heat according to claim 1, characterized in that, The security monitoring steps specifically include: Based on the generator inlet temperature and the concentration of the concentrated solution, the saturated crystallization temperature under the current operating conditions is calculated using the Durin equation. Calculate the difference between the current solution temperature and the saturation crystallization temperature to obtain the crystallization safety margin; The crystallization safety margin is compared with the preset safety threshold: if the crystallization safety margin is less than the preset safety threshold, it is determined that there is a crystallization risk; if the crystallization safety margin is greater than or equal to the preset safety threshold, it is determined that the system is safe to operate. coefficients in the formula The Durin plot coefficients, derived from experimental fitting data, correspond to the fitting slope and bias term in different temperature ranges, respectively; the approximate fitting formula is as follows: in, The source is the calculation result, and the physical meaning is the saturation crystallization temperature; This represents the solution concentration and must be entered as a percentage value when performing calculations. The source is an empirical constant obtained by performing least-squares nonlinear regression fitting on standard solubility data of lithium bromide aqueous solution within the working concentration range, with the following values: , , , The vertical asymptote characterizing the crystallization curve ensures formula convergence within the working concentration range.

6. The lithium bromide heat pump control method coupling solar thermal energy and computing waste heat according to claim 5, characterized in that, The anti-crystallization dilution process includes: In response to the determination of a crystallization risk, a melting and heating command is generated to increase the inlet temperature of the generator; Simultaneously, a bypass adjustment command is generated to open the refrigerant water bypass valve, directly mixing the refrigerant water into the concentrated solution circuit to reduce the concentration of the concentrated solution until the crystallization safety margin is restored to above the preset recovery threshold.

7. The lithium bromide heat pump control method coupling solar thermal energy and computing power waste heat according to claim 1, characterized in that, The process of generating flow regulation commands based on the exergy efficiency of the coupled heat source and the thermal response time constant includes: Calculate the thermal response time constant, which is defined as the time required for the system to reach 63.2% of the steady-state value from the time it receives the signal of a step change in the temperature of the heat source; The execution step size of the flow regulation command is corrected based on the thermal response time constant. If the thermal response time constant is less than the preset response threshold, the first step of the long adjustment strategy is adopted to quickly respond to fluctuations. If the thermal response time constant is greater than or equal to the preset response threshold, a gradual adjustment strategy with a second step length less than the first step length is adopted to prevent system oscillation.

8. The lithium bromide heat pump control method coupling solar thermal energy and computing waste heat according to claim 1, characterized in that, The method further includes: Establish a computing power heat dissipation guarantee mechanism to forcibly activate the backup heat dissipation bypass when the heat pump unit fails or is shut down for maintenance; Monitor the rate of change of computing load data. If the rate of change is greater than the preset surge threshold, prioritize locking the flow of the computing waste heat loop and adjust the flow of the solar thermal loop to balance the total heat input and ensure that the cooling of computing equipment is not interrupted. If the rate of change is less than or equal to the preset surge threshold, maintain the current flow adjustment strategy.

9. A lithium bromide heat pump control method coupling solar thermal and computing waste heat according to claim 2, characterized in that, The analysis of the asynchronous characteristics of the source load specifically includes: The solar irradiance and computing load within a preset time period are predicted using a long short-term memory network model. Based on the prediction results, the supply and demand matching index for a future preset time period is calculated. The supply and demand matching index is the ratio of the predicted solar energy supply to the computing power heat load demand. If the supply and demand matching index is less than the preset matching threshold, a thermal storage device intervention command is generated in advance to use the thermal storage device to smooth out asynchronous fluctuations in the source and load. Define the input tensor structure of the Long Short-Term Memory (LSTM) network model as follows: ,in, Batch size is set to 1 during the online inference phase and 64 during the offline training phase; time step... Set as Feature Dimension Set as Among them, in constructing the input tensor At that time, the feature channel index of the last dimension is clearly defined. Corresponding to the normalized solar irradiance data sequence, index The normalized computing load data sequence is used to ensure feature alignment between the inference and training phases; the model output layer is configured to predict the data sequence for the next 4 hours, with a time resolution of 15 minutes, i.e., the output sequence length is 16 time steps. The LSTM model contains two hidden layers. The first hidden layer has 64 nodes and uses the tanh activation function. The second hidden layer has 32 nodes and uses a linear activation function to connect to the output layer.

Images