Intelligent data monitoring and analysis method and system for solar desalination system

Abstract

The present application provides an intelligent data monitoring and analysis method and system for a solar desalination system. The method includes: calculating the condensation efficiency based on the phase change characteristic data and ambient temperature data of the steam in the condenser tube during the condensation process and the annular temperature gradient distribution data and the phase change characteristic data; inputting the condensation efficiency and ambient temperature data into a pre-trained reinforcement learning parameter optimization model, and generating the flow regulation parameters and guide plate deflection angle adjustment parameters of each partition in combination with the physical constraints of the condensation structure; converting the flow regulation parameters and guide plate deflection angle adjustment parameters of each partition into corresponding electric regulating valve opening instructions, and adjusting the regulating valve opening and the angle of the adjustable angle guide plate of the cooling water circulation system corresponding to different partitions by using reinforcement learning to optimize the cooling water flow and guide plate deflection angle in real time, and realizing efficient and energy-saving adaptive control of the condensation system through temperature and phase change data monitoring.

Description

Technical Field

[0001] The present application relates to the technical field of seawater desalination, and in particular to an intelligent data monitoring and analysis method and system for a solar seawater desalination system. Background Art

[0002] The condensation stage of a solar desalination system is a key link in determining the overall energy efficiency. Its technical requirements focus on dynamic monitoring and intelligent control of the thermodynamic state of the multi-partition condensation structure. Due to the strong coupling relationship between the annular temperature gradient distribution of the condenser tube wall, the steam phase change characteristics and the ambient temperature changes, it is necessary to collect multi-dimensional data in real time for the system, and based on this, dynamically adjust the cooling water flow and the deflection angle of the guide plate to balance the condensation efficiency and energy consumption. Especially under complex working conditions such as light intensity fluctuations and changes in seawater salinity, traditional static control strategies are difficult to adapt to dynamic heat loads. There is an urgent need for a method that can autonomously perceive, model and optimize the condensation process to achieve adaptive operation under multi-parameter coordination.

[0003] A representative solution to this demand is a multi-sensor feedback system based on Proportional Integral Derivative Control (PID). This solution deploys a temperature sensor array in different zones of the condenser tube wall to obtain annular temperature gradient data in real time. It then combines the steam flow rate sensor and the condensation efficiency calculation model to construct a closed-loop feedback control loop. The system uses a PID algorithm to dynamically adjust the opening of the electric valve of the cooling water circulation system and implements zoned flow distribution through a preset rule for the deflector deflection angle. For example, when the temperature gradient in a zone exceeds a threshold, the PID controller adjusts the valve opening increment based on historical data, while the deflector deflection angle is gradually corrected according to a preset step size to improve the uniformity of steam distribution.

[0004] However, the defects of the existing scheme are mainly reflected in the lack of adaptability between static rules and dynamic working conditions. First, the parameters of the PID controller rely on manual experience to set, and it is impossible to autonomously optimize the adjustment strategy according to the dynamic correlation between condensation efficiency and phase change characteristics, resulting in delayed response when the ambient temperature changes suddenly or the steam load fluctuates, and a large loss of energy efficiency. Secondly, the guide plate deflection angle adjustment rule is based on a fixed step size and threshold implementation, lacking quantitative modeling of the thermal inertia differences among multiple partitions, making it difficult to achieve global optimal thermal field balance. For example, when the thermal inertia coefficients of the condenser tube wall partitions vary greatly, fixed step size adjustment may cause local overcooling or overheating, exacerbating energy loss. In addition, the system does not integrate the collaborative optimization mechanism of condensation efficiency and ambient temperature, resulting in a bottleneck in energy efficiency improvement. Summary of the Invention

[0005] The present application provides an intelligent data monitoring and analysis method and system for a solar desalination system, which is used to solve the problems in the prior art of insufficient autonomous parameter optimization capability under dynamic working conditions, lack of quantitative modeling of multi-zone thermal inertia differences, and lack of a mechanism for coordinated optimization of condensation efficiency and ambient temperature.

[0006] In a first aspect, the present application provides an intelligent data monitoring and analysis method for a solar desalination system, comprising:

[0007] During the condensation phase of the solar desalination system, data on the annular temperature gradient distribution of the condenser tube wall, data on the phase change characteristics of the steam in the condenser tube during the condensation process, and ambient temperature data are obtained. The condenser tube wall is embedded with corresponding condensation structures in different partitions, and each condensation structure includes a cooling water circulation system and an adjustable angle guide plate.

[0008] Calculating condensation efficiency based on the annular temperature gradient distribution data and the phase change characteristic data;

[0009] Inputting the condensation efficiency and the ambient temperature data into a pre-trained reinforcement learning parameter optimization model, and combining the physical constraints of the condensation structure to generate flow regulation parameters and guide plate deflection angle regulation parameters for each partition;

[0010] The flow control parameters and guide plate deflection angle control parameters of each partition are converted into corresponding electric control valve opening instructions respectively, so as to synchronously adjust the control valve opening of the cooling water circulation system corresponding to different partitions and the angle of the adjustable angle guide plate.

[0011] Optionally, the step of inputting the condensation efficiency and the ambient temperature data into a pre-trained reinforcement learning parameter optimization model, and combining the physical constraints of the condensation structure to generate flow adjustment parameters and guide plate deflection angle adjustment parameters for each partition includes:

[0012] The condensation efficiency and ambient temperature data are temporally and spatially bound according to the partition identification code, and input into the feature extraction module of the reinforcement learning parameter optimization model to extract the dynamic feature sequence and thermal inertia coefficient;

[0013] The dynamic feature sequence is input into a dual-channel strategy network of a reinforcement learning parameter optimization model, wherein the first channel strategy network generates initial flow parameters of the cooling water circulation system corresponding to each partition through a fully connected layer and a flow constraint layer, and the second channel strategy network generates initial deflection angle parameters of the adjustable angle guide plate corresponding to each partition through a convolutional layer and an angle constraint layer. The flow constraint layer dynamically adjusts the constraint boundary based on the flow constraint strategy output by the reinforcement learning agent, and the angle constraint layer dynamically adjusts the constraint boundary based on the angle constraint strategy output by the reinforcement learning agent;

[0014] Using the initial flow parameters and initial deflection angle parameters as state inputs of the reinforcement learning agent, receiving parameter correction actions output by the reinforcement learning agent, performing parameter correction based on the physical constraint conditions, and generating intermediate flow parameters and intermediate deflection angle parameters;

[0015] Inputting the intermediate flow parameter and the intermediate deflection angle parameter into a multi-objective optimization module of a reinforcement learning parameter optimization model for parameter optimization;

[0016] The optimized intermediate flow parameters, optimized intermediate deflection angle parameters, and historical optimal parameters are partitioned and cascaded and fused. The fusion weight is dynamically allocated according to the thermal inertia coefficient of the condensation structure of each partition to generate the flow regulation parameters and guide plate deflection angle regulation parameters of each partition.

[0017] Optionally, the dynamic feature sequence is input into a dual-channel strategy network of a reinforcement learning parameter optimization model, wherein the first channel strategy network generates initial flow parameters of the cooling water circulation system corresponding to each partition through a fully connected layer and a flow constraint layer, and the second channel strategy network generates initial deflection angle parameters of the adjustable angle guide plate corresponding to each partition through a convolutional layer and an angle constraint layer, including:

[0018] Input each partition feature vector in the dynamic feature sequence into the fully connected layer and map it into a flow feature vector through dense connections between nodes;

[0019] The flow characteristic vector is input into the flow constraint layer, in which the maximum pressure threshold and the minimum flow velocity threshold of the pipeline of the cooling water circulation system are predefined. The flow characteristic vector is truncated element by element through the inequality constraint function to generate the initial flow parameters of the cooling water circulation system corresponding to each partition;

[0020] Inputting the two-dimensional spatial distribution features of each partition in the dynamic feature sequence into the convolution layer, performing sliding window convolution along the partition spatial dimension through multiple groups of convolution kernels, and extracting the deflection angle correlation features between adjacent partitions;

[0021] The deflection angle association characteristics of the guide plate are input into the angle constraint layer, in which the upper and lower limits of the mechanical rotation angle of the adjustable angle guide plate are predefined. The deflection angle association characteristics of the guide plate are mapped to the preset angle range through a piecewise linear mapping function to generate the initial deflection angle parameters of the adjustable angle guide plate corresponding to each partition.

[0022] Optionally, the calculating the condensation efficiency based on the annular temperature gradient distribution data and the phase change characteristic data includes:

[0023] Dividing the annular temperature gradient distribution data into a plurality of annular sub-regions at equal radial angles in the annular region, and extracting temperature gradient vector data of each annular sub-region;

[0024] Obtaining the phase change latent heat value and the upper limit of the phase change temperature range of each annular sub-region from the phase change characteristic data, and calculating the thermal response rate coefficient of each annular sub-region, wherein the thermal response rate coefficient is the ratio of the phase change latent heat value of the corresponding annular sub-region to the upper limit of the phase change temperature range;

[0025] Multiplying the temperature gradient vector data of each annular sub-region by the thermal response rate coefficient of the corresponding annular sub-region to generate a condensation efficiency contribution value of each annular sub-region;

[0026] According to the area proportion of each annular sub-region in the annular region, a weighted sum is performed on the condensation efficiency contribution value of each annular sub-region to generate the condensation efficiency.

[0027] Optionally, the flow characteristic vector is input into a flow constraint layer, in which a maximum pressure threshold and a minimum flow velocity threshold of the pipeline of the cooling water circulation system are predefined, and the flow characteristic vector is truncated element by element through an inequality constraint function to generate initial flow parameters of the cooling water circulation system corresponding to each partition, including:

[0028] Performing element-by-element constraint processing on each element of the flow characteristic vector, if the current element value is greater than the maximum allowable flow value corresponding to the maximum pressure threshold of the pipeline of the corresponding partition, then modifying the current element value to the maximum allowable flow value;

[0029] If the current element value is less than the minimum allowable flow value corresponding to the minimum flow rate threshold of the corresponding partition, the current element value is modified to the minimum allowable flow value;

[0030] If the current element value is between the minimum allowable flow rate value and the maximum allowable flow rate value, the current element value is retained unchanged;

[0031] The elements processed by element-by-element constraint are reorganized according to the partition mapping relationship to generate the initial flow parameters of the cooling water circulation system corresponding to each partition.

[0032] Optionally, using the initial flow parameter and the initial deflection angle parameter as state inputs of a reinforcement learning agent, receiving a parameter correction action output by the reinforcement learning agent, performing parameter correction based on the physical constraint conditions, and generating an intermediate flow parameter and an intermediate deflection angle parameter includes:

[0033] Comparing the initial flow parameter with the upper and lower bound values of the flow parameter;

[0034] Numerically comparing the initial deflection angle parameter with the upper constraint angle and the lower constraint angle of the deflection angle parameter;

[0035] The intermediate flow parameters are determined according to the flow comparison results, and the intermediate deflection angle parameters are determined according to the deflection angle comparison results.

[0036] Optionally, determining the intermediate flow parameter according to the flow comparison result and determining the intermediate deflection angle parameter according to the deflection angle comparison result includes:

[0037] If the flow comparison result indicates that the initial flow parameter is greater than the upper constraint limit of the flow parameter, the first fixed offset is used as the intermediate flow parameter; if the flow comparison result indicates that the initial flow parameter is less than the lower constraint limit of the flow parameter, the second fixed offset is used as the intermediate flow parameter;

[0038] If the deflection angle comparison result indicates that the initial deflection angle parameter is greater than the constrained upper limit angle of the deflection angle parameter, an intermediate deflection angle parameter is generated between the initial deflection angle parameter and the constrained upper limit angle of the deflection angle parameter through interpolation operation; if the initial deflection angle parameter is less than the constrained lower limit angle of the deflection angle parameter, an intermediate deflection angle parameter is generated between the constrained lower limit angle and the initial deflection angle parameter through interpolation operation.

[0039] In a second aspect, the present application provides an intelligent data monitoring and analysis system for a solar desalination system, comprising:

[0040] An acquisition module is used to obtain annular temperature gradient distribution data of the condenser tube wall, phase change characteristic data of the steam in the condenser tube during the condensation process, and ambient temperature data during the condensation phase of the solar desalination system, wherein the condenser tube wall is embedded with corresponding condensation structures in different partitions, and each condensation structure includes a cooling water circulation system and an adjustable angle guide plate;

[0041] a calculation module, configured to calculate a condensation efficiency based on the annular temperature gradient distribution data and the phase change characteristic data;

[0042] a generation module, configured to input the condensation efficiency and the ambient temperature data into a pre-trained reinforcement learning parameter optimization model, and generate flow adjustment parameters and guide plate deflection angle adjustment parameters for each partition in combination with the physical constraints of the condensation structure;

[0043] The regulating module is used to convert the flow regulation parameters and guide plate deflection angle regulation parameters of each partition into corresponding electric regulating valve opening instructions, so as to synchronously adjust the regulating valve opening of the cooling water circulation system corresponding to different partitions and the angle of the adjustable angle guide plate.

[0044] In a third aspect, the present application provides a computing device comprising a processing component and a storage component; the storage component stores one or more computer instructions; the one or more computer instructions are used to be called and executed by the processing component to implement an intelligent data monitoring and analysis method for a solar desalination system as described in any one of the first aspects.

[0045] In a fourth aspect, the present application provides a computer storage medium storing a computer program, which, when executed by a computer, implements an intelligent data monitoring and analysis method for a solar desalination system as described in any one of the first aspects.

[0046] In an embodiment of the present application, an intelligent data monitoring and analysis method for a solar desalination system is provided, the method comprising: obtaining annular temperature gradient distribution data of the condenser tube wall, phase change characteristic data of the steam in the condenser tube during the condensation process, and ambient temperature data during the condensation stage of the solar desalination system, wherein corresponding condensation structures are embedded in different partitions inside the condenser tube wall, and each of the condensation structures comprises a cooling water circulation system and an adjustable angle guide plate; calculating the condensation efficiency based on the annular temperature gradient distribution data and the phase change characteristic data; inputting the condensation efficiency and the ambient temperature data into a pre-trained reinforcement learning parameter optimization model, and generating flow regulation parameters and guide plate deflection angle adjustment parameters for each partition in combination with the physical constraints of the condensation structure; converting the flow regulation parameters and guide plate deflection angle adjustment parameters of each partition into corresponding electric control valve opening instructions, so as to synchronously adjust the control valve opening of the cooling water circulation system and the angle of the adjustable angle guide plate corresponding to different partitions.

[0047] This application uses real-time monitoring of multi-zone embedded condensation structures to accurately capture the temperature field distribution of the condenser tube wall and the dynamic characteristics of the steam phase change, solving the problem of inaccurate identification of local thermodynamic states caused by a single temperature measurement point or global average measurement in traditional systems, and providing high-resolution data support for subsequent refined regulation. Combining the dynamic characteristics of the steam phase change process and the temperature gradient distribution, a quantitative evaluation of the condensation efficiency at the zone level is achieved, breaking through the limitations of traditional methods that rely on overall efficiency estimation, providing a scientific basis for differentiated regulation of zones, and avoiding the drag of local inefficient areas on the overall performance of the system. The reinforcement learning model is used to integrate real-time environmental variables and physical constraints to dynamically optimize the flow rate of each zone and the deflection angle of the guide plate, solving the problem that traditional PID or rule control cannot adapt to nonlinear heat load changes, and improving the system's adaptability under variable working conditions. Through the coordinated action of the electric control valve and the guide plate, precise matching of multi-zone flow and airflow organization is achieved, overcoming the problems of uneven condensation or energy waste caused by traditional single-point regulation, and improving the system's energy efficiency and fresh water output stability.

[0048] Furthermore, by spatiotemporally binding condensation efficiency and ambient temperature data, the feature extraction module of the reinforcement learning model is input to extract dynamic feature sequences and thermal inertia coefficients. A dual-channel strategy network is used to separately optimize flow and guide plate parameters. Specifically, the first channel generates initial flow parameters through a fully connected layer and a dynamic flow constraint layer, while the second channel generates initial guide plate deflection angle parameters through a convolutional layer and an angle constraint layer. Both channels dynamically adjust constraint boundaries based on the reinforcement learning agent. The initial parameters are then constrained and optimized through multi-objective optimization. Finally, the historical optimal parameters are dynamically fused with the thermal inertia coefficient to output the optimal control parameters at the partition level. The flow constraint layer ensures safe pipeline operation through inequality functions, the angle constraint layer ensures mechanical limits through piecewise linear mapping, and the convolutional layer exploits spatial correlation features between adjacent partitions. Through dynamic feature extraction and dual-channel collaborative optimization, precise decoupling and regulation of flow rate and guide plate parameters are achieved, solving the response hysteresis problem caused by parameter coupling in traditional methods; dynamic boundary adjustment of the constraint layer ensures the physical safety of the system and avoids the risk of overload; parameter fusion based on thermal inertia coefficient improves the adaptability to nonlinear thermodynamic processes, especially optimizes the control stability under variable working conditions; the introduction of spatial correlation features enhances the multi-partition collaborative control capability, effectively reduces local condensation unevenness, and overall improves the system energy efficiency and water production consistency.

[0049] These and other aspects of the present application will become more readily apparent from the description of the following embodiments. BRIEF DESCRIPTION OF THE DRAWINGS

[0050] In order to more clearly illustrate the embodiments of the present application or the technical solutions in the prior art, a brief introduction will be given below to the drawings required for use in the embodiments or the description of the prior art. Obviously, the drawings described below are some embodiments of the present application. For ordinary technicians in this field, other drawings can be obtained based on these drawings without any creative work.

[0051] Figure 1 A flow chart of an intelligent data monitoring and analysis method for a solar desalination system provided in an embodiment of the present application;

[0052] Figure 2 A schematic diagram of the structure of an intelligent data monitoring and analysis system for a solar desalination system provided in an embodiment of the present application;

[0053] Figure 3 A schematic diagram of the structure of a computing device provided in an embodiment of the present application. DETAILED DESCRIPTION

[0054] In order to enable those skilled in the art to better understand the solution of the present application, the technical solution in the embodiments of the present application will be clearly and completely described below in conjunction with the drawings in the embodiments of the present application.

[0055] In some of the processes described in the specification and claims of this application and the above-mentioned figures, multiple operations that appear in a specific order are included, but it should be clearly understood that these operations may not be executed in the order in which they appear in this document or may be executed in parallel. The serial numbers of the operations, such as 11, 12, etc., are only used to distinguish between different operations, and the serial numbers themselves do not represent any order of execution. In addition, these processes may include more or fewer operations, and these operations may be executed in sequence or in parallel. It should be noted that the descriptions of "first", "second", etc. in this document are used to distinguish different messages, devices, modules, etc., and do not represent a sequential order, nor do they limit "first" and "second" to being different types.

[0056] The following will be combined with the drawings in the embodiments of this application to clearly and completely describe the technical solutions in the embodiments of this application. Obviously, the embodiments described are only part of the embodiments of this application, not all of the embodiments. Based on the embodiments in this application, all other embodiments obtained by those skilled in the art without making creative efforts are within the scope of protection of this application.

[0057] In order to solve the problems in the prior art of insufficient autonomous parameter optimization capability under dynamic working conditions, lack of quantitative modeling of multi-zone thermal inertia differences, and lack of collaborative optimization mechanism of condensation efficiency and ambient temperature, the embodiment of the present application provides an intelligent data monitoring and analysis method for a solar desalination system, which adopts the following concept: for the efficiency optimization problem of the condensation stage of the solar desalination system, first, the annular temperature gradient distribution, steam phase change characteristics and ambient temperature data of the condenser tube wall are collected in real time through a multi-zone embedded condensation structure to accurately perceive the local thermodynamic state; a partitioned condensation efficiency calculation model is established based on the temperature gradient and phase change data to quantitatively evaluate the performance differences of each region; then the condensation efficiency and environmental parameters are input into a pre-trained reinforcement learning optimization model, and combined with the physical constraints of the cooling water circulation system and the guide plate, the optimal control parameters of the flow rate and the deflection angle of each partition are dynamically generated; finally, the parameters are converted into electric control valve opening instructions to achieve multi-zone synchronous and precise adjustment of the cooling water flow and the diversion angle, thereby breaking through the bottleneck of traditional methods' control lag and uneven energy efficiency under dynamic working conditions, and improving the overall condensation efficiency and stability of the system.

[0058] Figure 1 This is a flow chart of an intelligent data monitoring and analysis method for a solar desalination system provided in an embodiment of the present application, such as Figure 1 As shown, the method includes:

[0059] S11. During the condensation stage of the solar desalination system, the annular temperature gradient distribution data of the condenser tube wall, the phase change characteristic data of the steam in the condenser tube during the condensation process, and the ambient temperature data are obtained. Corresponding condensation structures are embedded in different partitions inside the condenser tube wall, and each condensation structure includes a cooling water circulation system and an adjustable angle guide plate.

[0060] Among them, the annular temperature gradient distribution data refers to the temperature values and spatial change rates of each temperature measuring point along the circumference of the condenser tube. The phase change characteristic data includes physical quantities of the phase change process such as steam condensation rate, liquid film thickness, and latent heat release. The ambient temperature data refers to the real-time air temperature of the space where the condensing device is located when the solar seawater desalination system is running. The data source is collected through an anti-corrosion temperature sensor or a multi-point infrared temperature measurement module arranged on the outer wall of the condenser tube, and the sampling frequency is ≥1Hz. The cooling water circulation system may refer to an active heat exchange subsystem integrated in each condensation partition. The adjustable angle guide plate refers to a metal guide blade that can be rotated 0-90° around the axis, which is used to control the direction of the condensing airflow.

[0061] In an embodiment of the present application, during the condensation stage of the solar desalination system, annular temperature gradient distribution data is obtained by arranging a temperature sensor array in each partition of the condenser tube wall. At the same time, a steam phase change monitoring device is used to collect phase change characteristic data of the steam in the condenser tube, and the ambient temperature data is recorded synchronously. Among them, the condenser tube wall adopts a partitioned embedded design, and each partition integrates an independent condensation structure, which includes a cooling water circulation system and an adjustable angle guide plate. During data collection, the temperature sensors are arranged in a circular array and sampled every 5 seconds. The steam phase change data can be captured in real time by a high-precision optical sensor.

[0062] S12. Calculate the condensation efficiency based on the annular temperature gradient distribution data and the phase change characteristic data.

[0063] Among them, condensation efficiency refers to the ratio of the fresh water energy obtained by actual condensation to the input steam latent heat.

[0064] In this embodiment, the annular temperature gradient data is first averaged and its standard deviation calculated for each partition to characterize the temperature field uniformity. The condensation rate in the vapor phase change characteristic data is then divided by the theoretical maximum condensation rate for that partition to obtain the phase change efficiency coefficient. Finally, the temperature field uniformity index is multiplied by the phase change efficiency coefficient and then multiplied by a normalized weighting factor to output a partitioned condensation efficiency value within the range of 0-1. Optionally, to improve computational efficiency, the computation process for each partition can be performed in parallel.

[0065] S13. Input the condensation efficiency and ambient temperature data into a pre-trained reinforcement learning parameter optimization model, and generate flow adjustment parameters and guide plate deflection angle adjustment parameters for each partition in combination with the physical constraints of the condensation structure.

[0066] Among them, the reinforcement learning parameter optimization model is an intelligent decision-making system based on the proximal strategy optimization algorithm, which is specifically used to dynamically optimize the condensation control parameters of the solar desalination system. Physical constraints refer to the maximum / minimum flow, pressure, angle and other limit parameters allowed by the mechanical structure of the system. The flow regulation parameter refers to the optimization instruction output by the reinforcement learning model for dynamically adjusting the flow of the cooling water circulation system. The guide plate deflection angle adjustment parameter refers to the instruction output by the reinforcement learning model for optimizing the position of the adjustable angle guide plate. Therefore, S13 can input the condensation efficiency and the ambient temperature data into the reinforcement learning parameter optimization model pre-trained based on the proximal strategy optimization algorithm, and generate the flow regulation parameters and guide plate deflection angle adjustment parameters of each partition in combination with the physical constraints of the condensation structure; the input dimension of the reinforcement learning parameter optimization model is the joint feature vector of the condensation efficiency, ambient temperature and partition identification code, and the output dimension is the flow parameter and deflection angle parameter of each partition.

[0067] In an embodiment of the present application, the condensing efficiency and ambient temperature data are encoded by partition and then input into a pre-trained reinforcement learning parameter optimization model. The model first verifies the validity of the input data through a physical constraint check module. The constraints include the maximum flow limit of the cooling water circulation system and the mechanical angle range of the guide plate. The model's dual-branch output layer generates flow regulation parameters and guide plate deflection angle adjustment parameters, respectively. The flow parameters are mapped to the constraint range using an s-function, and the guide plate parameters are corrected using piecewise linear interpolation. Optionally, the embodiment of the present application can update the parameter matrix every 30 seconds.

[0068] S14. Convert the flow control parameters and guide plate deflection angle control parameters of each partition into corresponding electric control valve opening instructions to synchronously adjust the control valve opening and the angle of the adjustable angle guide plate of the cooling water circulation system corresponding to different partitions.

[0069] Among them, the electric control valve opening instruction refers to the standardized electrical signal that controls the degree of valve opening.

[0070] In this embodiment of the present application, the flow control parameters for each zone are converted into analog signals to drive the opening of the electric control valve. Simultaneously, the deflector deflection angle parameters are controlled by a pulse width modulation (PWM) signal to control the rotation of the servo motor. A PID synchronization algorithm can be used during the adjustment process to ensure that all zones are adjusted within 2 seconds. The cooling water flow rate change rate is limited to ±10% / s, and the deflector angular velocity does not exceed 5° / s to avoid hydraulic shock or mechanical vibration. The system provides real-time feedback of the valve opening and deflector deflection angle to the central controller, forming a closed loop.

[0071] Here's a specific example: After deploying this system at a solar desalination plant, an annular temperature gradient was detected in zone 3 during a period of intense midday sunlight, causing its condensation efficiency to drop to 0.65. The reinforcement learning model generated parameters based on this information: increasing the zone's flow rate by 8.2 L / min and adjusting the deflector angle to 52°. Within two seconds, the efficiency was restored to 0.82. During the same period, shadowing caused a uniform temperature gradient in zone 7. The system automatically reduced its flow rate by 5% and increased the deflector angle to 65°, maintaining a stable overall water production rate of 8.2 L / (m²·h).

[0072] By executing S11 to S14, and through multi-source data fusion and reinforcement learning dynamic optimization, precise zone control of the condensation process is achieved: the collaborative analysis of temperature gradient and phase change data improves the accuracy of efficiency calculations; the parameter generation of the reinforcement learning model under physical constraints ensures system safety and avoids overload failures; the synchronous control of the electric valve and guide plate shortens the operating condition adjustment time and improves steam utilization.

[0073] Therefore, the embodiment of the present application accurately captures the temperature field distribution of the condenser tube wall and the dynamic characteristics of the steam phase change through real-time monitoring of the multi-zone embedded condensation structure, solves the problem of inaccurate identification of local thermodynamic states caused by a single temperature measurement point or global average measurement in traditional systems, and provides high-resolution data support for subsequent refined regulation. Combining the dynamic characteristics of the steam phase change process and the temperature gradient distribution, a quantitative evaluation of the condensation efficiency at the zone level is achieved, breaking through the limitations of traditional methods that rely on overall efficiency estimation, providing a scientific basis for differentiated regulation of zones, and avoiding the drag of the overall system performance due to local inefficient areas. The reinforcement learning model is used to integrate real-time environmental variables and physical constraints, and dynamically optimize the flow rate of each zone and the deflection angle parameters of the guide plate, solving the problem that traditional PID or rule control cannot adapt to nonlinear heat load changes, and improving the system's adaptability under variable working conditions. Through the coordinated action of the electric control valve and the guide plate, precise matching of multi-zone flow and airflow organization is achieved, overcoming the problems of uneven condensation or energy waste caused by traditional single-point regulation, and improving the system's energy efficiency and fresh water output stability.

[0074] In one possible embodiment, S13, the condensation efficiency and ambient temperature data are input into a pre-trained reinforcement learning parameter optimization model, and the flow adjustment parameters and guide plate deflection angle adjustment parameters for each partition are generated in combination with the physical constraints of the condensation structure, including:

[0075] Step 131: Bind the condensing efficiency and ambient temperature data in time and space according to the partition identification code, and input them into the feature extraction module of the reinforcement learning parameter optimization model to extract the dynamic feature sequence and thermal inertia coefficient.

[0076] Spatiotemporal binding involves encoding data from different partitions by timestamp and spatial location, forming a data matrix with spatiotemporal attributes. A dynamic feature sequence is a multidimensional set of time-series features extracted from the spatiotemporally bound condensing efficiency and ambient temperature data, reflecting the system's real-time operating status. The thermal inertia coefficient is a dimensionless parameter that measures the thermal response delay of a partition.

[0077] In this embodiment, condensation efficiency and ambient temperature data are first spatially and temporally bound according to the partition identifier code and then input into the feature extraction module of the reinforcement learning parameter optimization model. This feature extraction module can employ a hybrid structure of a convolutional neural network and a long-short-term memory network. The convolutional layer extracts the local spatial correlations between temperature gradients and phase change features, while the long-short-term memory network layer captures the temporal dependencies of ambient temperature changes. Ultimately, the module outputs a dynamic feature sequence and thermal inertia coefficient.

[0078] In other embodiments, the multiplication result of the time interval and the heat flow is calculated, and the ratio of the partition temperature change to the multiplication result is used as the thermal inertia coefficient. The constraint boundaries of the flow constraint layer and the angle constraint layer are dynamically adjusted based on the real-time feedback of the pipeline pressure sensor data and the mechanical limit data of the guide plate angle sensor.

[0079] Step 132: Input the dynamic feature sequence into the dual-channel strategy network of the reinforcement learning parameter optimization model, wherein the first channel strategy network generates the initial flow parameters of the cooling water circulation system corresponding to each partition through the fully connected layer and the flow constraint layer, and the second channel strategy network generates the initial deflection angle parameters of the adjustable angle guide plate corresponding to each partition through the convolution layer and the angle constraint layer. The flow constraint layer dynamically adjusts the constraint boundary based on the flow constraint strategy output by the reinforcement learning agent, and the angle constraint layer dynamically adjusts the constraint boundary based on the angle constraint strategy output by the reinforcement learning agent.

[0080] The dual-channel policy network is a neural network architecture that processes different control objectives in parallel. It generates flow parameters and guide vane deflection angle adjustment parameters through two independent channels. A fully connected layer is a neural network layer whose nodes are densely connected to all nodes in the previous layer and is used for global feature integration and nonlinear mapping. The flow constraint layer is a neural network layer that ensures that the flow does not exceed the pipeline pressure and flow velocity limits. The initial flow parameter is the preliminary control value output by the first channel of the dual-channel policy network and represents the recommended flow distribution for the cooling water circulation system. The convolution layer extracts local spatial features through a sliding window operation of the convolution kernel and is suitable for processing data with spatial correlation. The angle constraint layer is a logic module that limits the guide vane deflection angle to within a mechanically feasible range. The initial deflection angle parameter is the preliminary angle recommendation output by the second channel of the dual-channel policy network and represents the recommended deflection angle for each zoned guide vane. The reinforcement learning agent can be an intelligent decision-making module using a deep Q-network or a proximal policy optimization algorithm, outputting parameter adjustment actions.

[0081] In an embodiment of the present application, a dynamic feature sequence is input into a dual-channel strategy network. The first channel strategy network maps the feature vector to a high-dimensional space through a fully connected layer to generate the initial value of the flow parameter; the flow constraint layer is based on the flow constraint strategy output by the reinforcement learning agent, and dynamically adjusts the parameter boundary through the inequality truncation function to generate the initial flow parameters of the cooling water circulation system of each partition. The second channel strategy network extracts the associated features of the deflection angles of the guide plates of adjacent partitions through a convolution layer. The angle constraint layer uses a piecewise linear mapping function to convert the features to a preset interval according to the upper and lower limits of the mechanical rotation angle, and generates the initial deflection angle parameters of the adjustable angle guide plates of each partition. The dual-channel design realizes the decoupling optimization of flow and angle. It should be noted that the embodiment of the present application does not specifically limit the expressions corresponding to the inequality truncation function and the piecewise linear mapping function.

[0082] Step 133: Use the initial flow parameters and initial deflection angle parameters as state inputs of the reinforcement learning agent, receive the parameter correction action output by the reinforcement learning agent, perform parameter correction based on physical constraints, and generate intermediate flow parameters and intermediate deflection angle parameters.

[0083] The parameter correction action refers to the flow / angle adjustment instruction output by the reinforcement learning agent. The intermediate flow parameters and intermediate deflection angle parameters are the correction results of the reinforcement learning agent on the initial flow parameters and initial deflection angle parameters.

[0084] In the embodiments of the present application, initial flow parameters and initial deflection angle parameters are used as state inputs for a reinforcement learning agent. The agent uses a commonly used deep deterministic policy gradient algorithm to correct actions based on the output parameters and evaluate the value of the actions. Corrected actions must satisfy physical constraints, and the commonly used constraint projection algorithm is used to constrain the corrected parameters to the feasible domain, ultimately generating intermediate flow parameters and intermediate deflection angle parameters. This process achieves a balance between local optimality and global constraints.

[0085] Step 134: Input the intermediate flow parameters and the intermediate deflection angle parameters into a multi-objective optimization module of the reinforcement learning parameter optimization model for parameter optimization. The multi-objective optimization module is an optimization algorithm component that balances efficiency, energy consumption, and safety.

[0086] In this embodiment of the present application, the intermediate flow rate parameters and the intermediate deflection angle parameters are input into a multi-objective optimization module. This module, based on the Pareto optimization framework, optimizes condensation efficiency, minimizes energy consumption, and maintains thermal field balance. The expressions for these optimization objectives are conventional mathematical expressions and are not detailed here. A non-dominated sorting genetic algorithm is used to generate the parameter optimization solution set. During the optimization process, the module dynamically adjusts the objective weights based on the real-time fluctuations of ambient temperature data and outputs the optimized parameters for the multi-objective collaboration.

[0087] Step 135: Perform partition-wise cascade fusion on the optimized intermediate flow parameters, optimized intermediate deflection angle parameters, and historical optimal parameters, dynamically assign fusion weights based on the thermal inertia coefficient of each partition's condensation structure, and generate flow regulation parameters and guide plate deflection angle regulation parameters for each partition.

[0088] Among them, partition cascade fusion is the process of weighted fusion of real-time parameters and historical parameters according to partition characteristics.

[0089] In the embodiments of the present application, optimized intermediate parameters are cascaded and fused with historically optimized parameters in a partitioned manner. This fusion process can be implemented using a weighted average algorithm, with weights dynamically assigned to each partition's thermal inertia coefficient. For example, for partitions with significant thermal hysteresis, the flow control parameters are primarily based on historically optimized values, while for partitions with rapid thermal response, the real-time optimized values are primarily used. The resulting flow control parameters and guide vane deflection angle control parameters are mapped to the actuators via partition coding, enabling global dynamic control.

[0090] Here's a specific example: During midday operation at a solar desalination plant, the system detected an annular temperature gradient anomaly in zone 3, resulting in a drop in condensation efficiency to 0.65. Meanwhile, zone 7 maintained a uniform temperature of 28°C due to shadowing. The system then bound zone 3's efficiency data (0.65) to the temperature gradient data, according to the zone code. The network extracted a dynamic feature sequence and calculated a thermal inertia coefficient of 1.1. For zone 7, the data was then bound simultaneously, extracting a feature sequence and a thermal inertia coefficient of 0.9. These features were then fed into a two-channel strategy network: the first channel generated the initial flow rate parameters for zone 3, which were dynamically limited to [+5 L / min, +10 L / min] by the flow constraint layer and outputted at +8.2 L / min. The second channel generated an initial deflection angle of 50° for the guide vanes, which was limited to [45°, 60°] by the angle constraint layer and outputted at 52°. For zone 7, a 6% flow reduction and an angle of 63° were generated. Based on physical constraints, the reinforcement learning agent fine-tuned the parameters for zone 3, maintaining a constant flow rate of +8.2 L / min and a guide angle of 52 degrees. Zone 7 was ultimately determined to have a flow rate of -5% and an angle of 65 degrees. Using a multi-objective optimization module that balanced efficiency with energy consumption, the parameters for zone 3 were optimized to +8.0 L / min and a guide angle of 53 degrees, while those for zone 7 were adjusted to -4.8% and an angle of 64 degrees. The optimized parameters were then weighted and fused with historical optimal values based on the thermal inertia coefficient. The final output for zone 3 was +7.7 L / min and a guide angle of 51.5 degrees, while the output for zone 7 was -4.9% and an angle of 62 degrees. This restored the efficiency of zone 3 to 0.82 within 2 seconds, stabilizing the water production rate of the all-solar desalination plant at 8.2 L / (m²·h).

[0091] By executing steps 131 to 135, the embodiment of the present application accurately depicts the system state through spatiotemporal binding and dynamic feature extraction, combines the dual-channel strategy network to optimize the flow rate and guide plate deflection angle respectively, utilizes the reinforcement learning agent and the multi-objective optimization module to achieve automatic parameter correction and global optimization, and finally dynamically weights and fuses historical data through the thermal inertia coefficient to improve the adaptability and energy efficiency of the condensing system while reducing equipment losses.

[0092] In one possible embodiment, step 132 inputs the dynamic feature sequence into a dual-channel strategy network of a reinforcement learning parameter optimization model, wherein the first channel strategy network generates initial flow parameters of the cooling water circulation system corresponding to each partition through a fully connected layer and a flow constraint layer, and the second channel strategy network generates initial deflection angle parameters of the adjustable angle guide plate corresponding to each partition through a convolutional layer and an angle constraint layer, including:

[0093] Step a1: Input each partition feature vector in the dynamic feature sequence into the fully connected layer and map it into a traffic feature vector through dense connections between nodes. The traffic feature vector is the traffic control-related feature encoded by the fully connected layer.

[0094] In this embodiment, the feature vectors of each partition in the dynamic feature sequence are input into a fully connected layer. This layer performs nonlinear transformations through dense connections between nodes and uses an activation function to map the features into a low-dimensional flow feature vector. This vector implicitly captures key factors required for flow regulation, such as pressure, flow rate, and heat exchange rate, providing the basis for subsequent flow constraints.

[0095] Step a2: Input the flow characteristic vector into the flow constraint layer. The flow constraint layer predefines the maximum pressure threshold and minimum flow rate threshold of the cooling water circulation system pipeline. The flow characteristic vector is truncated element by element through the inequality constraint function to generate the initial flow parameters of the cooling water circulation system corresponding to each partition.

[0096] Among them, the formula of the inequality constraint function is ,in, is the element of the ith partition of the traffic feature vector, is the flow upper limit threshold corresponding to the maximum pressure of the pipeline, is the minimum flow rate threshold allowed by the system, is the initial flow parameter corresponding to the i-th partition. Truncation refers to the operation of forcibly limiting the flow recommendation value that exceeds the physically feasible range to the preset boundary. If the initial parameter exceeds , then it is mandatory to take . Truncate downwards: If the initial parameter is lower than , then it is mandatory to take .

[0097] In this embodiment, the flow feature vector is input to the flow constraint layer, which predefines physical constraints: a maximum pressure threshold and a minimum flow velocity threshold for the pipeline. An inequality constraint function is used to clip the flow feature vector element by element, generating realistic initial flow parameters.

[0098] Step a3: Input the two-dimensional spatial distribution features of each partition in the dynamic feature sequence into the convolution layer, perform sliding window convolution along the partition space dimension through multiple groups of convolution kernels, and extract the associated features of the deflection angles of the guide plates between adjacent partitions.

[0099] The two-dimensional spatial distribution features refer to the thermal maps or distribution matrices of parameters such as the partitioned temperature field and the deflector deflection angle on a two-dimensional plane. The partitioned spatial dimension refers to the physical spatial arrangement of the condensation partitions in a solar desalination plant, with each partition having fixed spatial coordinates. Sliding window convolution involves sliding the convolution kernel along the partitioned spatial dimension at a fixed step size, calculating the weighted sum of local region features at each position. Features associated with the deflector deflection angle can include: airflow interference coefficient, temperature gradient propagation factor, efficiency coupling weight, mechanical stress distribution characteristics, etc.

[0100] In this embodiment, the two-dimensional spatial distribution features in the dynamic feature sequence are fed into a convolutional layer. A 3×3 convolution kernel is used to perform a sliding window calculation along the partitioned space to extract the correlation features of the deflector angles between adjacent partitions. Multiple sets of convolution kernels extract different spatial patterns in parallel, and the resulting fused angle correlation feature tensor is output.

[0101] Step a4: Input the deflection angle-related features of the guide plate into the angle constraint layer. The angle constraint layer predefines the upper and lower limits of the mechanical rotation angle of the adjustable angle guide plate. The deflection angle-related features of the guide plate are mapped to the preset angle range through a piecewise linear mapping function to generate the initial deflection angle parameters of the adjustable angle guide plate corresponding to each partition.

[0102] Among them, the formula of the piecewise linear mapping function is ,in, is the original angle suggestion value output by the convolution layer, is the initial deflection angle parameter, is the lower limit of the mechanical rotation of the guide plate, is the upper limit of the mechanical rotation of the guide plate, is the first segmentation threshold, is the second segmentation threshold, is the third segmentation threshold, is the slope of the rising segment, is the slope of the descending segment, where , and The value of can be a constant.

[0103] In the embodiment of the present application, the deflector deflection angle-related features are input into the angle constraint layer, which predefines the mechanical limits. The features are mapped to a preset interval through a piecewise linear mapping function to generate the initial deflection angle parameters.

[0104] The following is a specific example: In a solar desalination plant, the 12-dimensional dynamic features of partition 3 are fed into a fully connected layer, which outputs an 8-dimensional flow feature vector. A constraint layer truncates the vector element [1.1], corresponding to a flow rate of +9 L / min, to +8.2 L / min. Convolution of the temperature field matrix extracts the airflow interference feature 0.7 between partition 3 and its adjacent partition 4. This feature is mapped to 52°, ultimately generating control instructions for a flow rate of +8.2 L / min for partition 3 and a guide plate angle of 52°, restoring efficiency to 0.82 within 2 seconds.

[0105] By executing steps a1-a4, this embodiment of the application extracts flow control features and deflector deflection angle-related features through fully connected layers and convolutional layers, respectively. Combined with the dynamic boundary constraints of the physical constraint layer, this ensures that the generated control parameters always meet the system's mechanical load capacity and operational safety requirements. This method achieves collaborative optimization of multi-zone parameters. While ensuring the safe operation of the equipment, it uses feature-driven intelligent adjustment to achieve the optimal match between each zone's flow rate and the deflector deflection angle, effectively improving the overall system stability and control response speed while avoiding airflow interference or pressure fluctuations between zones.

[0106] In a possible embodiment, S12, calculating the condensation efficiency based on the annular temperature gradient distribution data and the phase change characteristic data, includes:

[0107] Step 121 : Divide the annular temperature gradient distribution data into a number of annular sub-regions at equal radial angles in the annular region, and extract the temperature gradient vector data of each annular sub-region.

[0108] The annular sub-region is a sector-shaped partition divided at a fixed angle and used for local temperature analysis. The temperature gradient vector data is a vector containing the direction and intensity of temperature change.

[0109] In the embodiment of the present application, the annular temperature gradient distribution data is divided into a number of annular sub-areas at equal angles along the radial direction of the annular area. The temperature gradient vector data of each annular sub-area is extracted using a temperature field interpolation algorithm in a polar coordinate system, including the temperature change direction and intensity to form vector data.

[0110] Step 122: Obtain the phase change latent heat value and the upper limit of the phase change temperature range of each annular sub-region from the phase change characteristic data, and calculate the thermal response rate coefficient of each annular sub-region. The thermal response rate coefficient is the ratio of the phase change latent heat value of the corresponding annular sub-region to the upper limit of the phase change temperature range.

[0111] The latent heat of phase change refers to the amount of latent heat absorbed or released during a phase change. The upper limit of the phase change temperature range refers to the highest critical temperature at which a material undergoes a phase change. The thermal response rate coefficient, the ratio of the latent heat of phase change to the upper limit of the temperature range, reflects the thermal response sensitivity.

[0112] In this embodiment, the latent heat of phase change and the upper limit of the phase change temperature range for each annular subregion are obtained from the phase change characteristics database. When calculating the thermal response rate coefficient, the latent heat of phase change is divided by the upper limit of the phase change temperature range. This coefficient represents the material's ability to absorb and release heat per unit temperature change and is used to quantify the subregion's potential contribution to condensation efficiency.

[0113] Step 123: Multiply the temperature gradient vector data of each annular sub-region by the thermal response rate coefficient of the corresponding annular sub-region to generate a condensation efficiency contribution value of each annular sub-region.

[0114] The condensation efficiency contribution value refers to the quantitative contribution of the sub-region to the overall condensation efficiency. The condensation efficiency contribution value refers to the quantitative contribution of each annular sub-region to the overall condensation efficiency based on its temperature gradient and thermal response characteristics.

[0115] In this embodiment of the present application, the temperature gradient vector data of each annular sub-region is scalar-multiplied by the corresponding thermal response rate coefficient to generate a condensation efficiency contribution value. This value comprehensively reflects the heat transfer rate driven by the temperature gradient and the thermal response characteristics of the phase change material.

[0116] Step 124 : Perform weighted summation on the condensation efficiency contribution value of each annular sub-region according to the area proportion of each annular sub-region in the annular region to generate the condensation efficiency.

[0117] The area ratio refers to the proportion of the sub-region area to the total area of the annular region.

[0118] In this embodiment, the condensation efficiency contribution values are weighted and summed based on the area proportion of each annular sub-region. Specifically, each sub-region's contribution value is multiplied by its area weight, and the results of all sub-regions are accumulated to finally generate the overall condensation efficiency value. This process ensures that large-sized sub-regions have a dominant influence on efficiency through area weighting.

[0119] The following is a specific example: In the condensing ring of a solar desalination plant, a 10-meter-diameter annular area is equally divided into 12 30-degree sub-areas. The temperature gradient vector of sub-area 3 is extracted [45°, 4.2°C / m]. The latent heat of phase change of 230 kJ / kg and the upper temperature limit of 75°C are obtained for this sub-area, and the thermal response rate coefficient is calculated as 3.07 kJ / (kg·°C). The condensing efficiency contribution value is generated as 4.2 × 3.07 = 12.89 kJ / (kg·m). Based on the 8.3% area proportion of sub-area 3, a weighted value of 12.89 × 0.083 = 1.07 is calculated. The final weighted sum of the 12 sub-areas is 0.85, which accurately reflects the actual condensing efficiency of the system.

[0120] By executing steps 121 to 124, the embodiment of the present application accurately quantifies the contribution of each local area to the condensation efficiency through refined segmentation of the annular sub-area and coupled calculation of the thermal response, and combines the area weight to achieve a scientific evaluation of the overall efficiency, thereby solving the problem of insufficient correlation between temperature gradient and phase change characteristics in traditional methods and improving the spatial resolution and accuracy of condensation efficiency measurement.

[0121] In one possible embodiment, step a2 inputs the flow characteristic vector into the flow constraint layer, which predefines the maximum pressure threshold and minimum flow rate threshold of the cooling water circulation system pipeline. The flow characteristic vector is truncated element by element using an inequality constraint function to generate the initial flow parameters of the cooling water circulation system corresponding to each partition, including:

[0122] Step a21: perform element-by-element constraint processing on each element of the flow characteristic vector. If the current element value is greater than the maximum allowable flow value corresponding to the maximum pressure threshold of the pipeline in the corresponding partition, the current element value is modified to the maximum allowable flow value.

[0123] The maximum pressure threshold for a pipeline is the pressure limit calculated based on the pipeline material, wall thickness, safety factor, etc. The maximum allowable flow rate is the upper limit of the flow rate converted based on the maximum pressure, and is related to the pipeline cross-sectional area and flow rate.

[0124] Step a22: If the current element value is less than the minimum allowable flow value corresponding to the minimum flow rate threshold of the corresponding partition, the current element value is modified to the minimum allowable flow value.

[0125] Among them, the minimum flow rate threshold refers to the minimum flow rate requirement to prevent siltation or freezing, and the minimum allowable flow value refers to the lower limit of the flow calculated by the minimum flow rate and the pipe cross-sectional area.

[0126] Step a23: If the current element value is between the minimum allowable flow rate and the maximum allowable flow rate, the current element value remains unchanged. The interval determination is performed using a conditional judgment algorithm to confirm that the current flow rate value is between the minimum allowable flow rate and the maximum allowable flow rate. This conditional judgment algorithm is known in the art and will not be further described here.

[0127] Step a24: Reorganize the elements after element-by-element constraint processing according to the partition mapping relationship to generate the initial flow parameters of the cooling water circulation system corresponding to each partition. The partition mapping relationship is the correspondence rule between the partition number and the physical / logical area.

[0128] Here's a specific example:

[0129] In the solar desalination plant, the characteristic flow value of +11 L / min in zone 3 was truncated to +10 L / min; the recommended value of -6% in zone 7 was increased to -5%; the reasonable value of +3.2 L / min in zone 5 was retained; and the zone codes [3, 5, 7] were reorganized into a vector [+10 L / min, +3.2 L / min, -5%] to generate initial flow parameters and send them to the circulation pump control system to ensure that zone 3 is not overpressurized, zone 7 is not blocked, and zone 5 is precisely regulated.

[0130] By executing steps a21 to a24, the embodiment of the present application uses partition-adaptive element-by-element constraint processing to convert the flow control recommendation value output by the neural network into a physically feasible initial parameter while ensuring pipeline safety and system stability. This not only avoids the risk of overpressure or insufficient flow, but also maximizes the optimization intention of the intelligent model, thereby achieving the unity of safety and control effect.

[0131] In one possible embodiment, step 133, using the initial flow parameters and the initial deflection angle parameters as state inputs of the reinforcement learning agent, receiving the parameter correction action output by the reinforcement learning agent, performing the parameter correction based on the physical constraints, and generating the intermediate flow parameters and the intermediate deflection angle parameters, includes:

[0132] Step b1: numerically compare the initial flow parameter with the upper and lower bound values of the flow parameter.

[0133] The upper constraint limit is the maximum allowable flow rate calculated based on parameters such as pipe material and design pressure. The lower constraint limit is the minimum allowable flow rate calculated based on the minimum flow rate threshold. Numerical comparison is the process of using an algorithm to compare parameters with preset thresholds and make corrections.

[0134] Step b2: numerically compare the initial deflection angle parameter with the upper constraint angle and the lower constraint angle of the deflection angle parameter.

[0135] The upper constraint angle refers to the maximum allowable deflection angle determined by the mechanical structure limit or fluid dynamics requirements, and the lower constraint angle refers to the minimum deflection angle required to maintain the basic function of the system.

[0136] Step b3: determining an intermediate flow parameter according to the flow comparison result, and determining an intermediate deflection angle parameter according to the deflection angle comparison result.

[0137] The flow comparison result refers to the set of flow parameters that have been corrected to ensure that the flow in each partition is within a safe range. The deflection angle comparison result refers to the set of deflection angle parameters that have been corrected to meet mechanical and fluid dynamics constraints.

[0138] The following is a specific example: During operation of a solar desalination plant, the initial flow rate in zone 7 was found to be -6% below the lower limit of -5% and marked as "needs upward adjustment." The deflector angle of 52° in zone 3 was verified to be within the range of 20° to 70° and marked as "compliant." The flow rate in zone 7 was corrected to -5%, while the parameters for zone 3 remained unchanged. The final output was the intermediate parameters: flow rate [+8.2 L / min, -5%] and deflector deflection angle 65°, which met pipeline safety requirements while maintaining control accuracy.

[0139] By executing steps b1 to b3, the embodiment of the present application uses a dual parameter boundary checking mechanism to ensure that all intermediate parameters strictly comply with the physical system constraints while retaining the intelligent control intention, providing safe and reliable input for subsequent optimization, and effectively avoiding equipment damage or efficiency loss due to parameter out-of-bounds.

[0140] In a possible embodiment, step b3, determining an intermediate flow parameter according to the flow comparison result, and determining an intermediate deflection angle parameter according to the deflection angle comparison result, includes:

[0141] Step c1: If the flow comparison result indicates that the initial flow parameter is greater than the upper limit value of the flow parameter constraint, the first fixed offset is used as the intermediate flow parameter; if the flow comparison result indicates that the initial flow parameter is less than the lower limit value of the flow parameter constraint, the second fixed offset is used as the intermediate flow parameter.

[0142] The first fixed offset is a safety downward adjustment when the flow rate exceeds the upper limit, used to prevent sudden pressure drops. The second fixed offset is a safety upward adjustment when the flow rate falls below the lower limit, used to prevent sudden changes in flow rate.

[0143] Step c2: If the deflection angle comparison result indicates that the initial deflection angle parameter is greater than the constrained upper limit angle of the deflection angle parameter, an intermediate deflection angle parameter is generated between the initial deflection angle parameter and the constrained upper limit angle of the deflection angle parameter through interpolation operation; if the initial deflection angle parameter is less than the constrained lower limit angle of the deflection angle parameter, an intermediate deflection angle parameter is generated between the constrained lower limit angle and the initial deflection angle parameter through interpolation operation.

[0144] Interpolation is a mathematical method for proportionally calculating transition values between boundary values and initial values. Interpolation coefficients are weighting factors that determine how close an intermediate parameter is to a boundary or initial value.

[0145] The following is a specific example: In a solar desalination plant, the first offset of -1.5L / min is applied to the over-limit flow of +11L / min in zone 3 (if the upper limit is +10L / min), and the intermediate value of 8.5L / min is output; for the out-of-bounds angle of 75° in zone 5 (the upper limit is 70°), the interpolation coefficient of 0.6 is used to generate 72°. The final intermediate parameters are flow [8.5L / min, -4.2%] and angle [72°, 18°], achieving a smooth transition from the out-of-bounds state to the safe range.

[0146] By executing steps c1 to c2, the embodiment of the present application generates a smoothly transitioned intermediate value when the parameter exceeds the limit through the combined application of the offset and interpolation algorithms, which not only satisfies the physical constraints but also retains the control continuity, effectively suppresses the system oscillation caused by parameter jumps, and improves the control stability and equipment life.

[0147] Figure 2A schematic diagram of the structure of an intelligent data monitoring and analysis system for a solar desalination system provided in an embodiment of the present application is shown in FIG. Figure 2 As shown, the system includes:

[0148] The acquisition module 21 is used to obtain the annular temperature gradient distribution data of the condenser tube wall, the phase change characteristic data of the steam in the condenser tube during the condensation process and the ambient temperature data during the condensation stage of the solar seawater desalination system. The corresponding condensation structures are embedded in different partitions inside the condenser tube wall, and each condensation structure includes a cooling water circulation system and an adjustable angle guide plate.

[0149] The calculation module 22 is used to calculate the condensation efficiency based on the annular temperature gradient distribution data and the phase change characteristic data.

[0150] The generation module 23 is used to input the condensation efficiency and ambient temperature data into a pre-trained reinforcement learning parameter optimization model, and generate the flow adjustment parameters and guide plate deflection angle adjustment parameters of each partition in combination with the physical constraints of the condensation structure.

[0151] The regulating module 24 is used to convert the flow regulating parameters and guide plate deflection angle regulating parameters of each partition into corresponding electric regulating valve opening instructions, so as to synchronously adjust the regulating valve opening and the angle of the adjustable angle guide plate of the cooling water circulation system corresponding to different partitions.

[0152] Figure 2 The intelligent data monitoring and analysis system of the solar desalination system can perform Figure 1 The implementation principles and technical effects of the intelligent data monitoring and analysis method for a solar desalination system described in the illustrated embodiment are not further elaborated. The specific manner in which each module and unit performs operations in the intelligent data monitoring and analysis system for a solar desalination system in the above embodiment has been described in detail in the relevant embodiments of the method and will not be further elaborated here.

[0153] In one possible design, Figure 2 The intelligent data monitoring and analysis system of a solar desalination system of the embodiment shown can be implemented as a computing device, such as Figure 3 As shown, the computing device may include a storage component 31 and a processing component 32 .

[0154] The storage component 31 stores one or more computer instructions, wherein the one or more computer instructions are called and executed by the processing component 32 .

[0155] The processing component 32 is used to obtain the annular temperature gradient distribution data of the condenser tube wall, the phase change characteristic data of the steam in the condenser tube during the condensation process, and the ambient temperature data during the condensation stage of the solar desalination system, wherein the corresponding condensation structure is embedded in different partitions inside the condenser tube wall, and each condensation structure includes a cooling water circulation system and an adjustable angle guide plate. Based on the annular temperature gradient distribution data and the phase change characteristic data, the condensation efficiency is calculated. The condensation efficiency and the ambient temperature data are input into a pre-trained reinforcement learning parameter optimization model, and the flow control parameters and the guide plate deflection angle adjustment parameters of each partition are generated in combination with the physical constraints of the condensation structure. The flow control parameters and the guide plate deflection angle adjustment parameters of each partition are converted into corresponding electric control valve opening instructions respectively, so as to synchronously adjust the control valve opening of the cooling water circulation system corresponding to different partitions and the angle of the adjustable angle guide plate.

[0156] The processing component 32 may include one or more processors to execute computer instructions to complete all or part of the steps in the above method. Of course, the processing component may also be implemented as one or more application-specific integrated circuits (ASICs), digital signal processors (DSPs), digital signal processing devices (DSPDs), programmable logic devices (PLDs), field programmable gate arrays (FPGAs), controllers, microcontrollers, microprocessors, or other electronic components to perform the above method.

[0157] The storage component 31 is configured to store various types of data to support operations on the terminal. The storage component can be implemented by any type of volatile or non-volatile memory device, or a combination thereof, such as random access memory (RAM), static random access memory (SRAM), electrically erasable programmable read-only memory (EEPROM), erasable programmable read-only memory (EPROM), programmable read-only memory (PROM), read-only memory (ROM), magnetic storage, flash memory, magnetic disk, or optical disk.

[0158] Of course, a computing device may also include other components, such as input / output interfaces, display components, communication components, etc.

[0159] The input / output interface provides an interface between the processing component and the peripheral interface module, which can be an output device, an input device, etc.

[0160] The communication component is configured to facilitate, among other things, wired or wireless communications between the computing device and other devices.

[0161] Among them, the computing device can be a physical device or an elastic computing host provided by a cloud computing platform, etc. In this case, the computing device can refer to a cloud server, and the above-mentioned processing components, storage components, etc. can be basic server resources rented or purchased from the cloud computing platform.

[0162] The present application also provides a computer storage medium storing a computer program, wherein the computer program can achieve the above-mentioned Figure 1 The embodiment shown is an intelligent data monitoring and analysis method for a solar seawater desalination system.

[0163] Those skilled in the art will clearly understand that, for the convenience and brevity of description, the specific working processes of the systems, devices and units described above can refer to the corresponding processes in the aforementioned method embodiments and will not be repeated here.

[0164] The device embodiments described above are merely illustrative. The units described as separate components may or may not be physically separate, and the components shown as units may or may not be physical units, i.e., they may be located in one location or distributed across multiple network units. Some or all of the modules may be selected based on actual needs to achieve the objectives of the present embodiment. Persons of ordinary skill in the art will be able to understand and implement the present invention without inventive effort.

[0165] Through the above description of the embodiments, those skilled in the art will clearly understand that each embodiment can be implemented using software plus a necessary general-purpose hardware platform, or of course, hardware. Based on this understanding, the essence of the above technical solution, or the portion that contributes to the prior art, can be embodied in the form of a software product. This computer software product can be stored in a computer-readable storage medium, such as ROM / RAM, a magnetic disk, or an optical disk, and includes a number of instructions for causing a computer device (such as a personal computer, server, or network device) to execute the methods described in each embodiment or certain portions of the embodiments.

[0166] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present application, rather than to limit them. Although the present application has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that they can still modify the technical solutions described in the aforementioned embodiments, or make equivalent replacements for some of the technical features therein. However, these modifications or replacements do not deviate the essence of the corresponding technical solutions from the spirit and scope of the technical solutions of the embodiments of the present application.

Claims

1. An intelligent data monitoring and analysis method for a solar desalination system, characterized in that: include: During the condensation phase of the solar desalination system, data on the annular temperature gradient distribution of the condenser tube wall, data on the phase change characteristics of the steam in the condenser tube during the condensation process, and ambient temperature data are obtained. The condenser tube wall is embedded with corresponding condensation structures in different partitions, and each condensation structure includes a cooling water circulation system and an adjustable angle guide plate. The ambient temperature data is the ambient temperature of the space in which the condenser tube is located when the solar desalination system is running. Calculating condensation efficiency based on the annular temperature gradient distribution data and the phase change characteristic data; Inputting the condensation efficiency and the ambient temperature data into a pre-trained reinforcement learning parameter optimization model, and combining the physical constraints of the condensation structure to generate flow regulation parameters and guide plate deflection angle regulation parameters for each partition; Convert the flow control parameters and deflection angle control parameters of each partition into corresponding electric control valve opening instructions, so as to synchronously adjust the opening of the control valve of the cooling water circulation system and the angle of the adjustable angle deflector corresponding to different partitions; The condensation efficiency and the ambient temperature data are input into a pre-trained reinforcement learning parameter optimization model, and the flow adjustment parameters and guide plate deflection angle adjustment parameters of each partition are generated in combination with the physical constraints of the condensation structure, including: The condensation efficiency and ambient temperature data are temporally and spatially bound according to the partition identification code, and input into the feature extraction module of the reinforcement learning parameter optimization model to extract the dynamic feature sequence and thermal inertia coefficient; The dynamic feature sequence is input into a dual-channel strategy network of a reinforcement learning parameter optimization model, wherein the first channel strategy network generates initial flow parameters of the cooling water circulation system corresponding to each partition through a fully connected layer and a flow constraint layer, and the second channel strategy network generates initial deflection angle parameters of the adjustable angle guide plate corresponding to each partition through a convolutional layer and an angle constraint layer. The flow constraint layer dynamically adjusts the constraint boundary based on the flow constraint strategy output by the reinforcement learning agent, and the angle constraint layer dynamically adjusts the constraint boundary based on the angle constraint strategy output by the reinforcement learning agent; Using the initial flow parameters and initial deflection angle parameters as state inputs of the reinforcement learning agent, receiving parameter correction actions output by the reinforcement learning agent, performing parameter correction based on the physical constraint conditions, and generating intermediate flow parameters and intermediate deflection angle parameters; Inputting the intermediate flow parameter and the intermediate deflection angle parameter into a multi-objective optimization module of a reinforcement learning parameter optimization model for parameter optimization; The optimized intermediate flow parameters, optimized intermediate deflection angle parameters, and historical optimal parameters are partitioned and cascaded and fused. The fusion weight is dynamically allocated according to the thermal inertia coefficient of the condensation structure of each partition to generate the flow regulation parameters and guide plate deflection angle regulation parameters of each partition.

2. The method according to claim 1, characterized in that The dynamic feature sequence is input into a dual-channel strategy network of a reinforcement learning parameter optimization model, wherein the first channel strategy network generates the initial flow parameters of the cooling water circulation system corresponding to each partition through a fully connected layer and a flow constraint layer, and the second channel strategy network generates the initial deflection angle parameters of the adjustable angle guide plate corresponding to each partition through a convolution layer and an angle constraint layer, including: Input each partition feature vector in the dynamic feature sequence into the fully connected layer and map it into a flow feature vector through dense connections between nodes; The flow characteristic vector is input into the flow constraint layer, in which the maximum pressure threshold and the minimum flow velocity threshold of the pipeline of the cooling water circulation system are predefined. The flow characteristic vector is truncated element by element through the inequality constraint function to generate the initial flow parameters of the cooling water circulation system corresponding to each partition; Inputting the two-dimensional spatial distribution features of each partition in the dynamic feature sequence into the convolution layer, performing sliding window convolution along the partition spatial dimension through multiple groups of convolution kernels, and extracting the deflection angle correlation features between adjacent partitions; The deflection angle association characteristics of the guide plate are input into the angle constraint layer, in which the upper and lower limits of the mechanical rotation angle of the adjustable angle guide plate are predefined. The deflection angle association characteristics of the guide plate are mapped to the preset angle range through a piecewise linear mapping function to generate the initial deflection angle parameters of the adjustable angle guide plate corresponding to each partition.

3. The method according to claim 1, characterized in that The calculating of the condensation efficiency based on the annular temperature gradient distribution data and the phase change characteristic data includes: Dividing the annular temperature gradient distribution data into a plurality of annular sub-regions at equal radial angles in the annular region, and extracting temperature gradient vector data of each annular sub-region; Obtaining the phase change latent heat value and the upper limit of the phase change temperature range of each annular sub-region from the phase change characteristic data, and calculating the thermal response rate coefficient of each annular sub-region, wherein the thermal response rate coefficient is the ratio of the phase change latent heat value of the corresponding annular sub-region to the upper limit of the phase change temperature range; Multiplying the temperature gradient vector data of each annular sub-region by the thermal response rate coefficient of the corresponding annular sub-region to generate a condensation efficiency contribution value of each annular sub-region; According to the area proportion of each annular sub-region in the annular region, a weighted sum is performed on the condensation efficiency contribution value of each annular sub-region to generate the condensation efficiency.

4. The method according to claim 2, characterized in that The flow characteristic vector is input into the flow constraint layer, in which the maximum pressure threshold and the minimum flow velocity threshold of the pipeline of the cooling water circulation system are predefined. The flow characteristic vector is truncated element by element through the inequality constraint function to generate the initial flow parameters of the cooling water circulation system corresponding to each partition, including: Performing element-by-element constraint processing on each element of the flow characteristic vector, if the current element value is greater than the maximum allowable flow value corresponding to the maximum pressure threshold of the pipeline of the corresponding partition, then modifying the current element value to the maximum allowable flow value; If the current element value is less than the minimum allowable flow value corresponding to the minimum flow rate threshold of the corresponding partition, the current element value is modified to the minimum allowable flow value; If the current element value is between the minimum allowable flow rate value and the maximum allowable flow rate value, the current element value is retained unchanged; The elements processed by element-by-element constraint are reorganized according to the partition mapping relationship to generate the initial flow parameters of the cooling water circulation system corresponding to each partition.

5. The method according to claim 1, wherein The method of using the initial flow parameter and the initial deflection angle parameter as the state input of the reinforcement learning agent, receiving the parameter correction action output by the reinforcement learning agent, performing the parameter correction based on the physical constraint condition, and generating the intermediate flow parameter and the intermediate deflection angle parameter includes: Comparing the initial flow parameter with the upper and lower bound values of the flow parameter; Numerically comparing the initial deflection angle parameter with the upper constraint angle and the lower constraint angle of the deflection angle parameter; The intermediate flow parameters are determined according to the flow comparison results, and the intermediate deflection angle parameters are determined according to the deflection angle comparison results.

6. The method according to claim 5, characterized in that The method of determining an intermediate flow parameter according to the flow comparison result and determining an intermediate deflection angle parameter according to the deflection angle comparison result includes: If the flow comparison result indicates that the initial flow parameter is greater than the upper constraint limit of the flow parameter, the first fixed offset is used as the intermediate flow parameter; if the flow comparison result indicates that the initial flow parameter is less than the lower constraint limit of the flow parameter, the second fixed offset is used as the intermediate flow parameter; If the deflection angle comparison result indicates that the initial deflection angle parameter is greater than the constrained upper limit angle of the deflection angle parameter, an intermediate deflection angle parameter is generated between the initial deflection angle parameter and the constrained upper limit angle of the deflection angle parameter through interpolation operation; if the initial deflection angle parameter is less than the constrained lower limit angle of the deflection angle parameter, an intermediate deflection angle parameter is generated between the constrained lower limit angle and the initial deflection angle parameter through interpolation operation.

7. An intelligent data monitoring and analysis system for a solar desalination system, characterized in that: The method for implementing the intelligent data monitoring and analysis of the solar desalination system according to claim 1 comprises: An acquisition module is used to obtain annular temperature gradient distribution data of the condenser tube wall, phase change characteristic data of the steam in the condenser tube during the condensation process, and ambient temperature data during the condensation stage of the solar desalination system, wherein the interior of the condenser tube wall is embedded with corresponding condensation structures in different partitions, each of which includes a cooling water circulation system and an adjustable angle guide plate, and the ambient temperature data is the ambient temperature of the space in which the condenser tube is located when the solar desalination system is running; a calculation module, configured to calculate a condensation efficiency based on the annular temperature gradient distribution data and the phase change characteristic data; a generation module, configured to input the condensation efficiency and the ambient temperature data into a pre-trained reinforcement learning parameter optimization model, and generate flow adjustment parameters and guide plate deflection angle adjustment parameters for each partition in combination with the physical constraints of the condensation structure; The regulating module is used to convert the flow regulation parameters and guide plate deflection angle regulation parameters of each partition into corresponding electric regulating valve opening instructions, so as to synchronously adjust the regulating valve opening of the cooling water circulation system corresponding to different partitions and the angle of the adjustable angle guide plate.

8. A computing device, characterized in that It comprises a processing component and a storage component; the storage component stores one or more computer instructions; the one or more computer instructions are used to be called and executed by the processing component to implement the intelligent data monitoring and analysis method of a solar desalination system as described in any one of claims 1 to 6.

9. A computer storage medium, characterized in that A computer program is stored, and when the computer program is executed by a computer, the intelligent data monitoring and analysis method for a solar seawater desalination system according to any one of claims 1 to 6 is implemented.

Images