Leakage detection and early warning device for photo-thermal fused salt storage tank

Abstract

The invention discloses a photo-thermal fused salt storage tank leakage detection early warning device. The device comprises a multi-sensing detection unit, a data acquisition and processing unit and an intelligent analysis early warning unit which are installed in a fused salt tank detection cavity. The multi-sensor detection unit obtains a tank wall temperature field, a leakage sound wave signal and a characteristic gas concentration in real time; the data acquisition and processing unit synchronously acquires and preprocesses multi-source data; a leakage judgment model based on multi-source information fusion is arranged in the intelligent analysis and early warning unit, and a graded early warning signal containing the leakage probability, the estimated position and the severity level is output by conducting intelligent analysis and cross validation on temperature abnormity, acoustic characteristics and gas changes; according to the invention, through a multi-dimensional sensing cooperation and intelligent evidence fusion mechanism, early capture of tiny leakage, accurate positioning of leakage points and effective suppression of environmental interference are realized, so that the reliability, timeliness and decision support capability of leakage detection are greatly improved.

Description

Technical Field

[0001] This invention relates to the field of molten salt tank leakage detection technology, and in particular to a solar thermal molten salt tank leakage detection and early warning device. Background Technology

[0002] Molten salt tanks are equipment in the thermal storage system of solar thermal power plants. Currently, there are some technical solutions in the industry for leak monitoring of large molten salt tanks. For example, a small number of point temperature sensors are pre-embedded in the insulation layer of the tank foundation to indirectly determine whether a leak has occurred by monitoring local temperature anomalies. Another solution proposes to install weighing sensors at the bottom of the tank to infer leaks by monitoring changes in the total weight of the tank. These methods constitute the main existing monitoring methods in this field.

[0003] Existing technologies rely on single parameters (such as temperature at an individual point or total weight) for judgment, resulting in insufficient reliability. Point-type temperature sensors have limited coverage and may fail to detect minute leak initiation points far from the sensor in a timely manner. Furthermore, local temperature fluctuations are easily affected by external environmental factors or internal molten salt convection, leading to false alarms or missed alarms. Weighing methods are insensitive to slowly occurring minute leaks and cannot provide location information for the leak point, making it difficult to provide accurate guidance for emergency repairs. Moreover, existing methods are mostly passive reactive monitoring, typically only triggering an alarm after the leak has developed to a certain extent and caused significant physical or temperature changes. They lack the ability to detect early, subtle leak characteristics, resulting in insufficient timeliness of warnings and potentially leading to the escalation of accidents.

[0004] Therefore, in response to the problems mentioned above, this invention proposes a leakage detection and early warning device for a solar thermal molten salt storage tank. Summary of the Invention

[0005] To overcome the problems of existing technologies, such as limited monitoring methods, poor reliability, inability to locate leaks, and delayed early warnings, this invention proposes a leak detection and early warning device for molten salt tanks in solar thermal energy storage. By simultaneously monitoring multi-source information such as temperature field, acoustic signals, and characteristic gas concentrations, it can achieve early and accurate identification and location of leaks in molten salt tanks, thereby improving the operational safety of the thermal energy storage system.

[0006] The technical solution of this invention is: a leakage detection and early warning device for a solar thermal molten salt storage tank, comprising:

[0007] A multi-sensor detection unit, installed in a sealed detection chamber between the molten salt tank and the basic insulation layer, is used to acquire multi-dimensional physicochemical parameters related to leakage in real time. This unit integrates:

[0008] The distributed fiber optic temperature measurement network consists of multiple high-temperature resistant (preferably with long-term tolerance >600℃) single-mode optical fibers laid close to the outer wall and bottom of the tank according to a preset grid pattern, used for continuous monitoring of temperature field distribution. The density of the grid pattern is set differently according to the risk level of the area. In the load-bearing area at the center of the tank bottom and the heat-affected zone of the circumferential weld and T-weld, the path spacing is preferably 0.3-0.5 meters to form a high-density monitoring area. In other areas of the tank wall, the path spacing can be 0.8-1.2 meters.

[0009] The acoustic sensor array includes multiple high-temperature resistant broadband acoustic sensors (preferably operating frequency range of 1kHz-100kHz) mounted in an equally spaced matrix on the upper surface of the concrete foundation layer at the bottom of the tank. These sensors are used to collect structural acoustic and stress wave signals generated during leakage molten salt impact or seepage. Each acoustic sensor is coupled to the foundation layer through a high-temperature resistant alloy waveguide rod. One end of the waveguide rod is connected to the sensor's sensitive element, and the other end is embedded and anchored within the foundation layer. This enhances the ability to capture weak conducted sound waves and isolates some environmental noise.

[0010] The gas composition monitoring module includes multiple suction sampling probes installed at the top, middle, and bottom of the detection chamber and a gas analyzer connected to them via sampling pipelines. This module is used to detect in real time the concentration of characteristic gases generated by the thermal decomposition of molten salt (especially nitrate) after leakage or by its reaction with insulation materials (such as rock wool). The characteristic gases include at least nitrogen oxides and chlorine. The gas analyzer preferably employs tunable diode laser absorption spectroscopy technology to achieve high sensitivity and rapid online measurement of specific gas components.

[0011] The data acquisition and processing unit communicates with each sensor of the multi-sensor detection unit via high-temperature shielded cable and optical cable. It is used to synchronously acquire, condition, convert analog to digital and preprocess raw data of temperature, sound wave and gas concentration. The unit has the ability to acquire data in a high-speed parallel manner through multiple channels, and performs bandpass filtering and time-frequency domain preprocessing on the sound wave signal, and performs spatial interpolation calculation on the temperature data to generate temperature field cloud map.

[0012] The intelligent analysis and early warning unit, connected to the data acquisition and processing unit, has a built-in leakage detection model trained based on machine learning algorithms. This unit receives pre-processed multi-source data and performs feature extraction and fusion analysis on it, specifically including:

[0013] (1) Analyze the abnormal low-temperature diffusion regions and their gradient change patterns in the temperature field;

[0014] (2) Identify the energy surge and spectral characteristics of specific frequency bands (such as 10-30kHz) in acoustic signals that are related to leakage impact;

[0015] (3) Calculate the correlation between the rate of change of characteristic gas concentration and spatial distribution;

[0016] (4) The leakage determination model (e.g., based on deep neural networks or support vector machines) takes the above multi-source feature vectors as input, calculates the probability of leakage events, and outputs a graded early warning signal containing the estimated leakage location coordinates and the initial scale level of the leakage.

[0017] Preferably, the device further includes an environmental compensation and verification unit, which includes multiple environmental temperature and humidity sensors and a broadband background noise sensor evenly arranged in the detection chamber. The data acquisition and processing unit uses real-time acquired environmental temperature and humidity data to perform drift compensation on the measured values ​​of the distributed fiber optic temperature measurement network, and uses the background noise spectrum to perform adaptive noise suppression on the signals acquired by the acoustic sensor array. The intelligent analysis and early warning unit only confirms the triggering of a leak warning after the abnormal characteristics of multi-source data (temperature, sound waves, gas) are correlated in time and space, and environmental interference (such as noise generated by heavy rain impacting the tank wall, local temperature changes caused by maintenance activities) is excluded, thereby significantly reducing the false alarm rate.

[0018] Preferably, the intelligent analysis and early warning unit performs multi-source information fusion analysis, specifically employing an improved DS evidence theory algorithm or a Bayesian network model. It uses "temperature anomaly evidence" from the distributed fiber optic temperature measurement network, "acoustic anomaly evidence" from the acoustic sensor array, and "gas anomaly evidence" from the gas composition monitoring module as three independent evidence sources. Data from the environmental compensation and verification unit is introduced as a confidence weighting factor for joint inference, ultimately outputting a quantified comprehensive confidence level as the leakage probability. The graded early warning signal is triggered based on the comprehensive confidence level threshold: for example, a comprehensive confidence level between 60% and 80% triggers a Level 1 warning (enhanced monitoring), between 80% and 95% triggers a Level 2 warning (prepared for intervention), and above 95% triggers a Level 3 warning (emergency response).

[0019] Preferably, the device further includes a corrosion monitoring unit, which comprises multiple corrosion rate probes (such as sensors based on the principle of galvanometers or linear polarization resistance) embedded in or attached to the anti-corrosion layer on the upper surface of the tank bottom foundation layer in a grid pattern. The corrosion monitoring unit is connected to the data acquisition and processing unit. When a leak occurs, the molten salt that seeps into the foundation layer causes the corrosion rate detected by the probes to rise sharply. The corrosion rate data is sent to the intelligent analysis and early warning unit as an auxiliary basis for assessing the severity of the leak, especially its impact on the safety of the foundation structure, and is used to correct the estimated level of the leak.

[0020] Preferably, the device also integrates a visualization interface and a communication module; the module includes a local industrial computer and / or a remote cloud server, providing a graphical human-machine interface for dynamically displaying a three-dimensional cloud map of the tank temperature field, a real-time spectrum of acoustic signals, characteristic gas concentration change curves, the status of each sensor and early warning information, and displaying the estimated leak location on the three-dimensional model. Its communication module (supporting 4G / 5G or fiber optic Ethernet) can synchronously send all early warning information and key data to the power plant central control room in accordance with historical trends.

[0021] The beneficial effects of this invention are:

[0022] 1. This invention constructs a multi-dimensional sensing system that coordinates a distributed optical fiber temperature measurement network, an acoustic sensor array, and a gas composition monitoring module. Combined with an intelligent analysis model based on multi-source information fusion, it achieves synchronous capture and cross-verification of weak signals throughout the entire chain of leakage in a solar thermal molten salt storage tank, from physical impact and heat conduction to chemical changes. This elevates leak detection from the traditional lagging and rough judgment based on a single parameter to early and accurate intelligent diagnosis supported by multiple evidence chains, thus solving the problem of delayed early warning.

[0023] 2. The improved DS evidence theory fusion algorithm of this invention combines three independent evidence sources—temperature anomaly patterns, acoustic spectrum characteristics, and gas concentration change rates—with real-time environmental compensation data for dynamic credibility weighting and joint inference. This enables the system to trigger an early warning only when multi-source anomaly signals have spatiotemporal correlation, significantly eliminating false alarms from a single sensor caused by local insulation damage, external rainfall impact, or fluctuations in normal operating conditions. This achieves high-reliability monitoring with near-zero false alarms and solves the problem of poor reliability in traditional methods. Attached Figure Description

[0024] Figure 1 The diagram shown is a schematic representation of the system framework of this invention. Detailed Implementation

[0025] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are some embodiments of the present invention, but not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0026] Please see Figure 1 The present invention provides an embodiment: a leakage detection and early warning device for a solar thermal molten salt storage tank, comprising:

[0027] In this embodiment, the multi-sensor detection unit will be described in detail:

[0028] (1) The distributed optical fiber temperature measurement network uses a metal-armored single-mode temperature measurement optical cable that can withstand high temperatures (long-term operating temperature ≥ 800 °C). Before laying, plan the path on the outer wall of the clean and dry tank (before installing the insulation layer). In the bottom area of the tank, use a "return" shaped dense grid laying method. Especially in the central bearing area at the bottom of the tank and within 0.5 meters on both sides of all circumferential welds and T-shaped welds, the distance between optical cables is controlled within 0.3 - 0.5 meters to form a high-sensitivity monitoring area. In the tank wall area, use a laying method that combines spiral upward and horizontal circumferential directions, with a spacing of 1.0 meters. The optical cable needs to be tightly fixed to the tank wall using high-temperature-resistant binding tapes or welding jigs to ensure good thermal contact. The beginning and end of the entire optical cable (usually a single length can reach several kilometers) are connected to the distributed optical fiber temperature measurement host located in the electrical cabinet. The host can demodulate the temperature values of thousands of continuous measurement points along the optical cable in real time with a spatial resolution not lower than 0.25 meters, a temperature measurement accuracy of ±1 °C, and a refresh rate of up to 1 second / time, and reconstruct the two-dimensional / three-dimensional temperature field of the entire tank wall and tank bottom.

[0029] (2) On the upper surface of the concrete foundation layer at the bottom of the tank, before the construction of the anti-corrosion layer, install high-temperature-resistant broadband acoustic sensors according to an equidistant matrix of 3 meters × 3 meters. The sensors are piezoelectric ceramic elements encapsulated in a corrosion-resistant alloy shell, with a working frequency band of 1 kHz - 100 kHz. Each sensor is coupled through a waveguide rod about 0.3 meters long and 10 mm in diameter. The lower end of the waveguide rod is processed with threads and directly anchored in the sleeve embedded in the concrete, and the upper end is connected to the sensor housing through threads. This design can efficiently conduct the structural stress waves and acoustic emission signals caused by the dripping, impact, and seepage of leaked molten salt to the sensors, while suppressing the interference of environmental air acoustic waves. All sensors are connected to a multi-channel acoustic data acquisition instrument through shielded cables, with a sampling rate not lower than 250 kSPS to capture high-frequency signals.

[0030] (3) In the detection chamber, at three height levels of the top (accumulated gas), middle (main space), and bottom (near the possible leakage point), install aspirating sampling probes respectively. The probes are equipped with sintered metal filters to prevent dust inhalation. The gas sample is transported to the centrally installed multi-channel TDLAS gas analyzer through a heat-traced sampling pipe (maintaining about 120 °C to prevent gas condensation). The analyzer conducts highly selective measurements on nitrogen oxides that may be produced by the decomposition of nitrate molten salt leakage and chlorine gas that may be produced by impurities in the raw materials. The response time of each measurement channel is less than 10 seconds, and the detection limit can reach ppm or even ppb levels. The system is regularly flushed with zero gas and high-standard gas for automatic calibration.

[0031] In this embodiment, the data acquisition and processing unit is described specifically as follows:

[0032] This unit is an industrial-grade data acquisition server equipped with a multi-slot backplane and includes a DTS host communication card (for receiving temperature field data), a high-speed multi-function I / O card (for acquiring analog / digital signals from all acoustic sensors and gas analyzers), and a synchronization clock card (for stamping all data channels with a unified and accurate time stamp to ensure strict time synchronization of multi-source data).

[0033] This unit preprocesses the acquired data, including: spatially interpolating the raw DTS temperature data to generate higher resolution temperature field grid data; performing digital bandpass filtering on the acoustic signal (e.g., retaining the main frequency band of 5-80 kHz) and calculating the short-time Fourier transform to obtain the real-time spectrum; and performing moving average filtering on the gas concentration data and calculating its rate of change per minute.

[0034] In this embodiment, the intelligent analysis and early warning unit is described in detail:

[0035] The core of this unit is the leakage detection model, which is built and run through the following steps:

[0036] Before system commissioning or during major overhaul, model training utilizes historical normal operating data (at least one complete four-season operating cycle) to establish normal baseline models for each sensor (such as temperature field statistical characteristics, background noise spectrum, and background gas concentration), and obtains abnormal samples through simulated leak experiments. For example, when the tank is shut down for maintenance, simulated molten salt fluid (such as high-temperature heat transfer oil or low-melting-point alloy) heated to approximately 300°C is dripped at different locations at the bottom of the tank at controllable flow rates (such as 0.1 L / min, 0.5 L / min, 2 L / min), while simultaneously recording the entire system response data of DTS, acoustic waves, and gases. These "leakage-response" data pairs are used to train a multi-classification model such as a deep convolutional neural network or random forest.

[0037] During online operation, the preprocessed data stream is continuously input into the model, and the analysis process is carried out in layers:

[0038] For feature layer extraction, the following are included:

[0039] (1) The temperature feature is the difference field between the real-time temperature field and the normal baseline field. The area, perimeter, minimum temperature and temperature gradient of the "low temperature patch" appearing in the difference field are calculated. A "temperature anomaly index" TI is defined, which combines the patch area and gradient.

[0040] (2) For the real-time spectrum of each acoustic sensor, calculate the total energy in the 10-30kHz frequency band (the leakage sensitive frequency band determined by the experiment), and compare it with the historical background energy of the sensor (dynamically updated) to obtain the energy surge ratio. At the same time, use the arrival time difference of the sensor array to make a preliminary estimate of the sound source location.

[0041] (3) Calculate whether the absolute values ​​of NOx and Cl2 concentrations exceed the first-level threshold (e.g., 10 ppm) and calculate their concentration change rate (ppm / min). Since the chamber is ventilated, the absolute concentration may not be high, but the change rate is a more sensitive indicator.

[0042] Explanation regarding integration and decision-making:

[0043] Using an improved Dempster evidence framework, the three extracted features—"temperature anomaly," "acoustic anomaly," and "gas anomaly"—are transformed into three independent evidence bodies through three basic probability allocation functions. Each evidence body contains the degree of support for the proposition {leakage, non-leakage, uncertainty}. The key to fusion lies in introducing an environmental credibility factor (derived from an environmental compensation unit) (for example, if the background noise sensor detects noise from heavy rain hitting the tank wall, the weight of the acoustic evidence is temporarily reduced; if there are drastic changes in environmental temperature and humidity, the credibility of the temperature evidence is adjusted). Finally, the Dempster combination rule is used to fuse the weighted evidence to obtain a comprehensive leakage confidence level P (0-100%).

[0044] Based on the comprehensive confidence level P and the spatial correlation of features (such as whether the sound source location point intersects spatially with the low-temperature patch center or high gas concentration point), the system makes a decision:

[0045] Level 1 Warning (Observation Level): 60% ≤ P < 80%, or a single strong piece of evidence but not supported by other evidence. The system will display a flashing yellow alert on both local and remote interfaces, indicating the anomaly type and location. Operators are advised to increase monitoring frequency in the affected area.

[0046] Level 2 Warning (Confirmed): 80% ≤ P < 95%, and multiple anomalies have a reasonable spatiotemporal correlation. The system issues an orange warning, clearly marking "suspected leak," providing the coordinates of the most likely leak point (within an error range, such as ±1.5 meters), and activating a predefined emergency response checklist.

[0047] Level 3 Warning (Emergency Level): P ≥ 95%, and multi-source evidence is highly consistent, or a sharp increase in corrosion rate (from the corrosion monitoring unit) is detected. The system issues a red alert, triggers an audible and visual alarm, and automatically pushes an alarm message containing all data snapshots and location information to the mobile phones of relevant personnel and the remote monitoring center, requiring immediate execution of emergency procedures such as shutdown or tank transfer.

[0048] In this embodiment, the auxiliary unit will be described in detail:

[0049] For the environmental compensation and verification unit, temperature and humidity sensors and reference microphones (monitoring background noise) are evenly arranged in the detection chamber, and the data are used for the aforementioned evidence weight adjustment.

[0050] For the corrosion monitoring unit, a corrosion rate probe based on electrochemical principles is embedded in the anti-corrosion layer on the upper surface of the tank bottom base to monitor the corrosion current density of the metal base in real time. When the corrosion rate exceeds the normal value (e.g., <0.1 mm / year) by an order of magnitude, it serves as irrefutable evidence that the leak has affected the structure.

[0051] For the visualization and remote platform, it is based on a human-machine interface, displays the molten salt tank in the form of a 3D model, dynamically renders temperature field cloud map (using color gradient to represent temperature), displays the location of sound wave events in the form of pulses, draws gas concentration curves in real time, and uploads data to the remote cloud through industrial protocols.

[0052] This invention provides Embodiment 1:

[0053] This invention was implemented in a 50MW parabolic trough solar thermal power plant, with the system installed on a pair of cold and hot salt tanks, each 40 meters in diameter and with a molten salt capacity of approximately 28,000 cubic meters. The detection chamber was 1.2 meters wide. The total length of the DTS optical cable was 32 kilometers (per tank), and 121 acoustic sensors (11×11 matrix) were deployed per tank, along with 6 gas sampling points per tank. After the system was put into operation, it underwent a full year of operation. During this period, the system successfully filtered out 37 instances of interference from normal operating conditions caused by summer rainstorms, winter strong winds, nearby construction vibrations, and the start-up and shutdown of the molten salt pumps, without generating a single false alarm. This demonstrates the invention's strong environmental resistance and low false alarm rate.

[0054] This invention provides comparative examples:

[0055] (1) In this example, a scaled-down (1:10) molten salt tank test bench was built in the laboratory, and two comparative examples and the present invention were installed respectively, wherein:

[0056] Comparative Example 1 is a traditional point-type temperature measurement system, which has 8 thermocouples arranged at the bottom of the tank.

[0057] Comparative Example 2 uses the weighing method, in which a high-precision weighing sensor is installed on the tank support structure.

[0058] This invention involves laying a DTS optical cable and installing four acoustic wave sensors and one gas sampling point.

[0059] During the experiment, simulated molten salt at 300°C was slowly leaked at a concealed location at the bottom of the container (away from all thermocouples) at a rate of 0.05 L / min. The experiment was repeated 10 times, and the statistical results are shown in Table 1.

[0060] Table 1 Comparison of detection performance for minute leaks (0.05 L / min)

[0061] Detection methods Comparative Example 1 Comparative Example 2 This invention Average time (min) to detect a leak >120 (Not all detected) Undetectable (change is within noise) 18.5 Is it accurate (yes / no)? no no yes Number of false alarms during the experiment (due to other thermal interference) 3 0 0

[0062] As shown in the table above, traditional point-based temperature measurement relies heavily on sensor layout. Once a leak point is located between sensors, the response is extremely slow or even undetectable. The weighing method is completely insensitive to such a small flow rate. The system of this invention, through the continuous temperature field formed by DTS, can quickly sense the expansion of local low-temperature areas (even if the starting point is very small). The acoustic sensor can capture the faint dripping sound, and the gas sensor detects an abnormal concentration change rate after about 15 minutes. The multi-source information quickly increases the confidence level to over 80% in the fusion model, achieving early warning and location.

[0063] (2) On the same test bench, the cracking of the tank wall weld was simulated, and the initial leakage rate was about 2L / min. The experiment also compared the three methods, and the statistical results are shown in Table 2.

[0064] Table 2 Comparison of Detection and Response to Sudden Leakage (2L / min)

[0065] Detection methods Comparative Example 1 Comparative Example 2 This invention Alarm response time (s) 45-60 Approximately 30 8-12 Positioning error (m) >2.0 Unable to locate <0.5 Richness of decision-making information provided Limited information; only knowledge is available regarding whether a certain area is excessively hot or cold. Unable to distinguish the location of the leak Includes precise coordinates, estimated leak size, diffusion trend, and gas hazard level.

[0066] As shown in the table above, for larger leaks, while point-based temperature measurement and weighing methods can provide rapid alarms, the amount of information is severely insufficient. The acoustic array of this invention captures a strong signal and performs preliminary localization within 2-3 seconds of a leak occurring; the DTS clearly outlines the low-temperature diffusion area within 10 seconds; and the gas concentration spikes within seconds. Multi-source data is fused within 10 seconds, providing a Level 3 warning with a confidence level greater than 95%, and accurately locating the leak point.

[0067] The above are merely preferred embodiments of the present invention and are not intended to limit the present invention in any other way. Any person skilled in the art may make changes or modifications to the above-disclosed technical content to create equivalent embodiments that can be applied to other fields. However, any simple modifications, equivalent changes, and modifications made to the above embodiments based on the technical essence of the present invention without departing from the scope of the present invention shall still fall within the protection scope of the present invention.

Claims

1. A leakage detection and early warning device for a solar thermal molten salt storage tank, characterized in that, Including: The multi-sensor detection unit is installed in the detection chamber between the molten salt tank and the basic insulation layer to acquire multi-dimensional physicochemical parameters related to leakage in real time. The multi-sensor detection unit includes a distributed fiber optic temperature measurement network, an acoustic sensor array, and a gas composition monitoring module. The distributed optical fiber temperature measurement network includes multiple high-temperature resistant single-mode optical fibers, which are laid close to the outer wall of the tank in a preset grid pattern to continuously monitor the temperature field distribution in the tank wall and bottom areas. The acoustic sensor array includes multiple high-temperature broadband acoustic sensors, which are fixed in an equally spaced matrix on the upper surface of the tank bottom base layer to collect broadband acoustic signals generated when leaking molten salt impacts the base layer or insulation material. The gas composition monitoring module includes multiple suction sampling probes and a gas analyzer connected to the sampling probes. The suction sampling probes are installed at the top and bottom of the detection chamber and are used to detect the concentration of specific gases generated by the decomposition of molten salt leakage or reaction with the insulation material. The data acquisition and processing unit is connected in communication with the multi-sensor detection unit to synchronously receive and preprocess temperature, sound wave and gas concentration data; The intelligent analysis and early warning unit, connected to the data acquisition and processing unit, includes a leak detection model. This model is used to perform fusion analysis on multiple pre-processed data, identify leak characteristics, and output graded early warning signals based on the leak risk assessment results. The leak detection model is trained based on historical data under normal operating conditions and simulated leak experimental data. Its inputs are temperature anomaly gradient pattern, acoustic signal spectrum feature vector, and gas concentration change rate. The outputs are leak probability, estimated leak location, and leak severity level.

2. The leakage detection and early warning device for a solar thermal molten salt storage tank according to claim 1, characterized in that: The preset grid-like path of the distributed optical fiber temperature measurement network has a grid density where the path spacing between the central area of ​​the tank bottom and the heat-affected zone of the weld is smaller than that in other areas of the tank wall.

3. The leakage detection and early warning device for a solar thermal molten salt storage tank according to claim 1, characterized in that: Each high-temperature broadband acoustic sensor in the acoustic sensor array is equipped with a waveguide rod. One end of the waveguide rod is coupled to the sensor sensing surface, and the other end extends and is fixed in the base layer at the bottom of the tank to enhance the ability to capture acoustic waves transmitted through the structure.

4. The leakage detection and early warning device for a solar thermal molten salt storage tank according to claim 1, characterized in that: The specific gases in the gas composition monitoring module include at least nitrogen oxides and chlorine, and the gas analyzer uses tunable diode laser absorption spectroscopy technology.

5. The leakage detection and early warning device for a solar thermal molten salt storage tank according to claim 1, characterized in that: The device also includes an environmental compensation unit, which includes an environmental temperature and humidity sensor and a background noise sensor arranged in the detection chamber for monitoring environmental parameters in the chamber. The data acquisition and processing unit uses the environmental parameters to compensate and correct the data of the multi-sensor detection unit.

6. The leakage detection and early warning device for a solar thermal molten salt storage tank according to claim 1, characterized in that: The intelligent analysis and early warning unit performs fusion analysis using the DS evidence theory algorithm, which fuses information from temperature anomaly gradient patterns, acoustic signal spectral feature vectors, and gas concentration change rates as independent evidence sources to calculate the leakage probability.

7. The leakage detection and early warning device for a solar thermal molten salt storage tank according to claim 1, characterized in that, The tiered early warning signal includes at least three levels: Level 1 warning indicates the detection of early abnormal characteristics, prompting enhanced monitoring; Level 2 warning indicates the confirmation of a minor leak, prompting preparations for intervention; and Level 3 warning indicates the occurrence of a leak of a certain scale, prompting immediate emergency measures to be taken.

8. The leakage detection and early warning device for a solar thermal molten salt storage tank according to claim 1, characterized in that: The device also includes a corrosion monitoring unit, which includes multiple corrosion sensors embedded in the anti-corrosion layer on the upper surface of the tank bottom foundation layer. These sensors are used to monitor the corrosion rate of the leaked molten salt on the foundation. The signals from the corrosion sensors are input to an intelligent analysis and early warning unit as an auxiliary basis for judging the severity of the leak.

9. The leakage detection and early warning device for a solar thermal molten salt storage tank according to claim 1, characterized in that: The device also includes a visualization interface and a communication module. The visualization interface is used to display temperature field cloud map, sound wave signal spectrum map, gas concentration trend, leakage early warning information and estimated leakage location in real time. The communication module is used to send early warning information and key data to a remote monitoring center.

10. The leakage detection and early warning device for a solar thermal molten salt storage tank according to claim 1, characterized in that: The detection chamber is a pre-constructed annular sealed cavity in the insulation layer of the molten salt tank foundation.

Images