Load monitoring method and system for photovoltaic energy storage battery

Abstract

The invention relates to the field of photovoltaic energy storage batteries, in particular to a load monitoring method and system for a photovoltaic energy storage battery, and the method comprises the steps: collecting multi-dimensional parameters of the photovoltaic energy storage battery under a set load in real time, and integrating the multi-dimensional parameters at the same collection moment into a comprehensive data point; and for any one comprehensive data point, clustering the comprehensive data points according to the calculated neighborhood range to obtain a clustering result, and carrying out load state evaluation on each comprehensive data point based on the clustering result. Through deep fusion of spatio-temporal conjoint analysis, dynamic parameter self-adaption and intelligent clustering technologies, a closed-loop monitoring system from data acquisition to fault early warning is constructed, and a load management solution with high reliability, low false alarm and high adaptability is provided for a photovoltaic energy storage system.

Description

Technical Field

[0001] The present invention relates to the field of photovoltaic energy storage batteries. More specifically, the present invention relates to a load monitoring method and system for photovoltaic energy storage batteries. Background Art

[0002] A photovoltaic energy storage battery system is a device that converts solar energy into electrical energy and stores it. It can provide power support during peak electricity demand, playing a role in regulating the grid load and improving energy utilization efficiency. In practical applications, the stability and safety of a photovoltaic energy storage battery system may be affected by various factors. The most important one is the load change of the photovoltaic energy storage battery. Frequent load changes will accelerate the aging process of the battery, affecting the charge and discharge efficiency and lifespan of the battery.

[0003] To address the impact of load changes on the stability and safety of a photovoltaic energy storage battery system, a smart monitoring system is used to real-time monitor parameters such as the battery state, voltage, current, and temperature, and to promptly detect and handle abnormal situations.

[0004] Existing technologies such as DBSCAN is a density-based clustering algorithm that identifies different clusters (i.e., sets of similar data points) by analyzing the distribution density of data points and can identify noise points (i.e., isolated points that do not belong to any cluster). In the context of a photovoltaic energy storage battery system, the DBSCAN algorithm can be used to perform clustering analysis on the load parameters of the battery (such as voltage, current, power, etc.), and can identify abnormal data points that do not belong to the normal operating mode, thereby promptly detecting potential problems in the system, such as load overload. However, when using the DBSCAN algorithm to cluster the load parameters of a photovoltaic energy storage battery system, due to the limitation of the fixed neighborhood radius, it is difficult to adapt to complex data distributions and dynamic changes, resulting in an unsatisfactory clustering result, thereby affecting the accurate monitoring of the load state of the photovoltaic energy storage battery. Summary of the Invention

[0005] To solve the technical problem that the clustering result after clustering the load data of a photovoltaic energy storage battery is unsatisfactory, which in turn affects the accurate monitoring of the load state of the photovoltaic energy storage battery, the present invention provides solutions in the following aspects.

[0006] In a first aspect, a load monitoring method for a photovoltaic energy storage battery includes: Real-time collecting multi-dimensional parameters of the photovoltaic energy storage battery under a set load, and integrating the multi-dimensional parameters at the same collection moment into a comprehensive data point; for any comprehensive data point, clustering the comprehensive data point according to the calculated neighborhood range to obtain a clustering result, and evaluating the load state of each comprehensive data point based on the clustering result; Wherein, the process of obtaining the neighborhood range is: Calculate the dimensional importance of each dimensional parameter, and calculate the weighted Euclidean distance between each comprehensive data point and all other comprehensive data points based on the dimensional importance; Calculate the isolation degree of each comprehensive data point in time, and combine the weighted Euclidean distance to obtain the abnormality of the comprehensive data point; Adjust the preset initial neighborhood range of the comprehensive data point according to the abnormality, and then obtain the final neighborhood range.

[0007] By calculating the dimensional importance of each dimensional parameter, the present invention can assign different weights to different parameters according to the actual working conditions, so as to calculate the weighted Euclidean distance more reasonably. Combining the isolation degree in time, the abnormality of the comprehensive data point can be evaluated more accurately. Dynamically adjusting the neighborhood range according to the abnormality enables the clustering process to more flexibly adapt to the distribution characteristics of the data, avoiding misjudgment or missed judgment caused by a fixed neighborhood range; Through the clustering results obtained by the above method, abnormal data points can be effectively identified, so as to timely discover possible abnormal states in the operation process of the photovoltaic energy storage battery, such as overcharging, over-discharging, short circuit, thermal runaway, etc.

[0008] Preferably, the clustering is the DBSCAN clustering algorithm.

[0009] Preferably, the load status evaluation of each comprehensive data point based on the clustering results includes: Based on the clustering results, all comprehensive data points are divided into core points, boundary points and isolated points, calculate the proportion of isolated points, and when the proportion of isolated points is greater than the preset abnormality threshold, an abnormality warning is issued.

[0010] Dividing the comprehensive data points into core points, boundary points and isolated points through the clustering results can more clearly identify the position and nature of the data points in the clustering. Core points usually represent normal and stable operating states, boundary points may be at the boundary of normal states, while isolated points are more likely to be manifestations of abnormal states; calculating the proportion of isolated points provides a quantitative index for abnormal states. When the proportion of isolated points exceeds the preset abnormality threshold, it can be clearly judged that the system operating state has an abnormality.

[0011] Preferably, the process of obtaining the dimensional importance includes: Based on a single dimensional parameter among all the collected dimensional parameters, calculate the standard deviation of the dimensional parameter, and obtain the maximum value and the minimum value of the dimensional parameter, and normalize the ratio of the standard deviation to the difference between the maximum value and the minimum value to obtain the dimensional importance.

[0012] In the load status monitoring of photovoltaic energy storage batteries, different dimensional parameters (such as voltage, current, temperature, etc.) may have different degrees of influence on the battery status. By calculating the dimensional importance through the above method, multi-dimensional data can be flexibly processed, avoiding misjudgment or missed judgment caused by fixed weight allocation; the standard deviation is an important indicator to measure the degree of data dispersion, reflecting the fluctuation of the dimensional parameter during the acquisition process. The difference between the maximum value and the minimum value represents the value range of the parameter. Normalizing the ratio of the standard deviation to the difference can comprehensively consider the fluctuation amplitude and value range of the parameter, so as to more accurately reflect the importance of the parameter in the overall data.

[0013] Preferably, the weighted Euclidean distance satisfies the relational expression: ; where represents the weighted Euclidean distance between the th comprehensive data point and the th comprehensive data point, is the value of the th dimensional parameter in the th comprehensive data point, is the value of the th dimensional parameter in the th comprehensive data point, is the importance of the th dimensional parameter, is the total number of acquired dimensional parameters.

[0014] By introducing the weighted Euclidean distance, when calculating the distance, important dimensional parameters contribute more to the distance, while unimportant dimensional parameters contribute less to the distance. This can more accurately reflect the actual differences between data points.

[0015] Preferably, calculating the isolation degree of each comprehensive data point in time includes: Select any one comprehensive data point as the target data set, obtain the acquisition times corresponding to the first K comprehensive data points in the target sequence of the target data set, calculate the absolute value of the difference between the target data set and the sampling times of the first K comprehensive data points in its target sequence as the first parameter, and calculate the standard deviation of the difference between the target data set and the sampling times of the first K comprehensive data points in its target sequence as the second parameter; Calculate the mean of the sum of the ratios of all the first parameters to the second parameters and normalize it, and use the normalized result as the isolation degree of the target data set in time.

[0016] The first parameter (absolute time difference) measures the absolute time distance between the target data point and the previous K points. If the time interval between the target point and the previous K points is significantly greater than the average level, it indicates that there may be a time deviation; the second parameter (standard deviation of time difference) reflects the degree of dispersion of the time distribution of the previous K points themselves. If the time distribution of the previous K points is concentrated (small standard deviation), the time difference of the target point is more likely to be amplified as an anomaly.

[0017] Preferably, the abnormality satisfies the relational expression: ; where is the abnormality of the th target data set, is the degree of isolation in time of the th target data set, is the weighted Euclidean distance between the th target data set and the th comprehensive data point in its target sequence, is the standard deviation of the weighted Euclidean distances between the th target data set and all comprehensive data points in its target sequence, represents the standard normalization function.

[0018] Preferably, adjusting the preset initial neighborhood range of the comprehensive data point according to the abnormality includes: Preset a hyperparameter. When the difference between the hyperparameter and the abnormality is less than 0, narrow the initial neighborhood range; conversely, when the difference between the hyperparameter and the abnormality is greater than 0, expand the initial neighborhood range.

[0019] Preferably, the process of obtaining the dimension importance includes: Based on a single dimension parameter among all the collected dimension parameters, calculate the standard deviation of this dimension parameter and obtain the quartiles of this dimension parameter, and normalize the ratio of the standard deviation to the quartiles to obtain the dimension importance.

[0020] In a second aspect, a load monitoring system for a photovoltaic energy storage battery includes: a processor and a memory. The memory stores computer program instructions, and when the computer program instructions are executed by the processor, any one of the load monitoring methods for a photovoltaic energy storage battery described above is implemented.

[0021] The beneficial effects of the present invention are: The present invention first collects multi-dimensional parameters in real time and integrates them into comprehensive data points, which can comprehensively reflect the operating state of the photovoltaic energy storage battery under a set load, avoiding the limitations of single-parameter monitoring. Then, it dynamically adjusts the neighborhood range according to the abnormality of the comprehensive data points, enabling better adaptation to different load conditions and operating environments, making the clustering algorithm adaptable to the data distribution under different load conditions, improving the clustering accuracy, reducing misclassification, and finally triggering an alarm based on the proportion of outliers, quickly locating abnormal load states (such as overload and short circuit), shortening the fault response time, and ensuring system safety. Description of the Drawings

[0022] Figure 1 It is a flowchart of the method from step S1 to step S2 in a load monitoring method for a photovoltaic energy storage battery according to an embodiment of the present invention.

[0023] Figure 2 It is a flowchart of the method for dynamically adjusting the neighborhood range according to an embodiment of the present invention. Detailed Embodiments

[0024] Next, the technical solutions in the embodiments of the present invention will be clearly and completely described in conjunction with the accompanying drawings in the embodiments of the present invention. Obviously, the described embodiments are some, but not all, of the embodiments of the present invention.

[0025] Refer to Figure 1 , a load monitoring method for a photovoltaic energy storage battery includes steps S1 - S2, specifically as follows: S1: Collect multi-dimensional parameters of the photovoltaic energy storage battery under a set load in real time, and integrate the multi-dimensional parameters at the same acquisition moment into a comprehensive data point.

[0026] Different loads have different requirements for parameters such as the charge and discharge current and voltage of the photovoltaic energy storage battery. By collecting multi-dimensional parameters in real time, dynamic monitoring of the photovoltaic energy storage battery can be achieved. Through real-time monitoring, abnormal situations of the battery, such as overcharging, over-discharging, and overheating, can be detected in a timely manner.

[0027] In one embodiment, by using an electronic load device to set load parameters (manually operated by the implementer in the embodiment of the present invention), and then selecting appropriate voltage sensors, current sensors, and temperature sensors, multi-dimensional parameters of the photovoltaic energy storage battery under a set load are collected in real time (in the embodiment of the present invention, three parameters, namely the charge and discharge current, voltage, and temperature of the photovoltaic energy storage battery, are selected for real-time collection). During the collection process, the multi-dimensional parameter data at the same acquisition moment are integrated into a comprehensive data point.

[0028] It should be noted that the sampling frequency is set to 20Hz. The collected data is preprocessed, such as filtering and denoising, to improve the data quality.

[0029] S2: For any comprehensive data point, cluster the comprehensive data point according to the calculated neighborhood range to obtain a clustering result, and evaluate the load status of each comprehensive data point based on the clustering result.

[0030] In the traditional DBSCAN algorithm, a fixed neighborhood range (radius) is usually used to divide clustering clusters. However, in high-dimensional data, the characteristics of different dimensions have different effects on clustering, and the distribution of data points may be very complex. It is difficult for a fixed neighborhood range to adapt to this complexity, and abnormal data points are relatively isolated in space and time. If a fixed neighborhood range is used, it may also cause abnormal points to be incorrectly classified into normal clustering clusters, thereby reducing the accuracy of anomaly detection.

[0031] In order to better adapt to the dynamic characteristics and complexity of photovoltaic energy storage battery load data, enhance the accuracy of anomaly detection, improve the clustering effect, and avoid the limitations of fixed parameters, in the embodiments of the present invention, by dynamically adjusting the neighborhood range, it is possible to dynamically determine whether a data point is an isolated point according to the abnormality of the data point, so as to more accurately identify the abnormal state and ensure the safe operation of the photovoltaic energy storage system.

[0032] As Figure 2 shown, the above process of dynamically adjusting the neighborhood range includes step S20-step S22.

[0033] S20: Calculate the dimension importance of each dimension parameter, and calculate the weighted Euclidean distance between each comprehensive data point and all other comprehensive data points based on the dimension importance.

[0034] In the photovoltaic energy storage battery load data, different dimensions (such as voltage, current, temperature) have different contributions to the load status, and the fluctuation ranges of data in different dimensions are different. If the same weight is used for each dimension, important features will be diluted by other dimensions, affecting the accuracy of subsequent clustering and anomaly detection.

[0035] In one embodiment, for a single dimension parameter among all the dimension parameters collected, calculate the standard deviation of the dimension parameter, and obtain the maximum value and the minimum value of the dimension parameter.

[0036] Calculate the importance of the dimension parameter based on the standard deviation, maximum value, and minimum value of the dimension parameter, and the importance satisfies the relational expression:

[0037] In the formula, is the importance of the th dimension parameter, is the standard deviation of the th dimension parameter, is the The maximum value of the dimensional parameter, is the minimum value of the dimensional parameter.

[0038] Among them, As the standard deviation, it reflects the fluctuation degree of the dimensional parameter. The greater the fluctuation degree, the more unstable the dimensional parameter; represents the size of the distribution range of the dimensional data. The larger the distribution range, the larger the standard deviation allowed for the dimensional data; represents the normalization process.

[0039] According to the above-mentioned calculation process of the importance of the dimensional parameter, the importance corresponding to all dimensional parameters can be calculated in the same way.

[0040] Generally speaking, by using the size of the distribution range of the dimensional parameter to correct the fluctuation degree of the dimensional parameter, and taking the corrected result as the importance of the dimensional parameter, the influence of irrelevant dimensions is reduced.

[0041] Considering the influence of extreme values, in another embodiment, the importance of the dimensional parameter is calculated using quartiles, that is, the relational expression is satisfied as:

[0042] In the formula, is the importance of the dimensional parameter, is the standard deviation of the dimensional parameter, is the quartile of all the dimensional parameters collected,

[0043] In addition, the traditional DBSCAN algorithm uses the standard Euclidean distance to calculate the similarity between two comprehensive data points. However, in high-dimensional data, the standard Euclidean distance cannot distinguish the relative importance of each dimension, resulting in a decline in the clustering effect. Therefore, by introducing the dimension importance into the calculation of the Euclidean distance, the dimension with a greater degree of importance contributes more to the distance, better adapting to the characteristics of the photovoltaic energy storage battery load data and improving the accuracy of clustering.

[0044] In one embodiment, the weighted Euclidean distance between two comprehensive data points is calculated by combining the traditional Euclidean distance formula and the dimension importance, that is, the relational expression is satisfied as:

[0045] In the formula, represents the weighted Euclidean distance between the th comprehensive data point and the is the value of the th dimensional parameter in the th comprehensive data point. is the value of the th dimensional parameter in the th comprehensive data point. is the importance of the th dimensional parameter. is the total number of the collected dimensional parameters. In the embodiment of the present invention, it takes the value of 3, that is, it includes the charge and discharge current, voltage and temperature of the photovoltaic energy storage battery collected in S1 above.

[0046] Further, according to the above operations, the weighted Euclidean distance between any one comprehensive data point and all other comprehensive data points can be obtained. Then, all the obtained weighted Euclidean distances are sorted in ascending order, and then a sequence of comprehensive data points sorted by the weighted Euclidean distance is obtained as the target sequence of the selected comprehensive data points.

[0047] S21: Calculate the isolation degree of each comprehensive data point in time, and combine the weighted Euclidean distance to obtain the abnormality of the comprehensive data point.

[0048] The traditional DBSCAN algorithm performs clustering by considering the Euclidean distance between comprehensive data points. However, in the load monitoring of photovoltaic energy storage batteries, the load parameters will have local changes with the user's usage behavior. For example, during the peak electricity consumption period, when the load is large and all parameters are close to the limit values, it is a normal state, while when it suddenly changes to the limit value during the low electricity consumption period, it is probably an abnormal state. Therefore, it is necessary to consider the time series factor of the data points to calculate the abnormality of multi-dimensional data points.

[0049] In one embodiment, select any one comprehensive data point as the target data set, obtain the acquisition times corresponding to the first K (in the embodiment of the present invention, K takes the value of 20) comprehensive data points in the target sequence of the target data set, and calculate the absolute value of the difference between the target data set and the sampling times of the first K comprehensive data points in its target sequence respectively (for the convenience of description, the sum of the calculated absolute values of the differences is denoted as the first parameter), and calculate the standard deviation of the differences between the target data set and the sampling times of the first K comprehensive data points in its target sequence (for the convenience of description, the calculated standard deviation of the differences is denoted as the second parameter).

[0050] Further calculate the mean value of the sum of the ratios of all the first parameters to the second parameters and normalize it, and use the normalized result as the isolation degree of the target data set in time.

[0051] The greater the degree of isolation, it indicates that for other comprehensive data points that are relatively similar to the target data set, the corresponding sampling times are farther away from the sampling time corresponding to the target data set. In other words, the target data set is relatively isolated in time, and the data sets similar to it are distributed far apart in time.

[0052] Furthermore, combine the degree of isolation of the target data set in time with the above-mentioned weighted Euclidean distance, that is, considering both the time sequence factor and the space factor, and then calculate the abnormality of the comprehensive data point.

[0053] Exemplarily, a calculation formula for combining the degree of isolation of the target data set in time with the weighted Euclidean distance to calculate the abnormality of the target data set is given:

[0054] In the formula, is the abnormality of the th target data set, is the degree of isolation of the th target data set in time, is the weighted Euclidean distance between the th target data set and the th comprehensive data point in its target sequence, is the standard deviation of the weighted Euclidean distances between the th target data set and all the comprehensive data points in its target sequence, represents the standard normalization function.

[0055] The weighted Euclidean distance takes into account the feature differences of the data and can identify abnormal data sets that are quite different from other comprehensive data points in the feature space. After combining with the time isolation degree, it can more comprehensively evaluate the abnormality of the data set and avoid judging from a single dimension (time or space) only.

[0056] Then, according to the above operations, the abnormality of all the comprehensive data points can be calculated.

[0057] S22: Adjust the preset initial neighborhood range of the comprehensive data point according to the abnormality, and then obtain the final neighborhood range.

[0058] In cluster analysis, data points are usually divided into different clusters according to their neighborhood ranges (i.e., the distances from surrounding data points). However, some data points may be isolated from other data points both in space and time, and these data points are called "isolated points" or "abnormal points". In order to better identify these isolated points, it is necessary to dynamically adjust the neighborhood range according to the abnormality of the data.

[0059] In one embodiment, the neighborhood range is dynamically adjusted according to the abnormality of all the comprehensive data points calculated in S21 above. For comprehensive data points with high abnormality, a smaller neighborhood range is used to ensure that they are more easily identified as outliers; for comprehensive data points with low abnormality, a larger neighborhood range is used to better cluster with other comprehensive data points.

[0060] Specifically, first, an initial neighborhood range used by a comprehensive data point in cluster analysis is preset (that is, how many neighboring comprehensive data points are considered in cluster analysis, and the value in the embodiment of the present invention is 20).

[0061] Next, a calculation formula for dynamically adjusting the initial neighborhood range of comprehensive data points based on abnormality is given:

[0062] In the formula, is the neighborhood range of the th comprehensive data point, is the preset initial neighborhood range, is the th comprehensive data point's abnormality, represents the hyperbolic tangent function; is a preset hyperparameter used to measure the standard of abnormality, and this hyperparameter can be determined by historical data statistics or cross-validation. In the embodiment of the present invention, the value of this hyperparameter is 0.5.

[0063] Among them, when < 0, at this time is a negative value, that is, the initial neighborhood range is reduced. After reducing the neighborhood, the number of neighboring points around the th comprehensive data point decreases, and it is more likely to be determined as an outlier; conversely, when > 0, at this time is a positive value, that is, the initial neighborhood range is expanded. After expanding the neighborhood, the th comprehensive data point is more likely to integrate into the surrounding clusters, reducing the risk of being misjudged as an outlier.

[0064] In addition, in the formula, through the coefficient, it is ensured that the neighborhood range fluctuates between to avoid excessive fluctuations in the neighborhood range due to abnormality, while retaining sufficient adjustment space to adapt to different scenarios.

[0065] According to the adjustment operation of the initial neighborhood range of the th comprehensive data point above, the initial neighborhood ranges of all other comprehensive data points can be adjusted in the same way, and then the corresponding final neighborhood ranges are obtained.

[0066] All the comprehensive data points are clustered using the final neighborhood range of all the comprehensive data points calculated above, and all the comprehensive data points are divided into core points, edge points and isolated points.

[0067] The above core points, edge points and isolated points are the existing technology of DBSCAN clustering algorithm, which will not be described in detail here.

[0068] The proportion of isolated points is further calculated, that is, the ratio of the number of isolated points to the sum of the number of core points, edge points and isolated points.

[0069] When the proportion of isolated points exceeds a preset abnormal threshold (set to 0.1 in the embodiment of the present invention), it is considered that there are many abnormal points or noise points in the collected data. In this case, an early warning is issued to the staff to indicate that the load status of the photovoltaic energy storage battery is abnormal.

[0070] The system includes a processor and a memory, wherein the memory stores computer program instructions, and when the computer program instructions are executed by the processor, a load monitoring method for a photovoltaic energy storage battery according to the first aspect of the present invention is implemented.

[0071] The system also includes other components familiar to those skilled in the art, such as a communication bus and a communication interface. The configuration and functions of these components are known in the art and will not be described in detail here.

[0072] It should be noted that, for those skilled in the art, several modifications and improvements can be made without departing from the concept of the present invention, and these modifications and improvements all belong to the protection scope of the present invention. Therefore, the protection scope of the patent of the present invention shall be subject to the attached claims.

Claims

1. A load monitoring method for a photovoltaic energy storage battery, characterized in that Including: Collecting multi-dimensional parameters of a photovoltaic energy storage battery under a set load in real time, and integrating the multi-dimensional parameters at the same collection moment into a comprehensive data point; For any comprehensive data point, clustering the comprehensive data point according to the calculated neighborhood range to obtain a clustering result, and evaluating the load status of each comprehensive data point based on the clustering result; Among them, the process of obtaining the neighborhood range is: Calculating the dimension importance of each dimension parameter, and calculating the weighted Euclidean distance between each comprehensive data point and all other comprehensive data points based on the dimension importance; Calculating the isolation degree of each comprehensive data point in time, and combining the weighted Euclidean distance to obtain the abnormality of the comprehensive data point; Adjusting the preset initial neighborhood range of the comprehensive data point according to the abnormality, and further obtaining the final neighborhood range.

2. The load monitoring method for a photovoltaic energy storage battery according to claim 1, wherein The clustering is the DBSCAN clustering algorithm.

3. The load monitoring method for a photovoltaic energy storage battery according to claim 2, characterized in that Evaluating the load status of each comprehensive data point based on the clustering result includes: Dividing all comprehensive data points into core points, edge points and isolated points based on the clustering result, calculating the proportion of isolated points, and when the proportion of isolated points is greater than a preset abnormality threshold, sending an abnormality warning.

4. A load monitoring method for a photovoltaic energy storage battery according to claim 3, characterized in that The process of obtaining the dimension importance includes: Based on a single dimension parameter among all the collected dimension parameters, calculating the standard deviation of the dimension parameter, and obtaining the maximum value and the minimum value of the dimension parameter, and normalizing the ratio of the standard deviation to the difference between the maximum value and the minimum value to obtain the dimension importance.

5. A load monitoring method for a photovoltaic energy storage battery according to claim 4, characterized in that, The weighted Euclidean distance satisfies the relational expression: ; wherein, represents the weighted Euclidean distance between the -th and the -th comprehensive data points, is the value of the -th dimension parameter in the -th comprehensive data point, is the value of the -th dimension parameter in the -th comprehensive data point, is the importance of the -th dimension parameter, is the total number of the collected dimension parameters.

6. A load monitoring method for a photovoltaic energy storage battery according to claim 5, characterized in that, The calculating the isolation degree of each comprehensive data point in time includes: Selecting any comprehensive data point as the target data set, obtaining the collection moments corresponding to the first K comprehensive data points in the target sequence of the target data set, calculating the absolute value of the difference between the target data set and the sampling moments of the first K comprehensive data points in its target sequence as the first parameter, and calculating the standard deviation of the difference between the target data set and the sampling moments of the first K comprehensive data points in its target sequence as the second parameter; Calculating the mean value of the sum of the ratios of all the first parameters to the second parameters and normalizing it, and taking the normalized result as the isolation degree of the target data set in time.

7. A load monitoring method for a photovoltaic energy storage battery according to claim 6, characterized in that, The abnormality satisfies the relational expression: ; wherein, is the abnormality of the th target data set, is the degree of isolation in time of the th target data set, is the weighted Euclidean distance between the th target data set and the th comprehensive data point in its target sequence, is the standard deviation of the weighted Euclidean distances between the th target data set and all the comprehensive data points in its target sequence, represents the standard normalization function.

8. A load monitoring method for a photovoltaic energy storage battery according to claim 7, characterized in that Adjusting the preset initial neighborhood range of the comprehensive data point according to the abnormality includes: Presetting a hyperparameter, when the difference between the hyperparameter and the abnormality is less than 0, narrowing the initial neighborhood range; conversely, when the difference between the hyperparameter and the abnormality is greater than 0, expanding the initial neighborhood range.

9. A load monitoring method for a photovoltaic energy storage battery according to claim 3, characterized in that, The process of obtaining the dimension importance includes: Based on a single dimension parameter among all the collected dimension parameters, calculating the standard deviation of the dimension parameter, and obtaining the quartiles of the dimension parameter, and normalizing the ratio of the standard deviation to the quartiles to obtain the dimension importance.

10. A load monitoring system for a photovoltaic energy storage battery, characterized in that, Including: A processor and a memory, the memory stores computer program instructions, and when the computer program instructions are executed by the processor, the load monitoring method for a photovoltaic energy storage battery according to any one of claims 1-9 is implemented.

Images