An on-line monitoring method and system for molten salt energy storage

Abstract

This invention relates to the field of molten salt energy storage technology, and particularly to an online monitoring method and system for molten salt energy storage. The method uses a distributed temperature sensing network and infrared thermal imaging equipment to periodically collect thermal data from each zone of the energy storage unit, and combines this with a deep learning model to analyze the phase transition degree index. Based on real-time operating parameters and thermodynamic state characteristics, it simulates the spatiotemporal evolution of phase transition, generating dynamic range vector data of the phase transition front. Based on the correlation between heat conduction topology and molten salt components, it fuses and recombines multi-scale data to form a phase transition heat distribution map, identifying inefficient latent heat flux accumulation areas. It collects micro-thermal environment parameters and analyzes gradient characteristics to quantify the latent heat flux per unit time. When the flux is lower than the warning threshold, it automatically sends an efficiency anomaly warning signal to the control center, achieving precise monitoring and intelligent early warning.

Description

Technical Field

[0001] This invention relates to the field of molten salt energy storage technology, and in particular to an online monitoring method and system for molten salt energy storage. Background Technology

[0002] Molten salt energy storage technology, due to its high heat storage density, wide operating temperature range, and good thermal stability, has become a key supporting technology for large-scale thermal energy storage systems such as concentrated solar power (CSP), industrial waste heat recovery, and grid peak shaving. Its core principle is to utilize the absorption or release of a large amount of latent heat during the phase change process of molten salt, achieving efficient storage and on-demand release of thermal energy. However, in actual operation, the phase change process within molten salt energy storage units often exhibits non-uniform development due to factors such as uneven temperature field distribution, molten salt component segregation, flow dead zones, or abnormal local thermal resistance, resulting in delayed or incomplete phase change in some areas. This not only significantly reduces the overall latent heat throughput efficiency of the system but may also lead to safety hazards such as local overheating, solidification blockage, and even structural damage. Traditional molten salt energy storage monitoring methods mainly rely on thermocouple temperature measurements at limited points or periodic manual inspections, making it difficult to comprehensively and in real-time capture the complex three-dimensional temperature field evolution and phase change front dynamics within the energy storage unit.

[0003] The above content is only used to help understand the technical solution of the present invention and does not represent an admission that the above content is prior art. Summary of the Invention

[0004] The main objective of this invention is to provide an online monitoring method and system for molten salt energy storage, which aims to solve the technical problems of existing molten salt energy storage monitoring technologies, such as the inability to identify inefficient regions of latent heat throughput caused by uneven phase change in real time and quantify their dynamic performance, and the lack of online intelligent diagnosis and early warning capabilities for the overall heat exchange efficiency of the system.

[0005] To achieve the above objectives, the present invention provides an online monitoring method for molten salt energy storage, the method comprising: Using an online monitoring system equipped with a distributed temperature sensing network and infrared thermal imaging equipment, periodic thermal data are collected from each functional zone of the molten salt energy storage unit in operation to obtain the molten salt temperature field distribution data of each functional zone. The phase transition degree index of each functional zone is analyzed based on a deep learning model. Based on the real-time operating parameter sequence of the molten salt energy storage unit and the phase change degree index of each functional zone, the thermodynamic state characteristics of each functional zone are determined. The spatiotemporal evolution behavior of the phase change process in the molten salt energy storage unit is numerically simulated using the thermodynamic state characteristics to obtain the dynamic range vector data of the phase change front of each functional zone. Based on the heat conduction topology between each functional zone and the correlation of molten salt components, multi-scale fusion and recombination of all phase change front dynamic range vector data are performed to generate a phase change heat distribution map of the molten salt energy storage unit. Based on the phase change heat distribution map, multiple latent heat throughput inefficient accumulation areas in the energy storage unit are identified. Collect microthermal environment parameters of each latent heat inefficient accumulation zone, and determine the latent heat flux of the molten salt phase change process in each latent heat inefficient accumulation zone per unit time based on the gradient change characteristics of each microthermal environment parameter. When the latent heat throughput is lower than the preset performance warning threshold, an abnormal latent heat throughput performance warning signal is sent to the energy storage system integrated management and control center.

[0006] Optionally, the phase transition degree index for analyzing each functional partition based on a deep learning model specifically includes: Acquire thermal remote sensing images of each functional zone and their corresponding spatiotemporal coordinate information; A pre-trained convolutional neural network model was used to extract the thermophysical property parameters and phase transition-specific thermal spectrum features of molten salt in thermal images. Based on the molten salt thermophysical property characteristics and phase transition-specific thermal spectrum characteristics, the standardized phase transition degree index of each functional zone is calculated.

[0007] Optionally, the determination of the thermodynamic state characteristics of each functional zone based on the real-time operating parameter sequence of the molten salt energy storage unit and the phase transition degree index of each functional zone includes: Acquire spatial location information of each functional zone and real-time operation monitoring data of the molten salt energy storage unit; The spatial heterogeneity index of the phase transition process is calculated based on the phase transition degree index of each functional zone; By using an ensemble learning algorithm, combined with the real-time operation monitoring data and the spatial heterogeneity index value of the phase change process, the thermodynamic state characteristics of each functional zone are determined.

[0008] Optionally, the step of numerically simulating the spatiotemporal evolution of the phase transition process in the molten salt energy storage unit based on the thermodynamic state characteristics to obtain dynamic range vector data of the phase transition front for each functional zone includes: Select any functional zone as the target functional zone, and determine the molten salt phase transition stage parameters and thermal diffusion direction vector set of the target functional zone; The phase transition front advance rate of the target functional zone is calculated based on the thermodynamic state characteristic parameters corresponding to the target functional zone and the molten salt phase transition stage parameters. Based on the vector set of phase change front propagation rate and thermal diffusion direction, the dynamic range vector data of the phase change front of the target functional area is generated; the dynamic range vector data of the phase change front of the remaining functional areas is calculated in sequence.

[0009] Optionally, the step of multi-scale fusion and reorganization of all phase change front dynamic range vector data based on the heat conduction topology between functional zones and the correlation of molten salt components to generate a phase change heat distribution map of the molten salt energy storage unit includes: The thermal resistance distance and molten salt composition similarity coefficient between each functional zone were determined. A molten salt phase change heat diffusion network topology is constructed based on the thermal resistance distance between each functional zone and the molten salt composition similarity coefficient. Select any functional partition as the target functional partition; Based on the topology of the molten salt phase change heat diffusion network and the real-time operating conditions of the target functional partition, the dynamic range vector data of the phase change front of the target functional partition is spatially fused to obtain the recombined phase change front dynamic range of the target functional partition. The phase change front dynamic range spatial fusion of the remaining functional zones is completed sequentially; the phase change heat distribution map of the molten salt energy storage unit is generated by integrating the recombined phase change front dynamic range data of all functional zones.

[0010] Optionally, identifying multiple inefficient latent heat throughput accumulation zones in the energy storage unit based on the phase change heat distribution map specifically includes: Determine the set of regions in the phase change heat distribution map whose phase change front advance rate is lower than a second preset threshold. Based on the aforementioned region set and historical phase change performance data, multiple inefficient latent heat throughput clusters in the energy storage unit are identified using a spatial clustering algorithm.

[0011] Optionally, determining the latent heat flux of the molten salt phase change process per unit time within each inefficient latent heat flux accumulation zone based on the gradient change characteristics of each microthermal environment parameter includes: Calculate the spatial gradient distribution characteristics of each micro-thermal environment parameter; Select any inefficient latent heat throughput accumulation region as the target accumulation region; Calculate the baseline latent heat throughput flux of the target aggregation region based on the gradient distribution characteristics; Based on the baseline latent heat throughput and the microthermal environment parameters of the target aggregation area, the latent heat throughput of the molten salt phase change process in the target aggregation area per unit time is calculated by a thermodynamic dynamic model. Calculate the latent heat flux of the remaining inefficient latent heat flux accumulation zones in turn.

[0012] Furthermore, to achieve the above objectives, the present invention also provides an online monitoring system for molten salt energy storage, the system comprising: The multi-source sensing module is used to periodically collect thermal data of each functional zone of the molten salt energy storage unit in operation by utilizing an online monitoring system equipped with a distributed temperature sensing network and infrared thermal imaging equipment, to obtain the molten salt temperature field distribution data of each functional zone, and to analyze the phase transition degree index of each functional zone based on a deep learning model. The phase change modeling module is used to determine the thermodynamic state characteristics of each functional zone based on the real-time operating parameter sequence of the molten salt energy storage unit and the phase change degree index of each functional zone. The module then uses these thermodynamic state characteristics to perform numerical simulation of the spatiotemporal evolution behavior of the phase change process in the molten salt energy storage unit, and obtains the dynamic range vector data of the phase change front of each functional zone. The map fusion module is used to perform multi-scale fusion and recombination of all phase change front dynamic range vector data based on the heat conduction topology relationship between each functional zone and the correlation degree of molten salt components, to generate a phase change heat distribution map of the molten salt energy storage unit, and to identify multiple latent heat throughput inefficient accumulation areas in the energy storage unit based on the phase change heat distribution map. The flux quantization module is used to collect the micro-thermal environment parameters of each latent heat inefficient accumulation zone and determine the latent heat influent flux of the molten salt phase change process in each latent heat inefficient accumulation zone per unit time based on the gradient change characteristics of each micro-thermal environment parameter. The intelligent early warning module is used to send an early warning signal of abnormal latent heat throughput efficiency to the energy storage system integrated management and control center when the latent heat throughput is lower than the preset efficiency early warning threshold.

[0013] In addition, to achieve the above objectives, the present invention also provides an online monitoring device for molten salt energy storage, the device comprising: a memory, a processor, and an online monitoring program for molten salt energy storage stored in the memory and executable on the processor, the online monitoring program for molten salt energy storage being configured to implement the steps of the online monitoring method for molten salt energy storage as described above.

[0014] In addition, to achieve the above objectives, the present invention also provides a medium storing a molten salt energy storage online monitoring program, which, when executed by a processor, implements the steps of the molten salt energy storage online monitoring method as described above.

[0015] This invention provides an online monitoring method for molten salt energy storage. This method achieves high-precision, full-area, real-time dynamic monitoring of the phase change process within the molten salt energy storage unit by integrating multimodal sensing of distributed temperature sensing and infrared thermal imaging, intelligent analysis of phase change state driven by deep learning, and dynamic tracking of the phase change front based on thermodynamic modeling. Furthermore, by combining thermal conduction topology and component correlation for multi-scale data fusion, it accurately identifies inefficient latent heat transfer accumulation areas and quantifies their effective heat exchange flux per unit time, thus proactively issuing early warnings before a significant decline in system efficiency. This method significantly improves the operational safety, energy utilization efficiency, and intelligent operation and maintenance level of molten salt energy storage systems, effectively avoiding risks such as performance degradation and solidification blockage caused by local phase change hysteresis or abnormal thermal resistance, providing key technical support for the reliable and efficient operation of large-scale phase change thermal storage devices. Attached Figure Description

[0016] Figure 1 This is a flowchart illustrating an embodiment of the online monitoring method for molten salt energy storage of the present invention.

[0017] The realization of the objective, functional features and advantages of the present invention will be further explained in conjunction with the embodiments and with reference to the accompanying drawings. Detailed Implementation

[0018] It should be understood that the specific embodiments described herein are merely illustrative of the invention and are not intended to limit the invention.

[0019] Reference Figure 1 , Figure 1 This is a flowchart illustrating an embodiment of the online monitoring method for molten salt energy storage of the present invention, which presents an embodiment of the online monitoring method for molten salt energy storage of the present invention.

[0020] In one embodiment, the online monitoring method for molten salt energy storage includes: Step S100: Using an online monitoring system equipped with a distributed temperature sensing network and infrared thermal imaging equipment, periodic thermal data is collected from each functional zone of the molten salt energy storage unit in operation to obtain the molten salt temperature field distribution data of each functional zone, and the phase transition degree index of each functional zone is analyzed based on a deep learning model.

[0021] The distributed temperature sensing network can be a continuous or quasi-continuous temperature sensing system deployed along the internal space of the molten salt energy storage unit, capable of simultaneously acquiring temperature data from multiple points. In this embodiment, the distributed temperature sensing network can achieve spatially resolved temperature measurement through a single optical fiber based on physical effects such as Raman scattering, Brillouin scattering, or fiber gratings. Furthermore, the distributed temperature sensing network can be used to provide high spatial resolution internal three-dimensional temperature field lattice data, supporting fine characterization of phase transition regions. Exemplarily, the distributed temperature sensing network can include, but is not limited to, one or more of the following: a Raman scattering-based DTS system, a Brillouin scattering-based BOTDA system, and a quasi-distributed temperature measurement system based on a fiber grating array. The infrared thermal imaging device can be a non-contact thermal radiation detection device used to acquire a two-dimensional temperature distribution image of an object's surface. In an exemplary embodiment, the infrared thermal imaging device can receive thermal radiation emitted from the target surface through an infrared detector and convert it into a temperature distribution image. Furthermore, the infrared thermal imaging device can be used to provide the global surface temperature field of the molten salt energy storage unit's outer shell or visible area, supplementing internal sensing blind spots. For example, infrared thermal imaging equipment can employ mid-wave infrared thermal imagers, long-wave infrared thermal imagers, cooled infrared focal plane arrays, etc. The online monitoring system can be a hardware and software platform integrating multi-source sensing devices and data processing units. In one specific embodiment, the online monitoring system can be used to undertake periodic thermal data acquisition tasks, achieving simultaneous acquisition and preprocessing of multi-modal data. Furthermore, the online monitoring system can be equipped with a distributed temperature sensing network and infrared thermal imaging equipment, providing them with power supply, communication, and time synchronization support.

[0022] A molten salt energy storage unit can be a physical container and its internal medium that contain molten salt and store and release thermal energy. In this embodiment, the molten salt energy storage unit can serve as the main structure for the phase change thermal storage process, and its internal state is the monitoring object. Exemplarily, the molten salt energy storage unit can be one or more of the following: a single-tank molten salt thermal storage unit, a dual-tank molten salt thermal storage unit, and a packed-bed molten salt thermal storage unit. Each functional zone can be a logical or physical sub-region divided into the molten salt energy storage unit based on its structure, flow channels, or thermal characteristics. In an exemplary embodiment, each functional zone can be used to facilitate local state modeling and abnormal region location, improving monitoring granularity. Exemplarily, each functional zone can include an inlet buffer zone, a main thermal storage zone, and an outlet transition zone. Molten salt temperature field distribution data can be a set describing the temperature values ​​of molten salt at various locations in space, including both temporal and spatial dimensions. In this embodiment, the molten salt temperature field distribution data can be generated by fusion after joint acquisition by a distributed temperature sensing network and an infrared thermal imaging device. Furthermore, the molten salt temperature field distribution data can serve as the basic input data for phase change state analysis and thermodynamic modeling.

[0023] Deep learning models can be nonlinear mapping functions built on neural network architectures, used to extract latent features from high-dimensional inputs. In one specific embodiment, the deep learning model can be trained under supervision using historical operating data and labeled phase transition states to establish a mapping relationship between the temperature field and the degree of phase transition. Furthermore, the deep learning model can be used to analyze the phase transition degree index of each functional zone, overcoming the misjudgment of complex phase transition processes by traditional thresholding methods. For example, deep learning models can employ convolutional neural networks (CNNs), graph neural networks (GNNs), and spatiotemporal attention networks (ST-Transformers). The phase transition degree index can be a dimensionless indicator characterizing the proportion of molten salt in a certain region that has completed phase transition, typically ranging from 0 to 1. In this embodiment, the phase transition degree index can be used to quantify the local phase transition state as an initial condition or boundary constraint for thermodynamic modeling.

[0024] An online monitoring system equipped with a distributed temperature sensing network and infrared thermal imaging equipment is used to periodically collect thermal data from each functional zone of the operating molten salt energy storage unit. This can be achieved by synchronously triggering distributed temperature sensing and infrared imaging at set time intervals to acquire internal and surface temperature data. Furthermore, this operation can be achieved by using a time synchronization triggering mechanism to ensure the timestamps of the two types of sensor data are aligned, or by establishing a geometric correspondence between infrared image pixels and internal sensing points through a spatial coordinate mapping algorithm. This allows for comprehensive, multi-dimensional thermal state coverage acquisition of the entire molten salt energy storage unit. Acquiring the molten salt temperature field distribution data for each functional zone can be achieved by spatially aggregating and interpolating the original sensor data according to functional zones. Further, this operation can be achieved by using Kriging interpolation to fill sparse sensing areas, or by discretizing and reconstructing the temperature field based on a finite element mesh, thus forming a structured, zoned three-dimensional temperature field input.

[0025] Analyzing the phase transition degree index of each functional zone using a deep learning model can be achieved by inputting molten salt temperature field distribution data into a trained deep learning model and outputting the phase transition degree index. Furthermore, this operation can be implemented by using graph neural networks to process irregular zone topologies, or by using spatiotemporal convolutional networks to simultaneously model the temporal evolution and spatial distribution of the temperature field, thereby enabling intelligent, nonlinear identification of complex phase transition states.

[0026] Step S200: Based on the real-time operating parameter sequence of the molten salt energy storage unit and the phase change degree index of each functional zone, determine the thermodynamic state characteristics of each functional zone, and perform numerical simulation of the spatiotemporal evolution behavior of the phase change process in the molten salt energy storage unit through the thermodynamic state characteristics to obtain the dynamic range vector data of the phase change front of each functional zone.

[0027] The real-time operating parameter sequence can be time-series data reflecting the current operating conditions of the molten salt energy storage unit, such as flow rate, inlet and outlet temperatures, and pressure. In this embodiment, the real-time operating parameter sequence can be used to provide external boundary conditions for correcting the calculation of thermodynamic state characteristics. Thermodynamic state characteristics can be a comprehensive vector describing the thermodynamic properties of the molten salt at a specific moment, including enthalpy, entropy, and phase fraction. In an exemplary embodiment, the thermodynamic state characteristics can be used as the initial field for numerical simulation, driving the dynamic evolution calculation of the phase transition front. The spatiotemporal evolution behavior of the phase transition process can be a dynamic pattern of the molten salt phase transition interface changing with time and space. In this embodiment, the spatiotemporal evolution behavior of the phase transition process can be used to reveal the development path of non-uniform phase transitions, providing a basis for anomaly identification.

[0028] The dynamic range vector data of the phase change front can be a set of vectors describing the direction, velocity, and expansion range of the phase change interface in each functional zone. In a specific embodiment, the dynamic range vector data of the phase change front can be used to quantitatively characterize the phase change inhomogeneity, supporting subsequent multi-scale fusion analysis. The thermodynamic state characteristics of each functional zone are determined based on the real-time operating parameter sequence of the molten salt energy storage unit and the phase change degree index. This can be achieved by combining operating parameters (such as flow rate and pressure) and the phase change degree index, and calculating thermodynamic properties through a property database or equation of state. Furthermore, this operation can be achieved by calling property libraries such as property calculation tools to calculate the state of the molten salt mixture, or by performing rapid online estimation based on a simplified thermodynamic model, thereby constructing a complete thermodynamic initial field that conforms to the current operating conditions.

[0029] Numerical simulation of the spatiotemporal evolution of phase transition processes in molten salt energy storage units, based on thermodynamic state characteristics, can be achieved by solving energy conservation equations including latent heat terms of phase transition and simulating the phase transition interface propagation process. Furthermore, this operation can be implemented using CFD simulations with the enthalpy-porosity method or by capturing complex interface morphologies using the phase-field method, thus obtaining physically consistent phase transition dynamic predictions. Obtaining the dynamic range vector data of the phase transition front for each functional zone can be achieved by extracting the position, normal velocity, and expansion direction of the phase transition interface from the numerical simulation results and constructing them into vector form. Further, this operation can be achieved by tracing the interface and calculating the normal velocity using the level set method, or by locating the interface based on the maximum value of the phase fraction gradient and fitting the motion vector, thereby transforming the phase transition dynamics into quantifiable and fusionable structured data.

[0030] Step S300: Based on the heat conduction topology between each functional zone and the correlation of molten salt components, multi-scale fusion and recombination of all phase change front dynamic range vector data are performed to generate a phase change heat distribution map of the molten salt energy storage unit. Based on the phase change heat distribution map, multiple inefficient latent heat throughput accumulation areas in the energy storage unit are identified.

[0031] The thermal conduction topology can be the connection structure and intensity distribution of thermal conduction paths between functional zones within a molten salt energy storage unit. In this embodiment, the thermal conduction topology can be used as a physical constraint for multi-scale fusion, reflecting the influence of structural thermal coupling on phase change propagation. Molten salt component correlation can be a measure of the similarity or diffusion correlation of molten salt chemical compositions in different functional zones. In an exemplary embodiment, molten salt component correlation can be used to reflect the influence of component segregation on local phase change temperature and latent heat release efficiency, serving as a fusion weighting factor. Multi-scale fusion and recombination can be a process of integrating dynamic range vector data of the phase change front at different spatial granularities. In this embodiment, multi-scale fusion and recombination can be used to generate a phase change heat distribution map that balances local details with global consistency. The phase change heat distribution map can be a spatial distribution map that visually expresses the phase change activity and heat flow efficiency within a molten salt energy storage unit. In a specific embodiment, the phase change heat distribution map can be used to intuitively present the spatial heterogeneity of latent heat throughput capacity, supporting the identification of inefficient regions. Inefficient latent heat throughput clusters can be continuous spatial sets in the phase change heat distribution map where the latent heat exchange rate is significantly lower than the surrounding area. In this embodiment, the inefficient accumulation area of ​​latent heat throughput can be used as a target area for performance early warning, indicating the location of potential risks.

[0032] Based on the thermal conduction topology and molten salt component correlation between functional zones, multi-scale fusion and recombination of all phase change front dynamic range vector data is performed. This can be achieved by using thermal conduction paths as graph edge weights and component correlation as node similarity, followed by graph signal smoothing or cluster fusion. Furthermore, this operation can be implemented by constructing a weighted graph Laplacian matrix for multi-scale graph signal filtering, or by using hierarchical clustering algorithms to merge similar dynamic regions at different scales. This can eliminate isolated outliers, enhance physical consistency, and improve the accuracy of inefficient region identification. Generating a phase change thermal distribution map of the molten salt energy storage unit can be achieved by mapping the fused phase change front dynamic data to a spatial grid and visualizing it in the form of color levels or contour lines. Further, this operation can be achieved by using thermal map rendering technology to generate a continuous color level map, or by generating a pseudo-color distribution map through equivalent heat flux density inversion, thus providing an intuitive and global view of phase change efficiency distribution. Based on the phase change thermal distribution map, multiple latent heat throughput inefficient clusters in the energy storage unit are identified. This can be achieved by applying image segmentation or clustering algorithms to extract connected regions with phase change activity below a threshold. Furthermore, this operation can be achieved by using OTSU adaptive threshold segmentation to identify inefficient areas, or by using the DBSCAN density clustering algorithm to discover spatial clustering anomalies, thereby locating potential risk areas and narrowing the scope of subsequent analysis.

[0033] Step S400: Collect micro-thermal environment parameters of each latent heat inefficient accumulation zone, and determine the latent heat flux of the molten salt phase change process in each latent heat inefficient accumulation zone per unit time based on the gradient change characteristics of each micro-thermal environment parameter.

[0034] The micro-thermal environment parameters can be finer-grained thermal parameters within the inefficient latent heat flux accumulation zone, such as local temperature differences and heat flux density gradients. In this embodiment, the micro-thermal environment parameters can be used to accurately quantify the heat exchange capacity of the region. The gradient change characteristics can be the rate of change pattern of the micro-thermal environment parameters in the spatial or temporal dimensions. In an exemplary embodiment, the gradient change characteristics can be used to reflect local thermal resistance anomalies or flow stagnation, supporting the calculation of latent heat flux. The latent heat flux can be the effective heat absorbed or released per unit volume of molten salt completing a phase change per unit time. In this embodiment, the latent heat flux can be used as a quantitative evaluation index of system performance to trigger early warning judgments.

[0035] Collecting micro-thermal environment parameters for each inefficient latent heat throughput accumulation zone can be done by retrieving higher-resolution local sensor data or refining parameters from model outputs for identified areas. Furthermore, this operation can be achieved by resampling a subset of the region from the original distributed sensor data, or by inverting microscale thermal parameters through numerical simulation with localized mesh refinement, thereby obtaining refined thermal characteristics to support accurate flux calculation. Determining the latent heat throughput flux per unit time of the molten salt phase change process within each inefficient latent heat throughput accumulation zone based on the gradient variation characteristics of each micro-thermal environment parameter can be done by calculating the effective heat exchange rate based on Fourier's law or energy balance equations, combined with temperature gradient and phase change rate. Further, this operation can be achieved by estimating the latent heat release rate per unit volume through local heat flux density integration, or by multiplying the time derivative of the phase change degree index by the unit latent heat value to obtain the instantaneous flux, thereby establishing comparable and predictable quantitative performance indicators.

[0036] Step S500: When the latent heat throughput is lower than the preset performance warning threshold, a latent heat throughput performance abnormality warning signal is sent to the energy storage system integrated management and control center.

[0037] The performance warning threshold can be a preset lower limit for latent heat throughput, used to determine whether the system has entered an inefficient operating state. In this embodiment, the performance warning threshold can be used as a criterion for triggering a warning, enabling proactive operation and maintenance intervention. The energy storage system integrated management and control center can be a central control platform responsible for the overall scheduling, monitoring, and decision-making of the molten salt energy storage system. In a specific embodiment, the energy storage system integrated management and control center can be used to receive warning signals and initiate corresponding operation and maintenance strategies. The latent heat throughput performance anomaly warning signal can be an alarm message generated when the latent heat throughput flux is lower than the performance warning threshold. In this embodiment, the latent heat throughput performance anomaly warning signal can be used to prompt operation and maintenance personnel or the automatic control system to take measures to prevent further performance deterioration.

[0038] When the latent heat throughput flux falls below a preset performance warning threshold, an abnormal latent heat throughput performance warning signal is sent to the integrated management and control center of the energy storage system. This can be achieved by comparing the calculated flux with the preset threshold; if it falls below, a warning is sent via the communication interface. Furthermore, this operation can be implemented by pushing structured warning messages to the control center via the OPCUA protocol, or by displaying an alarm on the local HMI interface and simultaneously uploading it to the cloud platform, thus achieving a closed loop from status awareness to proactive intervention.

[0039] Taking the operation monitoring of molten salt storage tanks in a concentrated solar power (CSP) plant as an example, the online monitoring method for molten salt energy storage in this embodiment can be implemented in a 50MW tower CSP plant's dual-tank system of hot and cold molten salt. A distributed fiber optic temperature measurement network based on Raman scattering is deployed throughout the height of the tanks, while a long-wave infrared thermal imager is installed on the outer wall of the tank. The online monitoring system collects data synchronously every 5 minutes. A deep learning model is trained based on historical charge-discharge cycle data, and outputs the phase change degree index at 1-meter height stratifications in real time. Combining pump flow rate and inlet / outlet temperature, the thermodynamic model simulates the dynamic vector of the phase change front advancing downwards (exothermic) or upwards (thermal storage). By integrating the thermal bridge path (heat conduction topology) of the tank's steel structure with the possible sodium and potassium segregation trends (component correlation) of the nitrate mixture, a phase change heat distribution map is generated, identifying inefficient accumulation zones formed by flow dead zones in the bottom corners. Gradient analysis of microthermal parameters in this area shows that the latent heat throughput is consistently below 60% of the design value. The system automatically issues an early warning to the power plant's DCS, prompting adjustments to the molten salt circulation strategy or the activation of local heating to prevent solidification and blockage.

[0040] In one embodiment, the process of analyzing the phase transition index of each functional partition using a deep learning model specifically includes: Acquire thermal remote sensing images of each functional zone and their corresponding spatiotemporal coordinate information.

[0041] The thermal remote sensing imagery can be a two-dimensional or three-dimensional thermal field image with spatial continuity and physical semantics, generated by fusing infrared thermal imaging and distributed temperature sensing data. It can provide physically consistent high-dimensional input to convolutional neural networks, replacing the original discrete point data and improving feature extraction efficiency and spatial context awareness. In this embodiment, the thermal remote sensing imagery can be fused by spatiotemporal alignment and interpolation of the surface temperature distribution acquired by the infrared thermal imager and the internal temperature field reconstructed by distributed fiber optic temperature measurement, forming a thermal image in a unified coordinate system. For example, the thermal remote sensing imagery can include, but is not limited to, one or more of the following: surface-internal fused thermal map, time-stacked thermal cube, and multispectral thermal response imagery. The spatiotemporal coordinate information can be the physical spatial location and acquisition timestamp corresponding to each pixel or voxel in the thermal remote sensing imagery, which can be used to ensure the traceability and alignment of thermal data in spatial structure and time series, supporting dynamic modeling and cross-modal fusion. Furthermore, the spatiotemporal coordinate information can include, but is not limited to, a three-dimensional spatial coordinate system, sampling timestamp sequences, and sensor topology mapping tables.

[0042] Acquiring thermal remote sensing images of each functional zone and their corresponding spatiotemporal coordinates can be achieved by cropping infrared thermal imaging and distributed temperature sensing data according to functional zones and attaching spatial location and time labels. In an exemplary embodiment, this operation can be achieved by mapping infrared pixels to a three-dimensional grid of energy storage units through a coordinate transformation matrix, or by automatically generating a partition mask using the sensor deployment topology and extracting the corresponding thermal sub-map. This allows for the construction of a model input with physical semantics and geometric structure, improving the spatial consistency of subsequent feature extraction.

[0043] The thermal properties of molten salt and phase transition-specific thermal spectra in thermal images were extracted using a pre-trained convolutional neural network model.

[0044] The pre-trained convolutional neural network model can be a deep convolutional architecture trained on a large number of historical thermal remote sensing images and corresponding phase change labels. It is used to automatically extract discriminative features from the thermal field and can be used to simultaneously extract molten salt thermophysical parameters and phase change-specific thermal spectrum features from thermal remote sensing images, achieving end-to-end phase change state mapping. In a specific embodiment, the pre-trained convolutional neural network model can employ a transfer learning strategy, pre-training on a general thermal imaging dataset and then fine-tuning using labeled data from molten salt thermal storage scenarios. For example, the pre-trained convolutional neural network model can use a ResNet-based thermal field encoder, a U-Net structured thermal spectrum segmentation network, etc. Specifically, this convolutional neural network can adopt a multi-layer encoder-decoder structure, containing 12 hidden layers. The input is multispectral thermal imaging data (size 256×256×6 channels, including parameters such as temperature field, heat flux density, and radiation intensity), and the output is a scalar value of the phase change degree index. The network integrates local thermal features with global spatial correlations through residual connections and attention mechanisms. It is pre-trained on a million-level molten salt thermal simulation dataset using transfer learning and then fine-tuned using experimental data. Its core evaluation logic is based on the nonlinear mapping relationship between thermal spectral features and latent heat of phase transition. It extracts key features such as abrupt temperature gradient changes and thermal hysteresis loop morphology at the molten salt solid-liquid interface through convolutional kernels. Finally, it regresses the standardized phase transition exponent through fully connected layers, with the error controlled within ±0.05.

[0045] The thermophysical properties of molten salt can be a deep characterization of the thermophysical properties (such as specific heat capacity, thermal conductivity, and density) implicitly reflected by the spatiotemporal evolution model of the temperature field. These properties can serve as the basis for calculating the degree of phase change, enabling the model to adapt to differences in molten salt behavior under different compositions or aging conditions. Furthermore, the thermophysical properties of molten salt can include, but are not limited to, thermal diffusivity response characteristics, thermal inertia index, and local thermal capacity gradient patterns. Phase change-specific thermal spectral characteristics can be unique thermal response patterns exhibited by molten salt within the critical phase change range, such as temperature plateaus, sudden changes in heat flux, and phase change hysteresis loops. These can be used to directly indicate whether a phase change has occurred and its extent, improving the sensitivity of identifying non-equilibrium phase change processes. In an exemplary embodiment, phase change-specific thermal spectral characteristics can include, but are not limited to, the persistence of the phase change plateau temperature range, sudden changes in latent heat release rate, and cooling-heating thermal spectral asymmetry.

[0046] Extracting molten salt thermophysical parameters and phase transition-specific thermal spectral features from thermal images using a pre-trained convolutional neural network model can be achieved by inputting thermal remote sensing images into a convolutional neural network and decoupling the property-related and phase transition-related feature maps through the activation responses of intermediate layers. Furthermore, this operation can be achieved by employing a multi-branch network structure to extract property features and phase transition features separately, or by using an attention mechanism to weight and highlight the thermal spectral response of the phase transition critical region. This allows for the automatic deconstruction of complex thermal response patterns, avoiding the subjectivity and limitations of manually designed features.

[0047] Based on the characteristics of molten salt thermophysical parameters and phase transition-specific thermal spectrum characteristics, the standardized phase transition degree index of each functional zone is calculated.

[0048] The standardized phase transition degree index can be a phase transition completion index that has been normalized and has uniform dimensions and comparability. Its value range is typically [0, 1]. It can be used to eliminate index deviations caused by differences in size, location, or physical properties between different functional zones, supporting cross-regional performance comparison and threshold warning. For example, the standardized phase transition degree index can include, but is not limited to, phase transition indices based on enthalpy normalization, standardized indices based on phase fraction mapping, and relative completion based on heat flux integrals. Based on the characteristics of molten salt thermophysical parameters and phase transition-specific thermal spectra, the standardized phase transition degree index of each functional zone can be calculated by concatenating two types of feature vectors and inputting them into a fully connected regression head, outputting a phase transition degree value constrained by a normalization function. In a specific embodiment, this operation can be achieved by compressing the original index to the [0, 1] interval using Min-Max normalization, or by normalizing the physical dimensions based on the theoretical maximum latent heat throughput, thereby generating physically interpretable and cross-regionally comparable quantitative indicators, improving the accuracy of the initial conditions for subsequent thermodynamic modeling.

[0049] Taking the intelligent monitoring of molten salt storage tanks for industrial waste heat recovery as an example, the online monitoring method for molten salt energy storage in this embodiment can be implemented in a waste heat recovery system of a steel plant. An infrared thermal imager is deployed on the outer wall of the molten salt storage tank, with a Raman DTS optical fiber embedded inside. The online monitoring system generates a thermal remote sensing image covering the entire tank every 2 minutes, along with millimeter-level spatial coordinates and precise timestamps. A pre-trained ResNet-34 model receives this image and automatically identifies a typical phase change plateau temperature zone (lasting more than 15 minutes) in the central region, but with slow heat diffusion, indicating a low thermal conductivity in this area. The model also extracts the thermophysical parameters of this region, showing an abnormally high effective heat capacity, suggesting nitrate segregation. Combining these two types of features, the model outputs a standardized phase change index of 0.38 for this functional area, significantly lower than the surrounding areas (0.72–0.85). This index is fed into the subsequent thermodynamic simulation module, triggering a judgment of inefficient accumulation in this area and issuing an early warning after flux assessment, prompting an adjustment of the molten salt stirring frequency.

[0050] In one embodiment, determining the thermodynamic state characteristics of each functional zone based on the real-time operating parameter sequence of the molten salt energy storage unit and the phase transition degree index of each functional zone specifically includes: Acquire spatial location information of each functional zone and real-time operation monitoring data of the molten salt energy storage unit.

[0051] The spatial location information of each functional zone can describe the geometric coordinates or topological relationships of each functional zone within the molten salt energy storage unit in three-dimensional space. This information can be used to provide spatial context for phase change state modeling and support the analysis of heat conduction paths and heterogeneity between regions. In an exemplary embodiment, the spatial location information of each functional zone can be extracted from the system CAD model or sensor deployment documents. For example, the spatial location information of each functional zone can include, but is not limited to, one or more of the following: voxel center coordinates in Cartesian coordinates, radial-axial hierarchical index based on the tank structure, and node topology numbers in the flow channel network.

[0052] Real-time operational monitoring data of the molten salt energy storage unit can be a set of externally measurable parameters reflecting the current operating conditions of the system, including flow rate, pressure, inlet and outlet temperatures, heating power, etc. These can be used as boundary conditions and driving variables for thermodynamic state inference, reflecting the system-level energy input and output status. Furthermore, the real-time operational monitoring data of the molten salt energy storage unit can be read from a DCS or PLC system. In a specific embodiment, the real-time operational monitoring data of the molten salt energy storage unit may include, but is not limited to, molten salt circulating pump flow signals, electric heater power feedback, inlet and outlet temperature differences of the storage tank, and system back pressure readings.

[0053] Acquiring spatial location information of each functional zone and real-time operational monitoring data of the molten salt energy storage unit can be achieved by extracting spatial coordinates from the system CAD model or sensor deployment documents and reading real-time operating parameters from the DCS or PLC system. Furthermore, this operation can be accomplished by automatically exporting the geometric center coordinates of each sensing zone from the BIM model, or by subscribing to the operational parameter stream from the industrial control system in real time using the OPCUA protocol. This allows for the construction of a joint context of physical space and operating conditions, providing multi-dimensional input for state modeling.

[0054] The spatial heterogeneity index of the phase transition process is calculated based on the phase transition degree index of each functional zone.

[0055] The spatial heterogeneity index of the phase change process can be a dimensionless index that quantifies the differences in the degree of phase change between different functional zones, reflecting the degree of asynchronous or non-uniform phase change. It can be used to characterize the differences in local phase change development caused by temperature field inhomogeneity, component segregation, or flow dead zones, serving as a key feature for thermodynamic modeling. In a specific embodiment, the spatial heterogeneity index of the phase change process can be calculated based on the phase change degree index of each functional zone, using methods such as statistical dispersion, spatial gradient, or graph Laplacian operator. For example, the spatial heterogeneity index of the phase change process may include, but is not limited to, the standard deviation of the phase change degree, the gradient magnitude of the phase change index between adjacent zones, and a phase change consistency score based on a graph structure.

[0056] The spatial heterogeneity index of the phase transition process is calculated based on the phase transition degree index of each functional zone. This can be achieved by performing spatial statistics or graph structure analysis on the phase transition degree index of each functional zone to generate scalars or vectors representing non-uniformity. Furthermore, this operation can be achieved by calculating the standard deviation or coefficient of variation of the phase transition degree index of all zones, or by constructing a zone adjacency graph and using a graph Laplace quadratic form to measure the smoothness of the phase transition. This transforms the differences in local phase transition states into quantifiable feature variables that can be input into the model.

[0057] By using an ensemble learning algorithm, combined with real-time operational monitoring data and spatial heterogeneity index values ​​of phase change processes, the thermodynamic state characteristics of each functional zone are determined.

[0058] Ensemble learning algorithms can be machine learning methods that combine multiple base learners to improve prediction performance and robustness. They can be used to fuse high-dimensional heterogeneous inputs (operational parameters and spatial heterogeneity indices) to more accurately infer the thermodynamic state characteristics of each partition. In an exemplary embodiment, an ensemble learning algorithm can train multiple weak models (such as decision trees) and weightedly fuse their outputs, utilizing a bias-variance tradeoff to optimize generalization ability. Exemplary examples include, but are not limited to, ensemble decision tree models, gradient boosting tree models, and stacking ensemble models.

[0059] The thermodynamic state characteristics of each functional zone can be a comprehensive vector describing the current thermodynamic properties of a specific functional zone, such as effective specific heat capacity, local latent heat utilization rate, and phase change rate. These characteristics can be used as the initial field and boundary conditions for dynamic numerical simulation of the phase change front, affecting the accuracy of subsequent inefficient zone identification. In this embodiment, the thermodynamic state characteristics of each functional zone are generated by an ensemble learning algorithm based on real-time operational monitoring data and spatial heterogeneity index values ​​of the phase change process.

[0060] By employing ensemble learning algorithms and combining real-time operational monitoring data with spatial heterogeneity indices of phase transition processes, the thermodynamic state characteristics of each functional zone can be determined. This can be achieved by concatenating the temporal characteristics of operational parameters with spatial heterogeneity indices into an input vector, which is then input into a pre-trained ensemble model to output the thermodynamic state characteristics. Furthermore, this operation can be implemented by using a gradient boosting tree regression model to predict the effective enthalpy of each zone, or by using an ensemble decision tree model to output the local latent heat utilization rate and phase transition rate as two separate target variables. This allows for high-fidelity modeling of complex phase transition states under multi-factor coupling, improving the accuracy of subsequent simulations and early warning systems.

[0061] For example, in the scenario of state modeling of an industrial waste heat recovery molten salt thermal storage system, the online monitoring method for molten salt energy storage in this embodiment can be as follows: In a waste heat recovery system of a steel plant, the molten salt storage tank is divided into 12 axial functional zones. The system obtains the center coordinates (spatial location information) of each zone from the BIM model and reads the molten salt flow rate, flue gas inlet temperature, and electric heating power (real-time operation monitoring data) updated every second from the PLC. The deep learning model has output the phase change degree index of each zone (e.g., ranging from 0.3 to 0.9), based on which the spatial heterogeneity index value of the phase change process is calculated to be 0.28 (standard deviation). This value, together with the operating parameters, is input into the pre-trained gradient boosting tree model ensemble model, which outputs the effective specific heat capacity and local latent heat utilization rate of each zone. The results show that the latent heat utilization rate of the bottom 10-12 zones is significantly low. Combined with their spatial location near the outlet dead zone, it is judged that there is a risk of flow stagnation, providing a high-precision initial field for subsequent phase change front simulation.

[0062] In one embodiment, the spatiotemporal evolution behavior of the phase transition process in the molten salt energy storage unit is numerically simulated based on thermodynamic state characteristics to obtain the dynamic range vector data of the phase transition front of each functional zone, specifically including: Select any functional zone as the target functional zone, and determine the molten salt phase transition stage parameters and thermal diffusion direction vector set of the target functional zone; The target functional partition can be a single functional partition selected for dynamic calculation of the phase change front during numerical simulation. It can serve as the basic calculation unit for generating dynamic range vector data of the phase change front, enabling sequential processing of the partition. In this embodiment, the target functional partition can participate in the propagation rate calculation together with thermodynamic state characteristic parameters, molten salt phase change stage parameters, and thermal diffusion direction vector set. The molten salt phase change stage parameters can be discrete or continuous variables describing the current phase change process state of the target functional partition, reflecting the solid-liquid ratio or the existence of the phase interface. They can be used to provide a basis for property switching in the calculation of the phase change front propagation rate, distinguishing different thermal response mechanisms such as completely solid, partially melted, and completely liquid states. For example, the molten salt phase change stage parameters can include, but are not limited to, one or more of the following: solid phase ratio coefficient, phase interface existence marker, and phase change completion index. The thermal diffusion direction vector set can be a set of unit vectors characterizing the dominant propagation direction of local heat flow within the target functional partition. It is determined by the temperature gradient and thermal conductivity tensor and can be used to constrain the spatial orientation of the phase change front expansion, ensuring that the dynamic evolution conforms to the real heat conduction path. In one exemplary embodiment, the set of heat diffusion direction vectors may include, but is not limited to, the main heat flow direction vector, the anisotropic heat diffusion basis vector, and the multimodal heat conduction direction cluster.

[0063] Selecting any functional partition as the target functional partition can be achieved by choosing one partition from all functional partitions in a preset order or using a parallel strategy as the current computation object. Furthermore, selecting any functional partition as the target functional partition can be implemented through round-robin scheduling or load balancing strategies, thereby enabling partitioned and modular phase transition dynamic modeling, facilitating parallel computing and local optimization. Determining the molten salt phase transition stage parameters and thermal diffusion direction vector set of the target functional partition can be based on the phase transition degree index and temperature field gradient of the partition, determining its phase transition stage and calculating the main thermal diffusion direction. In a specific embodiment, determining the molten salt phase transition stage parameters and thermal diffusion direction vector set of the target functional partition can be achieved by mapping the phase transition degree index threshold range to discrete phase transition stage labels, or by extracting the main thermal diffusion direction through temperature field gradient tensor eigenvalue decomposition, thus providing dual constraints of physical property state and spatial direction for the propulsion rate calculation.

[0064] The phase transition front advance rate of the target functional zone is calculated based on the thermodynamic state characteristic parameters and molten salt phase transition stage parameters corresponding to the target functional zone. The thermodynamic state characteristic parameters can be specific physical quantities extracted from thermodynamic state characteristics for calculating the propulsion rate, such as local specific heat capacity, latent heat density, density, and thermal conductivity. These can be used as input parameters for the phase change kinetic equations to support the physical consistency calculation of the propulsion rate. Furthermore, the thermodynamic state characteristic parameters can be jointly determined with the molten salt phase change stage parameters to determine the numerical value of the phase change front propulsion rate. The phase change front propulsion rate can be the distance the phase change interface moves along the direction of thermal diffusion per unit time, with definite physical dimensions (e.g., mm / s). It can be used to quantify the dynamic intensity of the phase change and is a key intermediate variable connecting the thermodynamic state and geometric evolution. In an exemplary embodiment, the phase change front propulsion rate can be inferred from the principle of energy conservation by combining local heat flux density, latent heat of phase change, and interface area to estimate the interface movement speed.

[0065] Based on the thermodynamic state characteristic parameters and molten salt phase transition stage parameters corresponding to the target functional area, the advancement rate of the phase transition front of the target functional area can be calculated. This can be achieved by substituting the thermodynamic state characteristic parameters into the phase transition kinetic model and activating the corresponding physical property equations according to the phase transition stage parameters to solve for the interface advancement rate. For example, this operation can be achieved by using the analytical approximate solution of the Stefan problem to calculate the one-dimensional advancement rate, or by numerically solving the multi-dimensional advancement rate based on the local energy balance equation, thereby obtaining a physically consistent local phase transition rate and avoiding simulation deviations caused by the homogeneity assumption.

[0066] Based on the vector set of phase change front propagation rate and thermal diffusion direction, dynamic range vector data of the phase change front of the target functional area is generated; Based on the propulsion rate and thermal diffusion direction vector set of the phase transition front, dynamic range vector data of the phase transition front for the target functional zone is generated. This can be achieved by combining the scalar propulsion rate and direction vector set to form a dynamic range vector with direction and amplitude. Furthermore, this operation can be achieved by projecting the propulsion rate onto multiple thermal diffusion directions to generate a multi-directional extension vector, or by multiplying the main direction vector by the rate to generate a single dominant dynamic vector. This allows for the construction of a structured output containing spatiotemporal evolution information, supporting subsequent fusion and visualization.

[0067] The dynamic range vector data calculations for the phase transition fronts of the remaining functional zones are completed sequentially.

[0068] The remaining functional zones can be any other functional zones besides the current target functional zone. These can be used to achieve dynamic modeling of the phase transition front covering the entire unit by sequentially rotating them to the target functional zone. The calculation of the dynamic range vector data of the phase transition front for the remaining functional zones can be performed by iterating through all functional zones and repeating the aforementioned steps until all zones have been calculated. In a specific embodiment, this operation can be implemented using a parallel computing framework to process multiple zones simultaneously, or by prioritizing the processing order of zones according to their thermal importance. This allows for dynamic modeling of the phase transition front covering the entire region, ensuring the integrity of monitoring.

[0069] For example, in the scenario of monitoring non-uniform phase change in molten salt storage tanks for industrial waste heat recovery, the online monitoring method for molten salt energy storage in this embodiment can be as follows: In a waste heat recovery system of a steel plant, the upper region of the molten salt storage tank is in a partially molten state for a long time due to the design of the inlet flow channel. The online monitoring system identifies this region as functional zone A. In the numerical simulation stage, the system selects it as the target functional zone. Based on the phase change degree index of 0.45 output by the deep learning model, it is determined to be in the "partially molten" stage, with a solid phase ratio of 0.55 corresponding to the molten salt phase change stage parameter. At the same time, the three-dimensional temperature gradient field reconstructed by the distributed temperature sensing network shows that the heat flow is mainly conducted downwards, generating a set of heat diffusion direction vectors containing the main directions [0, 0, -1]. Combining the thermodynamic state characteristic parameters of this zone (such as the latent heat of nitrate mixture of 160 kJ / kg and density of 1900 kg / m³), the phase change front advance rate is calculated to be 0.8 mm / min. Finally, a dynamic range vector of the phase change front pointing downwards with a rate of 0.8 mm / min is generated. The system then processed the remaining partitions in sequence and found that a certain partition at the bottom had a propagation rate of 0.1 mm / min and a disordered direction. It was marked as an inefficient clustering area and an early warning was triggered.

[0070] In one embodiment, based on the heat conduction topology between functional zones and the correlation of molten salt components, multi-scale fusion and recombination are performed on all phase change front dynamic range vector data to generate a phase change heat distribution map of the molten salt energy storage unit. Specifically, this includes: The thermal resistance distance and molten salt composition similarity coefficient between each functional zone were determined.

[0071] The thermal resistance distance between functional zones can be a quantitative indicator characterizing the equivalent thermal conduction resistance between any two functional zones. It can be used as a physical weighting basis for constructing the topology of a molten salt phase change heat diffusion network, replacing traditional Euclidean distance and more accurately depicting heat transfer efficiency. In this embodiment, the thermal resistance distance between functional zones can be obtained by combining system design parameters, online property inversion, or offline calibration experiments. For example, the thermal resistance distance between functional zones can include, but is not limited to, one or more of the following: contact thermal resistance-dominated thermal resistance distance, thermally conductive material path thermal resistance distance, and equivalent thermal resistance distance of multilayer composite structures.

[0072] The molten salt component similarity coefficient can be a dimensionless index measuring the similarity of the chemical composition of molten salts in any two functional zones. It can be used as a constraint factor for property consistency during the fusion process, correcting dynamic deviations in phase transition caused by differences in local phase transition temperature or latent heat. Furthermore, the molten salt component similarity coefficient can be calculated by inferring local components and calculating the similarity coefficient using distributed conductivity or spectral sensing. In an exemplary embodiment, the molten salt component similarity coefficient may include, but is not limited to, similarity coefficients based on principal component concentration ratios, component similarity coefficients based on mixing entropy, and similarity coefficients based on ion mobility distribution.

[0073] Determining the thermal resistance distance and molten salt composition similarity coefficient between functional zones can be achieved by obtaining quantitative values ​​of thermal resistance and composition similarity between functional zone pairs through system design parameters, online property inversion, or offline calibration experiments. Furthermore, determining the thermal resistance distance and molten salt composition similarity coefficient between functional zones can be achieved by inverting the equivalent thermal resistance distance based on finite element thermal simulation, inferring local composition through distributed conductivity or spectral sensing, and calculating the similarity coefficient. This allows for the establishment of a physical-chemical dual-dimensional zone correlation metric, providing a foundation for constructing a realistic thermal diffusion topology.

[0074] The topology of the molten salt phase change heat diffusion network is constructed based on the thermal resistance distance between each functional zone and the similarity coefficient of the molten salt composition.

[0075] The molten salt phase change thermal diffusion network topology can be a graph model consisting of functional zones as nodes and edges weighted by thermal resistance distance and molten salt component similarity coefficient. This model can provide physical-chemical coupling adjacency relationships and weighting criteria for the spatial fusion of dynamic range vector data at the phase change front. In a specific embodiment, the molten salt phase change thermal diffusion network topology can be quantified using thermal resistance distance and molten salt component similarity coefficient, guiding the subsequent adaptive fusion process of phase change data from each functional zone.

[0076] Constructing a molten salt phase change heat diffusion network topology based on the thermal resistance distance between functional zones and the molten salt component similarity coefficient can be achieved by treating the functional zones as graph nodes and using the product or weighted sum of the thermal resistance distance and the component similarity coefficient as edge weights to build a weighted undirected graph. Furthermore, this topology can be implemented by mapping the thermal resistance distance to connection strength using a Gaussian kernel function and employing a product fusion strategy (edge ​​weights equal to the component similarity coefficient multiplied by the exponential decay function of the thermal resistance distance normalized to the characteristic length). This allows for the formation of a non-uniform adjacency relationship model that integrates heat transfer efficiency and physical property consistency.

[0077] Select any functional partition as the target functional partition.

[0078] The target functional partition can be a specific functional partition currently being processed during the multi-scale fusion process. It can serve as the central node for spatial fusion operations, and its phase transition front dynamic range will be reorganized based on its thermodynamic neighbors. Selecting any functional partition as the target functional partition can be done by selecting the currently processed functional partitions in a preset order. Furthermore, selecting any functional partition as the target functional partition can be achieved by scanning line by line in spatial coordinate order and dynamically prioritizing high-uncertainty regions based on phase transition activity, thereby initiating the optimized fusion process for the phase transition data of that partition.

[0079] Based on the topology of the molten salt phase change thermal diffusion network and the real-time operating conditions of the target functional partition, the dynamic range vector data of the phase change front of the target functional partition is spatially fused to obtain the recombined phase change front dynamic range of the target functional partition.

[0080] The real-time operating conditions of the target functional area can be a set of operating parameters of the target functional area at the current moment, such as local flow, pressure, and boundary heat flux. These parameters can be used to dynamically adjust the fusion weights, adapting the spatial fusion process to the current thermodynamic environment. The dynamic range vector data of the phase change front of the target functional area can be the original dynamic information of the phase change front obtained through numerical simulation. This data can be used as input data for spatial fusion, optimized and recombined under the topological constraints of the thermal diffusion network. The recombined dynamic range of the phase change front of the target functional area can be an optimized dynamic representation of the phase change front obtained by fusing its thermodynamic neighbor information under the guidance of the molten salt phase change thermal diffusion network topology. This can be used to improve the robustness and physical consistency of local phase change state estimation and suppress isolated anomalies caused by sensor noise or model errors.

[0081] Based on the topology of the molten salt phase change thermal diffusion network and the real-time operating conditions of the target functional zone, spatial fusion is performed on the dynamic range vector data of the phase change front of the target functional zone. This can be achieved by weighted averaging or graph filtering of the phase change front vectors of the target functional zone and its neighbors within its thermal diffusion network neighborhood. Furthermore, this operation can be implemented using graph convolution (the recombined vector equals the sum of the products of the neighbor weights and the neighbor vectors divided by the total weights) and adaptive weights modulated by operating conditions (reducing the influence of neighbors in high-flow areas and enhancing fusion in low-flow areas). This allows for adaptive data smoothing under physical constraints, preserving the true phase change boundary and suppressing spurious fluctuations.

[0082] Obtaining the dynamic range of the recombined phase transition front of the target functional zone can be achieved by outputting an optimized dynamic representation of the phase transition front after spatial fusion. Furthermore, obtaining the dynamic range of the recombined phase transition front of the target functional zone can be achieved by retaining the original vector direction and only correcting the amplitude, while updating the direction and velocity to match the heat flow-dominated propagation path, thereby forming a local phase transition state description that is more consistent with the actual thermophysical process.

[0083] The phase transition front dynamic range spatial fusion of the remaining functional zones is completed sequentially.

[0084] The phase transition front dynamic range of the remaining functional partitions can be the original phase transition front dynamic range vector data of all functional partitions except the current target functional partition. This data can be used sequentially as targets for spatial fusion in round-robin processing, ultimately forming a full-region reconstructed dataset. The sequential spatial fusion of the phase transition front dynamic range of the remaining functional partitions can be achieved by iteratively executing the target functional partition selection and fusion operations until all functional partitions are covered. Furthermore, the sequential spatial fusion of the phase transition front dynamic range of the remaining functional partitions can be achieved by parallel processing of multiple non-adjacent functional partitions to improve efficiency and by using iterative multi-round fusion to enhance global consistency, thereby achieving consistency optimization of the phase transition dynamic data across the entire region.

[0085] By integrating the dynamic range data of the recombined phase change frontier of all functional zones, a phase change heat distribution map of the molten salt energy storage unit is generated.

[0086] The recombined phase change front dynamic range data of all functional zones can be a globally optimized dynamic phase change data set obtained after spatial fusion of all functional zones, which can be used as direct input for generating a high-fidelity phase change heat distribution map. By integrating the recombined phase change front dynamic range data of all functional zones, the phase change heat distribution map of the molten salt energy storage unit can be generated by mapping the recombined vector data to a unified spatial grid and generating a continuous distribution map through visualization algorithms. Furthermore, this operation can be achieved by generating a heat map based on the modulus of the recombined vectors and combining directional information to generate a phase change propulsion field map with arrows, thereby outputting a high-fidelity, physically consistent spatial distribution map of phase change efficiency, supporting accurate identification of inefficient areas.

[0087] Taking the monitoring of component segregation in molten salt storage tanks for industrial waste heat recovery as an example, the online monitoring method for molten salt energy storage in this embodiment can be used in a waste heat recovery system of a chemical plant where local segregation of sodium nitrate and potassium nitrate occurs in the molten salt storage tank due to long-term low flow rate operation. The system first calibrates the thermal resistance distance between each 1-meter cubic functional zone through historical thermal response tests, and then uses distributed Raman spectroscopy to invert local components and calculate the component similarity coefficient. The constructed molten salt phase change heat diffusion network shows that the connection weight between the bottom southwest corner and the main region is extremely low. During the heat release process, the original phase change front vector in this region shows abnormal hysteresis. During fusion, due to the high thermal resistance and large component differences, the contribution weight of its neighbors is automatically suppressed, and the recombination result still retains its independent hysteresis characteristics, rather than being averaged out. The final phase change heat distribution map clearly identifies this region as an inefficient accumulation area, and the system accordingly starts local stirring in advance to avoid the risk of solidification.

[0088] In one embodiment, identifying multiple inefficient latent heat throughput accumulation zones in the energy storage unit based on phase change heat distribution maps specifically includes: Determine the set of regions in the phase change heat distribution map where the advance rate of the phase change front is lower than a second preset threshold.

[0089] The phase transition front advance rate can be the distance the phase transition interface moves in space per unit time, characterizing the local phase transition dynamic activity level. It can be used as a quantitative indicator of dynamic phase transition efficiency to identify phase transition lag regions. In this embodiment, the phase transition front advance rate can be obtained by calculating the magnitude of the phase transition front dynamic range vector data or by estimating it using the time derivative of the phase transition degree index. Further, the phase transition front advance rate can be one or more of radial advance rate, axial advance rate, and normal advance rate. The second preset threshold can be a preset critical value used to determine whether the phase transition front advance rate is abnormally low. It can be used to screen potential phase transition lag regions and form a preliminary abnormal region set. In an exemplary embodiment, the second preset threshold can be set according to system design parameters or historical operation statistics. The region set can be a set of all spatial locations in the phase transition heat distribution map where the phase transition front advance rate is lower than the second preset threshold. It can be used as a candidate input for identifying inefficient clusters and narrowing the scope of subsequent analysis. For example, the region set can be generated by threshold filtering of the phase transition heat distribution map using a spatial mask method.

[0090] Determining the set of regions in the phase change heat distribution map where the advance rate of the phase change front is lower than a second preset threshold can be achieved by extracting the advance rate of the phase change front at each location in the phase change heat distribution map, filtering out locations with advance rates lower than the second preset threshold, and aggregating them into a set of regions. Furthermore, this operation can be implemented by calculating the advance rate based on the modulus of the dynamic range vector data of the phase change front and performing threshold filtering, or by estimating the local advance rate using the time derivative of the phase change degree index and then generating a spatial mask. This transforms dynamic phase change efficiency anomalies into a processable spatial candidate set, enhancing the physical basis for anomaly detection.

[0091] Based on regional set and historical phase change performance data, multiple inefficient latent heat throughput clusters in the energy storage unit are identified through spatial clustering algorithms.

[0092] Historical phase change efficiency data can be time-series data recorded by molten salt energy storage units during past operating cycles, including phase change efficiency, distribution of inefficient areas, and performance degradation trajectories. This data can be used to provide long-term system behavior patterns and help distinguish between true inefficient areas and transient noise. In one specific embodiment, historical phase change efficiency data may include, but is not limited to, one or more of the following: historical inefficient area maps, phase change efficiency decay curves, and fault-related event logs. Spatial clustering algorithms can be unsupervised learning methods that group geographical or geometric locations based on spatial proximity and density features. These algorithms can be used to identify inefficient latent heat throughput clusters with spatial continuity and structural consistency from a set of regions. Furthermore, spatial clustering algorithms can employ DBSCAN clustering, hierarchical clustering, spectral clustering, etc.

[0093] Based on the region set and historical phase change performance data, multiple inefficient latent heat throughput clusters in energy storage units are identified using a spatial clustering algorithm. This can be achieved by taking the region set as spatial point input, combining it with prior weights or constraints provided by historical phase change performance data, and then running the spatial clustering algorithm to output connected inefficient regions. Furthermore, this operation can be implemented by introducing historical inefficiency probabilities as point density weighting factors into the DBSCAN algorithm, or by using hierarchical clustering with historical fault regions as guiding anchor points for top-down segmentation. This allows for the fusion of current state and historical patterns, enhancing the robustness and physical consistency of inefficient region identification, and reducing false alarms and missed alarms.

[0094] Taking the performance evaluation of molten salt storage tanks for industrial waste heat recovery as an example, the online monitoring method for molten salt energy storage in this embodiment can be used in a waste heat recovery system of a steel plant where the molten salt storage tank experiences a localized decrease in charging efficiency after long-term operation. The phase change heat distribution map generated by the online monitoring system shows that the phase change front advancement rate in several regions in the middle of the tank is consistently below 0.5 mm / min (the second preset threshold), forming a preliminary region cluster. The system retrieves historical phase change performance data from the past 6 months and finds that some regions have appeared multiple times in the inefficiency records. Using the DBSCAN spatial clustering algorithm, with Euclidean distance and historical inefficiency frequency as joint metrics, two significant latent heat throughput inefficiency clusters are identified: one located below the inlet due to a flow dead zone, and the other near the tank wall due to increased heat loss caused by the aging of the insulation layer. Based on this, the system issues graded warnings, guiding maintenance personnel to prioritize the repair of the insulation layer and optimize the molten salt inlet flow velocity distribution.

[0095] In one embodiment, determining the latent heat flux of the molten salt phase change process per unit time within each inefficient latent heat flux accumulation zone based on the gradient change characteristics of each microthermal environment parameter specifically includes: Calculate the spatial gradient distribution characteristics of each micro-thermal environment parameter.

[0096] The spatial gradient distribution features can be expressed as a vector field or tensor field representing the rate of change of micro-thermal environment parameters in three-dimensional space, reflecting the driving force and non-uniformity of local heat transfer. In this embodiment, the spatial gradient distribution features can be used as the physical basis for constructing a baseline latent heat throughput flux, characterizing the heat flow path and thermal resistance distribution characteristics within the region. Furthermore, the spatial gradient distribution features can include, but are not limited to, one or more of the following: temperature gradient tensor, heat flux density gradient field, and component concentration gradient vector. Calculating the spatial gradient distribution features of each micro-thermal environment parameter can be achieved by performing partial derivative or difference operations on the micro-thermal environment parameters on a spatial grid to generate a gradient vector field. For example, this operation can be achieved by calculating the first / second-order gradient on a structured grid using the finite difference method, or by using the Sobel or Scharr operators to perform image gradient estimation on the field after interpolation of discrete sensing points, thereby extracting spatial non-uniformity features reflecting local heat transfer efficiency and the causes of anomalies.

[0097] Select any inefficient latent heat throughput region as the target aggregation region.

[0098] The target cluster region can be a specific area selected as the current computational object from a set of inefficient latent heat throughput cluster regions. In an exemplary embodiment, the target cluster region can be used as an iterative unit for region-by-region flux computation, supporting sequential quantification evaluation of multiple inefficient regions. Furthermore, the target cluster region can receive spatial information from gradient distribution characteristics and serve as the input region for a thermodynamic dynamic model. Selecting any inefficient latent heat throughput cluster region as the target cluster region can be achieved by selecting a region from a list of identified inefficient cluster regions in a preset order (e.g., spatial location, area size) for processing. In a specific embodiment, this operation can be implemented by traversing from top to bottom according to the region's center coordinates, or by prioritizing the region with the largest area or the lowest predicted flux, thereby enabling ordered and independent quantification analysis of multiple anomalous regions.

[0099] The baseline latent heat throughput flux of the target cluster area is calculated based on the gradient distribution characteristics.

[0100] The baseline latent heat throughput flux can be a reference latent heat exchange capacity of the target agglomeration region under ideal thermal coupling conditions, derived from gradient distribution characteristics. In this embodiment, the baseline latent heat throughput flux can be used to provide a standardized comparison benchmark for calibrating property deviations and boundary condition effects in actual flux calculations. Furthermore, the baseline latent heat throughput flux can be obtained by substituting gradient distribution characteristics into a simplified steady-state heat conduction model or empirical correlation formula. Calculating the baseline latent heat throughput flux of the target agglomeration region based on gradient distribution characteristics can be achieved by substituting the average or dominant gradient characteristics within the target agglomeration region into a simplified thermodynamic relationship to output a reference flux value. For example, this operation can be achieved by estimating the baseline heat flux based on Fourier's law combined with the average temperature gradient, or by using the flux-gradient mapping relationship of similar regions under historical normal operating conditions to back-calculate the baseline value, thereby establishing a benchmark for the heat exchange capacity that the region should possess under ideal conditions for subsequent deviation analysis.

[0101] Based on the baseline latent heat throughput flux and the microthermal environment parameters of the target agglomeration area, the latent heat throughput flux of the molten salt phase change process in the target agglomeration area per unit time is calculated using a thermodynamic dynamic model.

[0102] The thermodynamic dynamic model can be a numerical model that integrates latent heat of phase change, unsteady-state heat conduction equations, and time-varying relationships of molten salt properties. In a specific embodiment, the thermodynamic dynamic model can be used to dynamically simulate the phase change rate and effective heat exchange process of molten salt in a real microthermal environment within the target agglomeration zone. Furthermore, the thermodynamic dynamic model can include, but is not limited to, enthalpy-porosity transient models, local non-equilibrium phase change models, and heat-mass transfer models coupled with component diffusion. Based on the baseline latent heat throughput flux and the microthermal environment parameters of the target agglomeration zone, the latent heat throughput flux of the molten salt phase change process per unit time within the target agglomeration zone is calculated using the thermodynamic dynamic model. This can be achieved by using the baseline flux as an initial or boundary constraint, combined with measured microthermal environment parameters to drive the thermodynamic dynamic model for transient solution. In an exemplary embodiment, this operation can be achieved by embedding a custom phase change material property library into multiphysics simulation software or engineering simulation software for transient simulation, or by using a reduced-order model (ROM) to quickly solve the local energy equations online, thereby obtaining a real effective flux value considering property nonlinearity, phase change hysteresis, and local thermal resistance.

[0103] Calculate the latent heat flux of the remaining inefficient latent heat flux accumulation zones in turn.

[0104] The latent heat throughput flux of the remaining inefficient clusters is calculated sequentially. This can be achieved by iteratively executing the aforementioned steps, quantifying the flux of each remaining inefficient cluster. Furthermore, this operation can be parallelized to process multiple independent clusters to improve computational efficiency, or an incremental update strategy can be adopted to recalculate only regions with significant state changes. This allows for a complete performance profile of the inefficient regions across the entire system, supporting multi-point collaborative early warning and control.

[0105] Taking the performance evaluation of an industrial waste heat recovery molten salt thermal storage system as an example, the online monitoring method for molten salt energy storage in this embodiment can be implemented in a waste heat recovery system of a steel plant, where monitoring reveals two inefficient latent heat transfer accumulation zones in the middle of the storage tank. The system first calculates the spatial gradient distribution characteristics of temperature and heat flux density in this area, identifying one zone due to lateral heat leakage caused by insulation layer damage, and another zone due to flow dead zone formed by inlet channel offset. The former is selected as the target accumulation zone, and the baseline latent heat transfer flux is calculated to be 85 kW / m³ based on its radial temperature gradient. Then, its actual micro-thermal parameters (including the melting point increase caused by local component segregation) are input into the thermodynamic dynamic model, yielding a true flux of only 42 kW / m³. Subsequently, the same process is performed on the second zone, obtaining a flux of 38 kW / m³. Both are below the warning threshold of 60 kW / m³, so the system sends a graded warning to the control center and suggests activating local electric heating and adjusting the opening of the molten salt distribution valve to restore heat exchange efficiency before solidification occurs.

[0106] Furthermore, to achieve the above objectives, the present invention also provides an online monitoring system for molten salt energy storage, the system comprising: The multi-source sensing module is used to periodically collect thermal data of each functional zone of the molten salt energy storage unit in operation by utilizing an online monitoring system equipped with a distributed temperature sensing network and infrared thermal imaging equipment, to obtain the molten salt temperature field distribution data of each functional zone, and to analyze the phase transition degree index of each functional zone based on a deep learning model. The phase change modeling module is used to determine the thermodynamic state characteristics of each functional zone based on the real-time operating parameter sequence of the molten salt energy storage unit and the phase change degree index of each functional zone. The module then uses these thermodynamic state characteristics to perform numerical simulation of the spatiotemporal evolution behavior of the phase change process in the molten salt energy storage unit, and obtains the dynamic range vector data of the phase change front of each functional zone. The map fusion module is used to perform multi-scale fusion and recombination of all phase change front dynamic range vector data based on the heat conduction topology relationship between each functional zone and the correlation degree of molten salt components, to generate a phase change heat distribution map of the molten salt energy storage unit, and to identify multiple latent heat throughput inefficient accumulation areas in the energy storage unit based on the phase change heat distribution map. The flux quantization module is used to collect the micro-thermal environment parameters of each latent heat inefficient accumulation zone and determine the latent heat influent flux of the molten salt phase change process in each latent heat inefficient accumulation zone per unit time based on the gradient change characteristics of each micro-thermal environment parameter. The intelligent early warning module is used to send an early warning signal of abnormal latent heat throughput efficiency to the energy storage system integrated management and control center when the latent heat throughput is lower than the preset efficiency early warning threshold.

[0107] Other embodiments or specific implementations of the molten salt energy storage online monitoring system of the present invention can be referred to the above-described method embodiments, and will not be repeated here.

[0108] In addition, to achieve the above objectives, the present invention also provides an online monitoring device for molten salt energy storage, the device comprising: a memory, a processor, and an online monitoring program for molten salt energy storage stored in the memory and executable on the processor, the online monitoring program for molten salt energy storage being configured to implement the steps of the online monitoring method for molten salt energy storage as described above.

[0109] In addition, to achieve the above objectives, the present invention also provides a medium storing a molten salt energy storage online monitoring program, which, when executed by a processor, implements the steps of the molten salt energy storage online monitoring method as described above.

[0110] The above are merely preferred embodiments of the present invention and do not limit the scope of the patent. Any equivalent structural or procedural transformations made based on the description and drawings of the present invention, or direct or indirect applications in other related technical fields, are similarly included within the scope of patent protection of the present invention.

Claims

1. A method for online monitoring of molten salt energy storage, characterized in that, The method includes: Using an online monitoring system equipped with a distributed temperature sensing network and infrared thermal imaging equipment, periodic thermal data are collected from each functional zone of the molten salt energy storage unit in operation to obtain the molten salt temperature field distribution data of each functional zone. The phase transition degree index of each functional zone is analyzed based on a deep learning model. Based on the real-time operating parameter sequence of the molten salt energy storage unit and the phase change degree index of each functional zone, the thermodynamic state characteristics of each functional zone are determined. The spatiotemporal evolution behavior of the phase change process in the molten salt energy storage unit is numerically simulated using the thermodynamic state characteristics to obtain the dynamic range vector data of the phase change front of each functional zone. Based on the heat conduction topology between each functional zone and the correlation of molten salt components, multi-scale fusion and recombination of all phase change front dynamic range vector data are performed to generate a phase change heat distribution map of the molten salt energy storage unit. Based on the phase change heat distribution map, multiple latent heat throughput inefficient accumulation areas in the energy storage unit are identified. Collect microthermal environment parameters of each latent heat inefficient accumulation zone, and determine the latent heat flux of the molten salt phase change process in each latent heat inefficient accumulation zone per unit time based on the gradient change characteristics of each microthermal environment parameter. When the latent heat throughput is lower than the preset performance warning threshold, an abnormal latent heat throughput performance warning signal is sent to the energy storage system integrated management and control center.

2. The online monitoring method for molten salt energy storage as described in claim 1, characterized in that, The phase transition index for analyzing each functional partition based on a deep learning model includes: Acquire thermal remote sensing images of each functional zone and their corresponding spatiotemporal coordinate information; A pre-trained convolutional neural network model was used to extract the thermophysical property parameters and phase transition-specific thermal spectrum features of molten salt in thermal images. Based on the molten salt thermophysical property characteristics and phase transition-specific thermal spectrum characteristics, the standardized phase transition degree index of each functional zone is calculated.

3. The online monitoring method for molten salt energy storage as described in claim 1, characterized in that, The real-time operating parameter sequence based on the molten salt energy storage unit and the phase transition degree index of each functional zone are used to determine the thermodynamic state characteristics of each functional zone, including: Acquire spatial location information of each functional zone and real-time operation monitoring data of the molten salt energy storage unit; The spatial heterogeneity index of the phase transition process is calculated based on the phase transition degree index of each functional zone; By using an ensemble learning algorithm, combined with the real-time operation monitoring data and the spatial heterogeneity index value of the phase change process, the thermodynamic state characteristics of each functional zone are determined.

4. The online monitoring method for molten salt energy storage as described in claim 1, characterized in that, The method involves numerically simulating the spatiotemporal evolution of the phase transition process in the molten salt energy storage unit based on the thermodynamic state characteristics, to obtain dynamic range vector data of the phase transition front for each functional zone, including: Select any functional zone as the target functional zone, and determine the molten salt phase transition stage parameters and thermal diffusion direction vector set of the target functional zone; The phase transition front advance rate of the target functional zone is calculated based on the thermodynamic state characteristic parameters corresponding to the target functional zone and the molten salt phase transition stage parameters. Based on the vector set of phase change front propagation rate and thermal diffusion direction, the dynamic range vector data of the phase change front of the target functional area is generated; the dynamic range vector data of the phase change front of the remaining functional areas is calculated in sequence.

5. The online monitoring method for molten salt energy storage as described in claim 1, characterized in that, Based on the heat conduction topology between functional zones and the correlation of molten salt components, multi-scale fusion and recombination of all phase change front dynamic range vector data are performed to generate a phase change heat distribution map of the molten salt energy storage unit, including: The thermal resistance distance and molten salt composition similarity coefficient between each functional zone were determined. A molten salt phase change heat diffusion network topology is constructed based on the thermal resistance distance between each functional zone and the molten salt composition similarity coefficient. Select any functional partition as the target functional partition; Based on the topology of the molten salt phase change heat diffusion network and the real-time operating conditions of the target functional partition, the dynamic range vector data of the phase change front of the target functional partition is spatially fused to obtain the recombined phase change front dynamic range of the target functional partition. The phase change front dynamic range spatial fusion of the remaining functional zones is completed sequentially; the phase change heat distribution map of the molten salt energy storage unit is generated by integrating the recombined phase change front dynamic range data of all functional zones.

6. The online monitoring method for molten salt energy storage as described in claim 1, characterized in that, The identification of multiple inefficient latent heat throughput accumulation zones in the energy storage unit based on the phase change heat distribution map specifically includes: Determine the set of regions in the phase change heat distribution map whose phase change front advance rate is lower than a second preset threshold. Based on the aforementioned region set and historical phase change performance data, multiple inefficient latent heat throughput clusters in the energy storage unit are identified using a spatial clustering algorithm.

7. The online monitoring method for molten salt energy storage as described in claim 1, characterized in that, The determination of the latent heat flux of the molten salt phase transition process per unit time in each latent heat inefficient accumulation zone based on the gradient change characteristics of each microthermal environment parameter includes: Calculate the spatial gradient distribution characteristics of each micro-thermal environment parameter; Select any inefficient accumulation zone of latent heat throughput as the target accumulation zone; Calculate the baseline latent heat throughput flux of the target aggregation region based on the gradient distribution characteristics; Based on the baseline latent heat throughput and the microthermal environment parameters of the target aggregation area, the latent heat throughput of the molten salt phase change process in the target aggregation area per unit time is calculated by a thermodynamic dynamic model. Calculate the latent heat flux of the remaining inefficient latent heat flux accumulation zones in turn.

8. An online monitoring system for molten salt energy storage, characterized in that, The system includes: The multi-source sensing module is used to periodically collect thermal data of each functional zone of the molten salt energy storage unit in operation by utilizing an online monitoring system equipped with a distributed temperature sensing network and infrared thermal imaging equipment, to obtain the molten salt temperature field distribution data of each functional zone, and to analyze the phase transition degree index of each functional zone based on a deep learning model. The phase change modeling module is used to determine the thermodynamic state characteristics of each functional zone based on the real-time operating parameter sequence of the molten salt energy storage unit and the phase change degree index of each functional zone. The module then uses these thermodynamic state characteristics to perform numerical simulation of the spatiotemporal evolution behavior of the phase change process in the molten salt energy storage unit, and obtains the dynamic range vector data of the phase change front of each functional zone. The map fusion module is used to perform multi-scale fusion and recombination of all phase change front dynamic range vector data based on the heat conduction topology relationship between each functional zone and the correlation degree of molten salt components, to generate a phase change heat distribution map of the molten salt energy storage unit, and to identify multiple latent heat throughput inefficient accumulation areas in the energy storage unit based on the phase change heat distribution map. The flux quantization module is used to collect the micro-thermal environment parameters of each latent heat inefficient accumulation zone and determine the latent heat influent flux of the molten salt phase change process in each latent heat inefficient accumulation zone per unit time based on the gradient change characteristics of each micro-thermal environment parameter. The intelligent early warning module is used to send an early warning signal of abnormal latent heat throughput efficiency to the energy storage system integrated management and control center when the latent heat throughput is lower than the preset efficiency early warning threshold.

9. An online monitoring device for molten salt energy storage, characterized in that, The device includes: a memory, a processor, and a molten salt energy storage online monitoring program stored in the memory and executable on the processor, the molten salt energy storage online monitoring program being configured to implement the steps of the molten salt energy storage online monitoring method as described in any one of claims 1 to 7.

10. A medium, characterized in that, The medium stores an online monitoring program for molten salt energy storage, which, when executed by a processor, implements the steps of the online monitoring method for molten salt energy storage as described in any one of claims 1 to 7.

Images