An energy storage system operating state determination system

Abstract

This invention relates to the field of data processing technology and discloses a system for determining the operating status of an energy storage system. The system includes a data acquisition module for acquiring temperature and remote sensing images of a water storage tank at fixed time intervals; a data filtering module for analyzing the temperature of the water storage tank at multiple times to determine temperature fluctuation ranges and selecting remote sensing images of the water storage tank at the times corresponding to these temperature fluctuation ranges as valid images; an image analysis module for dividing the water storage tank into several equal sub-regions horizontally and analyzing each sub-region in the valid images using a detection model to determine the pollution risk of each sub-region; and a joint verification module for determining the characteristic distribution of pollution risk in the sub-regions of the water storage tank based on the location of the water outlet, and verifying the pollution risk of each sub-region based on the characteristic distribution to obtain a verification judgment of the pollution risk of each sub-region.

Description

Technical Field

[0001] This invention relates to the field of data processing technology, and more specifically, to a system for determining the operating status of an energy storage system. Background Technology

[0002] In the operation of a reservoir-based hydroelectric energy storage system, the pollution risk of the reservoir directly affects the operating status of the hydroelectric power generation equipment. Pollution risks include contaminants such as sediment and particulate matter in the reservoir. Current technologies typically rely on manual monitoring and assessment of the pollution risk of the reservoir in reservoir-based energy storage systems, resulting in low monitoring efficiency and inaccurate assessment results. Summary of the Invention

[0003] This invention provides a system for determining the operating status of an energy storage system, thereby solving the technical problems mentioned in the background art.

[0004] This invention provides a system for determining the operating status of an energy storage system, comprising:

[0005] The data acquisition module is used to collect the temperature of the water storage tank at fixed time intervals, and to acquire remote sensing images of the water storage tank through monitoring equipment at fixed time intervals.

[0006] The data filtering module is used to analyze the temperature of the reservoir at multiple times, determine the temperature fluctuation range of the reservoir, and select the remote sensing image of the reservoir at the time corresponding to the temperature fluctuation range as the effective image.

[0007] The image analysis module is used to divide the reservoir into several equal sub-regions in the horizontal direction, and analyze each sub-region of the reservoir in the effective image through a preset detection model to determine the pollution risk of each sub-region.

[0008] The joint verification module is used to determine the characteristic distribution of pollution risk in sub-regions of the reservoir based on the location of the reservoir's outlet, and to verify the pollution risk of each sub-region based on the characteristic distribution, thereby obtaining the verification judgment of the pollution risk of each sub-region, as follows:

[0009] If the pollution risk assessment for each sub-region is successful, the operating status of the energy storage system built on the reservoir is determined based on the pollution risk of each sub-region.

[0010] If the pollution risk assessment for each sub-region fails, the detection model is updated in reverse based on the feature distribution.

[0011] Furthermore, the temperature of the reservoir at multiple points in time was analyzed, including:

[0012] The critical Rayleigh number is calculated based on the temperature of the reservoir at multiple moments, using the following formula:

[0013]

[0014] Among them, R a The critical Rayleigh number is represented by g, g represents the acceleration due to gravity, b represents the coefficient of thermal expansion of the water in the reservoir, and T represents the thermal expansion coefficient of the water in the reservoir. b T represents the temperature at the bottom of the reservoir. s The water surface temperature of the reservoir is represented by h, the height difference between the water surface and the bottom of the reservoir is represented by v, the kinematic viscosity of the water in the reservoir is represented by k, and the heat dissipation efficiency of the water in the reservoir is represented by k. w The wind speed, r, represents the wind speed at the surface of the reservoir. w This represents the wind speed weighting coefficient, t represents the fixed time interval, and r represents the wind speed weighting coefficient. e Weighting coefficients representing the rate of change of water surface temperature in a reservoir;

[0015] R a The time corresponding to >1700 is taken as the starting time of the temperature fluctuation segment, and R is set as follows: a The time corresponding to <1700 is taken as the end time of the temperature fluctuation segment.

[0016] Furthermore, a pre-defined detection model is used to analyze each sub-region of the reservoir in the valid image. The detection model includes an input layer, a segmentation layer, a detection layer, and an output layer.

[0017] The input layer is used to input valid images into the detection model;

[0018] The cutting layer is used to cut the effective image according to the sub-region division of the reservoir, resulting in several cut images;

[0019] The detection layer, used to identify each cut pattern, specifically includes:

[0020] Step 1, construct the energy functional E(M0),

[0021] Where E(M0) represents the energy functional, used to make a smoothness assumption on the uncontaminated matrix M0 to be solved, m represents the number of rows in the cut image, n represents the number of columns in the cut image, M0 represents the uncontaminated matrix, M0(i,j) represents the pixel value in the i-th row and j-th column of the uncontaminated matrix, M0(i+1,j) represents the pixel value in the (i+1)-th row and j-th column of the cut image, M0(i-1,j) represents the pixel value in the (i-1)-th row and j-th column of the cut image, M0(i,j+1) represents the pixel value in the i-th row and (j+1)-th column of the cut image, and M0(i,j-1) represents the pixel value in the i-th row and (j-1)-th column of the cut image.

[0022] Step 2: Solve for the uncontaminated matrix M0 using an optimization algorithm, with the constraint that the uncontaminated matrix M0 and the cutting map are the same size;

[0023] Step 3: For the uncontaminated matrix M0, the number of updates to the uncontaminated matrix M0 is equal to the cutting graph when it reaches 0 iterations;

[0024] Step 4, construct the Lagrange function: Where λ represents the Lagrange multiplier and C represents the sum of the number of pixels in the cut image;

[0025] Step 5: Update the uncontaminated matrix M0 to obtain the updated uncontaminated matrix M0. The update formula is as follows:

[0026]

[0027] Among them, (M0) t+1 (i,j) represents the pixel value in the i-th row and j-th column updated by the uncontaminated matrix M0, where (M0) t (i,j) represents the pixel value in the i-th row and j-th column of the uncontaminated matrix M0. Let represent the discrete Laplace operator, η represent the update iteration step size, and t represent the number of update iterations;

[0028] Step 6: Subtract the z-th cut image from the updated uncontaminated matrix M0 point by point to obtain the deviation matrix S. Compare the element value of any element in the deviation matrix S with the preset eigenvalue. If the element value is less than the preset eigenvalue, replace it with 0. If the element value is greater than the preset eigenvalue, replace it with 1 to obtain the replacement matrix.

[0029] Step 7: Perform connectivity checks on each element in the replacement matrix to obtain the number L of connected clusters in the replacement matrix;

[0030] Step 8: Calculate the contamination risk R of the z-th cut image. totle The calculation formula is as follows:

[0031] R totle =MLP(ω1×‖S‖) F +ω2×L); where, ‖S‖ F Let S denote the Frobenius norm of the bias matrix S, ω1 and ω2 denote the sovereign weight parameter and the secondary weight parameter, respectively, and MLP denotes a multilayer perceptron.

[0032] The output layer is used to output the contamination risk R of the z-th cut image. totle .

[0033] Furthermore, based on the location of the reservoir's outlet, the characteristic distribution of pollution risk in the reservoir's sub-areas is determined, including:

[0034] Obtain the center coordinates (X) of each sub-region z ,Y zThe coordinates of the outlet (X0, Y0) are also provided.

[0035] For each sub-region, the Euclidean distance to the outlet is calculated using the center coordinates;

[0036] The characteristic distribution formula is as follows:

[0037]

[0038] Where XT represents the intensity of the pollution risk characteristic distribution in the z-th sub-region, d z Let d represent the Euclidean distance from the center of the z-th sub-region to the outlet, p represent the risk intensity coefficient, R0 represent the maximum pollution risk of the sub-region, φ1 represent the exponential decay parameter, e represent the natural base, φ2 represent the linear decay weight, and d max This represents the maximum distance between any two points in the reservoir.

[0039] Based on the characteristic distribution intensity of each sub-region, the characteristic distribution of pollution risk in the sub-regions of the reservoir is obtained.

[0040] Furthermore, the pollution risk of each sub-region is verified based on the characteristic distribution, including:

[0041] The average contamination risk of each sub-region is determined by averaging the multiple contamination risks of each sub-region in the output of several valid images based on the detection model.

[0042] Based on the maximum and minimum values ​​of the average pollution risk in each sub-region, the average pollution risk of each sub-region is normalized to obtain the actual distribution intensity of pollution risk in each sub-region.

[0043] Calculate the sum of squared differences between the actual distribution intensity of pollution risk and the characteristic distribution intensity of pollution risk, and compare the sum of squared differences with a preset deviation threshold.

[0044] If the sum of squared differences is less than the preset deviation threshold, the pollution risk assessment for each sub-region is successful; otherwise, the pollution risk assessment for each sub-region fails.

[0045] Furthermore, if the pollution risk assessment for each sub-region is successful, the operational status of the water storage energy system is determined based on the pollution risk of each sub-region, including:

[0046] Take q consecutive pollution risks from multiple times in each sub-region as training samples, and use the (q+1)th pollution risk as the sample label to construct training data to train the prediction model.

[0047] The prediction model outputs the pollution risk of each sub-region at the next time step. If the pollution risk of at least one sub-region at the next time step is greater than the preset warning threshold, the sub-region is determined to be in an abnormal operating state.

[0048] Furthermore, if the pollution risk assessment for each sub-region fails, the detection model is updated in reverse based on the feature distribution, including:

[0049] The custom parameters in the detection model are updated in reverse using the gradient descent method, which reduces the difference between the normalized value of the average pollution risk of the detection model at multiple time points for each sub-region and the intensity of the pollution risk characteristic distribution of the sub-region; the custom parameters are those set manually.

[0050] Furthermore, the monitoring device is equipped with a memory, which changes the state of the storage medium through electrical signals and stores remote sensing images at multiple times.

[0051] The beneficial effects of this invention are as follows: by accurately analyzing the pollution risk of the reservoir sub-area, and combining remote sensing images, dynamic model optimization, and regional feature distribution, an automated and intelligent energy storage system operation status assessment mechanism has been established. This not only improves the system's operating efficiency but also allows for real-time response to environmental changes and dynamic model adjustment to ensure accurate judgment of pollution risks. This provides scientific decision support for the management of hydroelectric energy storage systems and significantly enhances the safety and reliability of hydroelectric energy storage systems. Attached Figure Description

[0052] Figure 1 This is a block diagram of an energy storage system operation status determination system according to the present invention. Detailed Implementation

[0053] The subject matter described herein will now be discussed with reference to exemplary embodiments. It should be understood that these embodiments are discussed only to enable those skilled in the art to better understand and implement the subject matter described herein, and changes may be made to the function and arrangement of the elements discussed without departing from the scope of this specification. Various processes or components may be omitted, substituted, or added as needed in the examples. Furthermore, features described in some examples may be combined in other examples.

[0054] like Figure 1 As shown, a system for determining the operating status of an energy storage system includes:

[0055] The data acquisition module is used to collect the temperature of the water storage tank at fixed time intervals, and to acquire remote sensing images of the water storage tank through monitoring equipment at fixed time intervals.

[0056] The data filtering module is used to analyze the temperature of the reservoir at multiple times, determine the temperature fluctuation range of the reservoir, and select the remote sensing image of the reservoir at the time corresponding to the temperature fluctuation range as the effective image.

[0057] The image analysis module is used to divide the reservoir into several equal sub-regions in the horizontal direction, and analyze each sub-region of the reservoir in the effective image through a preset detection model to determine the pollution risk of each sub-region.

[0058] The joint verification module is used to determine the characteristic distribution of pollution risk in sub-regions of the reservoir based on the location of the reservoir's outlet, and to verify the pollution risk of each sub-region based on the characteristic distribution, thereby obtaining the verification judgment of the pollution risk of each sub-region, as follows:

[0059] If the pollution risk assessment for each sub-region is successful, the operating status of the energy storage system built on the reservoir is determined based on the pollution risk of each sub-region.

[0060] If the pollution risk assessment for each sub-region fails, the detection model is updated in reverse based on the feature distribution.

[0061] In one embodiment of the present invention, the temperature of the water storage tank at multiple moments is analyzed, including:

[0062] The critical Rayleigh number is calculated based on the temperature of the reservoir at multiple moments, using the following formula:

[0063]

[0064] Among them, R a The critical Rayleigh number is represented by g, g represents the acceleration due to gravity, b represents the coefficient of thermal expansion of the water in the reservoir, and T represents the thermal expansion coefficient of the water in the reservoir. b T represents the temperature at the bottom of the reservoir. s The water surface temperature of the reservoir is represented by h, the height difference between the water surface and the bottom of the reservoir is represented by v, the kinematic viscosity of the water in the reservoir is represented by k, and the heat dissipation efficiency of the water in the reservoir is represented by k. w The wind speed, r, represents the wind speed at the surface of the reservoir. w This represents the wind speed weighting coefficient, t represents the fixed time interval, and r represents the wind speed weighting coefficient. e Weighting coefficients representing the rate of change of water surface temperature in a reservoir;

[0065] R a The time corresponding to >1700 is taken as the starting time of the temperature fluctuation segment, and R is set as follows: a The time corresponding to <1700 is taken as the end time of the temperature fluctuation segment.

[0066] Specifically, during normal periods, due to the lack of significant temperature differences within the water body, convection may not occur, and particles and sediment at the bottom will not be stirred up. In this situation, assessments of pollution risk may be inaccurate due to a lack of information on the movement of actual pollutants, leading to an underestimation of the pollution situation. However, when convection occurs, sediment and particles at the bottom of the reservoir are carried up by the water flow, polluting the upper water layers and potentially even entering the hydroelectric generator, threatening the safe operation of the equipment. Convection is usually accompanied by significant temperature changes (such as natural convection caused by temperature differences). Therefore, identifying the time periods of convection occurrence (temperature fluctuation periods) by analyzing temperature data is crucial; these periods are the critical moments when sediment and particles are stirred up and pollute the reservoir. By screening temperature fluctuation periods, the focus can be placed on the time periods when convection is significant, avoiding analysis interference from irrelevant time periods.

[0067] In one embodiment of the present invention, each sub-region of the reservoir in the effective image is analyzed using a preset detection model. The detection model includes an input layer, a segmentation layer, a detection layer, and an output layer.

[0068] The input layer is used to input valid images into the detection model;

[0069] The cutting layer is used to cut the effective image according to the sub-region division of the reservoir, resulting in several cut images;

[0070] The detection layer, used to identify each cut pattern, specifically includes:

[0071] Step 1, construct the energy functional E(M0),

[0072] Where E(M0) represents the energy functional, used to make a smoothness assumption on the uncontaminated matrix M0 to be solved, m represents the number of rows in the cut image, n represents the number of columns in the cut image, M0 represents the uncontaminated matrix, M0(i,j) represents the pixel value in the i-th row and j-th column of the uncontaminated matrix, M0(i+1,j) represents the pixel value in the (i+1)-th row and j-th column of the cut image, M0(i-1,j) represents the pixel value in the (i-1)-th row and j-th column of the cut image, M0(i,j+1) represents the pixel value in the i-th row and (j+1)-th column of the cut image, and M0(i,j-1) represents the pixel value in the i-th row and (j-1)-th column of the cut image.

[0073] Step 2: Solve for the uncontaminated matrix M0 using an optimization algorithm, with the constraint that the uncontaminated matrix M0 and the cutting map are the same size;

[0074] Step 3: For the uncontaminated matrix M0, the number of updates to the uncontaminated matrix M0 is equal to the cutting graph when it reaches 0 iterations;

[0075] Step 4, construct the Lagrange function: Where λ represents the Lagrange multiplier and C represents the sum of the number of pixels in the cut image;

[0076] Step 5: Update the uncontaminated matrix M0 to obtain the updated uncontaminated matrix M0. The update formula is as follows:

[0077]

[0078] Among them, (M0) t+1 (i,j) represents the pixel value in the i-th row and j-th column updated by the uncontaminated matrix M0, where (M0) t (i,j) represents the pixel value in the i-th row and j-th column of the uncontaminated matrix M0. Let represent the discrete Laplace operator, η represent the update iteration step size, and t represent the number of update iterations;

[0079] Step 6: Subtract the z-th cut image from the updated uncontaminated matrix M0 point by point to obtain the deviation matrix S. Compare the element value of any element in the deviation matrix S with the preset eigenvalue. If the element value is less than the preset eigenvalue, replace it with 0. If the element value is greater than the preset eigenvalue, replace it with 1 to obtain the replacement matrix.

[0080] Step 7: Perform connectivity checks on each element in the replacement matrix to obtain the number L of connected clusters in the replacement matrix;

[0081] Step 8: Calculate the contamination risk R of the z-th cut image. totle The calculation formula is as follows:

[0082] R totle =MLP(ω1×‖S‖) F +ω2×L); where, ‖S‖ F Let S denote the Frobenius norm of the bias matrix S, ω1 and ω2 denote the sovereign weight parameter and the secondary weight parameter, respectively, and MLP denotes a multilayer perceptron.

[0083] The output layer is used to output the contamination risk R of the z-th cut image. totle .

[0084] Specifically, by dividing the reservoir into multiple sub-regions and independently assessing the pollution risk of each region, the overall analysis's granularity and accuracy are improved. A pollution-free matrix is ​​constructed using energy functional theory, and then the pollution risk is calculated based on the deviation matrix, ensuring the mathematical rigor of the assessment process. The detection model consists of an input layer, a segmentation layer, a detection layer, and an output layer, with the functions of each layer as follows:

[0085] Input layer: Receives the filtered valid images as input data. These images have been filtered to identify the remote sensing images corresponding to key moments through analysis of temperature fluctuations.

[0086] Features of the input image may include the distribution characteristics of pollution in the water body (e.g., sediment diffusion).

[0087] Segmentation Layer: The effective image is segmented according to the sub-regions of the reservoir, generating segmentation maps of multiple sub-regions. Each segmentation map corresponds to a sub-region of the reservoir, facilitating subsequent individual analysis of each region and quantification of pollution risk.

[0088] Detection layer: For each cut map, the risk of contamination is calculated. Multiple mathematical tools, such as energy functionals, Lagrangian functions and discrete Laplace operators, are used to perform in-depth analysis of the characteristics of contamination in sub-regions.

[0089] Output layer: Outputs the pollution risk value of each sub-region, which serves as the basis for subsequent pollution risk verification and system operation status assessment.

[0090] In one embodiment of the present invention, determining the characteristic distribution of pollution risk in a sub-region of the water storage tank based on the location of the water outlet includes:

[0091] Obtain the center coordinates (X) of each sub-region z ,Y z The coordinates of the outlet (X0, Y0) are also provided.

[0092] For each sub-region, the Euclidean distance to the outlet is calculated using the center coordinates;

[0093] The characteristic distribution formula is as follows:

[0094]

[0095] Where XT represents the intensity of the pollution risk characteristic distribution in the z-th sub-region, d z Let d represent the Euclidean distance from the center of the z-th sub-region to the outlet, p represent the risk intensity coefficient, R0 represent the maximum pollution risk of the sub-region, φ1 represent the exponential decay parameter, e represent the natural base, φ2 represent the linear decay weight, and d max This represents the maximum distance between any two points in the reservoir.

[0096] Based on the characteristic distribution intensity of each sub-region, the characteristic distribution of pollution risk in the sub-regions of the reservoir is obtained.

[0097] Specifically, by calculating the distance from each sub-region to the outlet and combining it with a comprehensive pollution risk attenuation model, the characteristic distribution intensity of pollution risk in the reservoir is determined. This method not only considers the influence of spatial distance but also incorporates inverse proportional, exponential, and linear attenuation laws, providing a scientific basis for subsequent risk verification and system optimization. This distribution calculation method has the advantages of accuracy, efficiency, and strong adaptability.

[0098] In one embodiment of the present invention, verifying the pollution risk of each sub-region based on the characteristic distribution includes:

[0099] The average contamination risk of each sub-region is determined by averaging the multiple contamination risks of each sub-region in the output of several valid images based on the detection model.

[0100] Based on the maximum and minimum values ​​of the average pollution risk in each sub-region, the average pollution risk of each sub-region is normalized to obtain the actual distribution intensity of pollution risk in each sub-region.

[0101] Calculate the sum of squared differences between the actual distribution intensity of pollution risk and the characteristic distribution intensity of pollution risk, and compare the sum of squared differences with a preset deviation threshold.

[0102] If the sum of squared differences is less than the preset deviation threshold, the pollution risk assessment for each sub-region is successful; otherwise, the pollution risk assessment for each sub-region fails.

[0103] Specifically, the evaluation results of the detection model are quantitatively verified by calculating the average value, normalization, and sum of squared differences of the pollution risk. By setting a deviation threshold, the reliability and accuracy of the detection model can be effectively judged, thereby ensuring the accuracy of the pollution risk assessment results for each sub-region.

[0104] In one embodiment of the present invention, if the pollution risk of each sub-region is successfully determined, the operating status of the water storage energy system is determined based on the pollution risk of each sub-region, including:

[0105] Take q consecutive pollution risks from multiple times in each sub-region as training samples, and use the (q+1)th pollution risk as the sample label to construct training data to train the prediction model.

[0106] The prediction model outputs the pollution risk of each sub-region at the next time step. If the pollution risk of at least one sub-region at the next time step is greater than the preset warning threshold, the sub-region is determined to be in an abnormal operating state.

[0107] Specifically, the pollution risk of each sub-region is predicted by training a time-series prediction model and compared with warning thresholds to identify potential operational anomalies in advance. Its advantages include: utilizing historical data for trend prediction, enhancing the system's foresight and early warning capabilities; independent analysis of each sub-region, improving the precision of pollution risk assessment; and automated anomaly detection, reducing human intervention and enhancing the system's intelligence level.

[0108] In one embodiment of the present invention, if the pollution risk determination for each sub-region fails, the detection model is updated in reverse according to the feature distribution, including:

[0109] The custom parameters in the detection model are updated in reverse using the gradient descent method, which reduces the difference between the normalized value of the average pollution risk of the detection model at multiple time points for each sub-region and the intensity of the pollution risk characteristic distribution of the sub-region; the custom parameters are those set manually.

[0110] In one embodiment of the present invention, the monitoring device is equipped with a memory, which stores remote sensing images at multiple times by changing the state of the memory storage medium through electrical signals.

[0111] Specifically, by updating the custom parameters in the detection model in reverse, the model's assessment of pollution risk for each sub-region becomes more accurate. The custom parameters are optimized using gradient descent to dynamically reduce the difference between the actual pollution risk distribution and the characteristic distribution, enabling intelligent adjustment of the detection model.

[0112] The embodiments of this example have been described above. However, this example is not limited to the specific implementation methods described above. The specific implementation methods described above are merely illustrative and not restrictive. Those skilled in the art can make many other forms based on the guidance of this example, and all of them are within the protection scope of this example.

Claims

1. A system for determining the operating status of an energy storage system, characterized in that, include: The data acquisition module is used to collect the temperature of the water storage tank at fixed time intervals, and to acquire remote sensing images of the water storage tank through monitoring equipment at fixed time intervals. The data filtering module is used to analyze the temperature of the reservoir at multiple times, determine the temperature fluctuation range of the reservoir, and select the remote sensing image of the reservoir at the time corresponding to the temperature fluctuation range as the effective image. The image analysis module is used to divide the reservoir into several equal sub-regions in the horizontal direction, and analyze each sub-region of the reservoir in the effective image through a preset detection model to determine the pollution risk of each sub-region. The joint verification module is used to determine the characteristic distribution of pollution risk in sub-regions of the reservoir based on the location of the reservoir's outlet, and to verify the pollution risk of each sub-region based on the characteristic distribution, thereby obtaining the verification judgment of the pollution risk of each sub-region, as follows: If the pollution risk assessment for each sub-region is successful, the operating status of the energy storage system built on the reservoir is determined based on the pollution risk of each sub-region. If the pollution risk assessment for each sub-region fails, the detection model is updated in reverse based on the feature distribution. The temperature of the reservoir at multiple points in time was analyzed, including: The critical Rayleigh number is calculated based on the temperature of the reservoir at multiple moments, using the following formula: ； in, This represents the critical Rayleigh number. Represents gravitational acceleration. This represents the coefficient of thermal expansion of the water in the reservoir. This indicates the temperature at the bottom of the reservoir. This indicates the temperature of the water surface in the reservoir. This indicates the height difference between the water surface and the bottom of the reservoir. This indicates the kinematic viscosity of the water in the reservoir. This indicates the heat dissipation efficiency of the water in the reservoir. The wind speed at the surface of the reservoir is indicated. This represents the wind speed weighting coefficient. Indicates a fixed time interval. Weighting coefficients representing the rate of change of water surface temperature in a reservoir; Will The time corresponding to >1700 is taken as the starting time of the temperature fluctuation segment. The time corresponding to <1700 is taken as the end time of the temperature fluctuation segment.

2. The system for determining the operating status of an energy storage system according to claim 1, characterized in that, The detection model analyzes each sub-region of the reservoir in the valid image using a pre-defined detection model, which includes an input layer, a segmentation layer, a detection layer, and an output layer. The input layer is used to input valid images into the detection model; The cutting layer is used to cut the effective image according to the sub-region division of the reservoir, resulting in several cut images; The detection layer, used to identify each cut pattern, specifically includes: Step 1, Construct the energy functional , ; in, Represents the energy functional used to solve for the uncontaminated matrix. Make a smoothness assumption. Indicates the number of rows in the cut graph. Indicates the number of columns in the cut diagram. Represents a pollution-free matrix. Represents the first contamination in the uncontaminated matrix. Line 1 Column pixel values, Indicates the first cutting diagram Line 1 Column pixel values, Indicates the first cutting diagram Line 1 Column pixel values, Indicates the first cutting diagram Line 1 Column pixel values, Indicates the first cutting diagram Line 1 The pixel values ​​of the column; Step 2, for the uncontaminated matrix The solution is obtained through an optimization algorithm, with the constraint being a pollution-free matrix. Same size as the cutout image; Step 3, for the uncontaminated matrix Pollution-free matrix When the number of update iterations is 0, it equals the cut map; Step 4, construct the Lagrange function: ;in, Represents the Lagrange multipliers. This represents the sum of the number of pixels in the cut image; Step 5, for the uncontaminated matrix Perform an update and solve the problem to obtain the updated uncontaminated matrix. The updated formula is as follows: ； in, Represents a pollution-free matrix Update the pixel value in the i-th row and j-th column. Represents a pollution-free matrix The pixel value in the i-th row and j-th column. Represents the discrete Laplace operator. Indicates the update iteration step size. Indicates the number of update iterations; Step 6, the first Each cut image and the updated pollution-free matrix By subtracting each point, we obtain the deviation matrix. and the deviation matrix The element value of any element is compared with a preset feature value. If the element value is less than the preset feature value, it is replaced with 0. If the element value is greater than the preset feature value, it is replaced with 1, thus obtaining a replacement matrix. Step 7: Perform connectivity checks on each element in the replacement matrix to obtain the number of connected clusters in the replacement matrix. ; Step 8, calculate the first... Risk of contamination of individual cutout images The calculation formula is as follows: ;in, Represents the deviation matrix The Frobenius norm, and These represent the sovereign weight parameter and the secondary weight parameter, respectively. This represents a multilayer perceptron; Output layer, used to output the first... Risk of contamination of individual cutout images .

3. The energy storage system operation status determination system according to claim 2, characterized in that, Based on the location of the reservoir's outlet, the characteristic distribution of pollution risk in the reservoir's sub-areas is determined, including: Obtain the center coordinates of each sub-region and the coordinates of the outlet ; For each sub-region, the Euclidean distance to the outlet is calculated using the center coordinates; The characteristic distribution formula is as follows: ； in, Indicates the first Distribution intensity of pollution risk characteristics in individual sub-regions Indicates the first The European-style distance from the center of each sub-area to the outlet. Indicates the risk intensity coefficient. Indicates the maximum pollution risk of the sub-region. Represents the exponential decay parameter. Represents the natural base. Indicates linearly decaying weights. This represents the maximum distance between any two points in the reservoir. Based on the characteristic distribution intensity of each sub-region, the characteristic distribution of pollution risk in the sub-regions of the reservoir is obtained.

4. The energy storage system operation status determination system according to claim 3, characterized in that, The pollution risk of each sub-region is verified based on the characteristic distribution, including: The average contamination risk of each sub-region is determined by averaging the multiple contamination risks of each sub-region in the output of several valid images based on the detection model. Based on the maximum and minimum values ​​of the average pollution risk in each sub-region, the average pollution risk of each sub-region is normalized to obtain the actual distribution intensity of pollution risk in each sub-region. Calculate the sum of squared differences between the actual distribution intensity of pollution risk and the characteristic distribution intensity of pollution risk, and compare the sum of squared differences with a preset deviation threshold. If the sum of squared differences is less than the preset deviation threshold, the pollution risk assessment for each sub-region is successful; otherwise, the pollution risk assessment for each sub-region fails.

5. The energy storage system operation status determination system according to claim 4, characterized in that, If the pollution risk assessment for each sub-region is successful, the operational status of the water storage energy system will be determined based on the pollution risk of each sub-region, including: Take q consecutive pollution risks from multiple times in each sub-region as training samples, and use the (q+1)th pollution risk as the sample label to construct training data to train the prediction model. The prediction model outputs the pollution risk of each sub-region at the next time step. If the pollution risk of at least one sub-region at the next time step is greater than the preset warning threshold, the sub-region is determined to be in an abnormal operating state.

6. The energy storage system operation status determination system according to claim 5, characterized in that, If the pollution risk assessment for each sub-region fails, the detection model will be updated in reverse based on the feature distribution, including: The custom parameters in the detection model are updated in reverse using the gradient descent method, which reduces the difference between the normalized value of the average pollution risk of the detection model at multiple time points for each sub-region and the intensity of the pollution risk characteristic distribution of the sub-region; the custom parameters are those set manually.

7. The energy storage system operation status determination system according to claim 1, characterized in that, The monitoring device is equipped with a memory, which changes the state of the storage medium through electrical signals and stores remote sensing images at multiple times.

Images