A hierarchical processing system for energy storage data

Abstract

The present invention discloses a hierarchical processing system for energy storage data, including an optimized feature extraction module: collecting initial energy storage data, constructing a unified time series through a specified time interval on the time dimension of the initial energy storage data for aligned sampling, introducing the aligned sampled initial energy storage data into an autoencoder for noise reduction processing, extracting features from the data after noise reduction processing to obtain optimized energy storage feature data; a dynamic hierarchical adjustment module: performing standardization processing on the optimized energy storage feature data, calculating the information entropy of the standardized features to obtain the initial weights, introducing the Q reinforcement learning algorithm to optimize the weights online, designing a third-order hierarchical model based on the optimized initial weights, classifying different combinations of optimized energy storage feature data and dynamic weights according to the third-order hierarchical model, and dividing the data into emergency level, important level and ordinary level based on the third-order hierarchical model, and providing targeted processing and management strategies for data at different levels.

Description

Technical Field

[0001] The present invention relates to the technical field of data hierarchical processing, and particularly to a hierarchical processing system for energy storage data. Background Art

[0002] Modern energy storage data covers various types of device data (such as lithium batteries, lead-acid batteries, etc.) and is connected to a complex energy network. The generated data has the characteristics of multi-source heterogeneity, including device operation parameters, environmental monitoring data, user operation records, etc. There are significant differences in format, acquisition frequency, timestamp, etc. among different data sources. Lack of an effective fusion processing mechanism, it is difficult to form an organic whole, seriously affecting the accuracy and efficiency of data analysis.

[0003] Traditional data grading methods usually based on fixed rules or simple statistical features cannot adapt to the dynamic changes of energy storage data characteristics and business requirements. In terms of weight determination, they cannot fully reflect the importance of data characteristics and their changes, reducing the overall effectiveness of data processing and increasing the operation risk of the energy storage system.

[0004] At present, due to the lack of an effective association processing mechanism for multi-source heterogeneous energy storage data, it is impossible to form an organic whole. And traditional data grading systems are difficult to dynamically adjust with the changes of data characteristics and business requirements, resulting in the disconnection between the grading results and the importance and application scenarios of actual data. This not only reduces the efficiency of data processing, but also may cause key data not to be processed and concerned in time, increasing the operation risk of the energy storage system. Therefore, a hierarchical processing system for energy storage data is proposed here. Summary of the Invention

[0005] In order to overcome the above-mentioned defects of the prior art and achieve the above object, the present invention proposes the following technical solutions:

[0006] A hierarchical processing system for energy storage data, comprising:

[0007] Optimized feature extraction module: Collect initial energy storage data, construct a unified time series for alignment sampling at a specified time interval in the time dimension of the initial energy storage data, introduce the aligned sampled initial energy storage data into an autoencoder for noise reduction processing, and perform feature extraction on the noise-reduced data to obtain optimized energy storage feature data; Dynamic grading adjustment module: Perform standardization processing on the optimized energy storage feature data, calculate the information entropy of the standardized features to obtain the initial weights, introduce the Q reinforcement learning algorithm to optimize the weights online, design a third-order grading model based on the optimized initial weights, and perform data grading on different combinations of optimized energy storage feature data and dynamic weights according to the third-order grading model;

[0008]

[0009] ​Hierarchical storage optimization module: According to the hierarchical results of dynamic hierarchical classification, construct a cost function, calculate the storage cost change rate of data at each level in different time stages through the cost function, and combine the cost sensitivity coefficient and time weight to obtain the calculated data heat index, and set a storage threshold , when the data heat index exceeds the storage threshold , it is high-frequency access data and is migrated to SSD. When the data heat index is lower than or equal to the storage threshold , it is low-frequency access data and is migrated to HDD.

[0010] The initial energy storage data includes energy storage device operation data, energy storage environment monitoring data, and user operation record data;

[0011] Let the total number of data points collected from all data sources be , and the data point set of the th data source collected is , constituting the initial energy storage data of multi-source heterogeneous.

[0012] The process of aligning and sampling the data of each data source according to the unified time series is as follows:

[0013] Perform alignment processing on the multi-source heterogeneous data from the time dimension. For each data point in the data source , its corresponding timestamp is , and unify the data of all data sources to an optimal time interval . The determination process of is as follows:

[0014] Analyze and collect the original acquisition frequencies of each data source. For the th data source , its original acquisition frequency is denoted as , where is the number of data points in the th data source ;

[0015] Each data point is accompanied by a timestamp. Set a search range for the time scale, starting from the time interval 0.1, and gradually increasing to the time interval 10 seconds at a certain step of 0.1 seconds per second. Denote the time scale set as , where is the minimum time scale, is the maximum time scale. Obtain the change indicators of the data at each time scale, analyze the change trend of these change indicators with the time scale, and find the time scale that makes the data change indicators tend to be stable. The time interval corresponding to this time scale is .

[0016] The process of obtaining the optimized energy storage characteristic data is as follows:

[0017] Based on the time interval Construct a unified time series , is the starting sampling time point, is the last sampling time point. Align and sample the data of each data source according to the unified time series to obtain the initial energy storage data after alignment sampling , where is the number of all data points, and the total number of data points remains unchanged after alignment;

[0018] Introduce the initial energy storage data after alignment sampling into the autoencoder for noise reduction processing;

[0019] Take the initial energy storage data after alignment sampling as the input of the autoencoder. The autoencoder maps the high-dimensional input data to a low-dimensional hidden layer representation through the encoder, and the decoder then restores the hidden layer representation to the reconstructed data, and adjusts the encoder and decoder by minimizing the reconstruction error value of the reconstructed data. The formula is expressed as:

[0020]

[0021] where is the number of all input data, is the th data in the input data, is the th element in the reconstructed data;

[0022] Train the autoencoder through the backpropagation algorithm, transfer the reconstruction error back to each layer of the encoder and decoder, calculate the gradient and update the weight matrix and bias vector, so that the reconstruction error gradually decreases. After training, extract features from the data after noise reduction processing to obtain the optimized energy storage characteristic data .

[0023] The process of obtaining the optimized initial weight is as follows:

[0024] Suppose the optimized energy storage characteristic data contains different feature dimensions, and through a optimized energy storage characteristic data matrix , where is the number of data points of the optimized energy storage characteristic data, is the number of feature dimensions of the optimized energy storage characteristic data. Each row corresponds to a data point, and each column corresponds to a feature dimension. Standardize the optimized energy storage characteristic data. For the th feature dimension, the standardization formula is:

[0025]

[0026] Among them, represents the data point corresponding to the th feature dimension, is the mean of the th feature dimension, is the standard deviation of the th feature dimension;

[0027] Calculate the information entropy of each standardized feature. For the th feature dimension, the steps to obtain the information entropy are as follows:

[0028] For the th feature dimension, calculate the probability of each data point in the th dimension, and divide the standardized feature value of each data point in the th dimension by the sum of the standardized feature values of all data points in this feature dimension, to obtain the probability of each data point in the th dimension;

[0029] According to the calculated probability , further calculate the information entropy of the th feature dimension, and the formula is expressed as:

[0030]

[0031] Then, based on the information entropy obtain the initial weight of each feature dimension. Divide the complementary value of each feature dimension by the sum of the complementary values of all feature dimensions, and the obtained result is the initial weight of this feature dimension;

[0032] Then introduce the Q reinforcement learning algorithm to optimize the weights online;

[0033] Determine a state space according to the combination of the value ranges of the feature dimensions;

[0034] Determine an action space according to the set of adjustment operations for each feature weight;

[0035] Create and initialize a value matrix of , where is the size of the state space, is the size of the action space. At each time step, the current state is determined and an action is selected from the action space with a probability of , and an action with the maximum value in the current state is selected with a probability of . After the selection, the weights are adjusted and an immediate reward is given according to the processing result . When the accuracy of the classification result improves, a positive reward is given, and when the accuracy decreases, a negative reward is given, and then values are updated. Repeat until the preset maximum number of learning steps is reached or the value matrix converges, and finally the optimized initial weights are obtained

[0036] The implementation process of designing a third-order classification model based on the optimized initial weights is as follows:

[0037] Construct a third-order classification model, and the third-order classification of the model is the emergency level, the important level, and the normal level;

[0038] Based on the different feature dimensions included in the optimized energy storage feature data and the optimized initial weight vector obtained in the previous steps, denoted as , where represents the optimized weight of the th feature dimension ;

[0039] The data is divided into three levels: the emergency level , the important level , and the normal level . Two thresholds and are set as the basis for classification determination;

[0040] Among them, the determination condition for the emergency level is: for the th data point ([[]] ) of the th feature dimension, its optimized energy storage feature data is . When , the corresponding data point is determined to be at the emergency level, where is the number of feature dimensions used for emergency level determination ;

[0041] The determination condition for the important level is: not meeting the emergency level condition and meeting , then this data point is determined to be of high importance level;

[0042] Condition for determining the ordinary level L3: If a data point neither meets the conditions for the emergency level nor the conditions for the high importance level, it is determined to be at the ordinary level;

[0043] After classification processing by the three - level classification model, three sets of classified data are obtained, which are respectively , and .

[0044] The construction process of the cost function is as follows:

[0045] Let the set of storage media be , where represents SSD, represents HDD. For the storage media, when storing data at the th level, where, = 1, 2, 3, and respectively corresponds to , , , the unit storage cost of the data is , the storage capacity is , the data volume of each data level is . Define a cost function to calculate the storage cost of using the storage media to store the data at the th level:

[0046]

[0047] Among them, represents the index of the storage media set being or .

[0048] The process of obtaining the data heat index is as follows:

[0049] Calculate the storage cost change rate of each level of data at different time stages through the cost function, and calculate the relative change ratio of the comprehensive storage cost of the th level data in two adjacent time stages. By subtracting the storage cost of the th time stage from the storage cost of the th time stage, the change amount of the storage cost is obtained, and then this change amount is divided by the storage cost of the previous time stage to obtain the relative change rate ;

[0050] Set a cost - sensitivity coefficient for each level according to the level importance of different data ;

[0051] Set a time weight according to different time periods , where each time stage has a corresponding time weight , and the sum of the weights of all time stages is 1;

[0052] By the storage cost change rate , the cost sensitivity coefficient and the time weight , calculate the data heat index: where represents the grading where the data point is located.

[0053] The present invention has the following beneficial effects:

[0054] In the present invention, first of all, by optimizing the feature extraction module, in the processing of multi-source heterogeneous data, the unified time series and aligned sampling solve the differences in format, acquisition frequency and timestamp of different data sources. On the time dimension of the initial energy storage data, a unified time series is constructed with a time interval , and the data of each data source is aligned and sampled. Subsequently, denoising processing is performed using an autoencoder, effectively solving the problem that there is a lack of an effective correlation processing mechanism for multi-source heterogeneous energy storage data, integrating the originally scattered and differently formatted data into an organic whole, laying a foundation for subsequent data processing and analysis, and improving the usability and consistency of the data;

[0055] Secondly, through the dynamic grading adjustment module, the dimension influence is eliminated by using data standardization processing, making the data of different feature dimensions comparable, laying a foundation for accurately calculating the information entropy and weight. Then, the initial weight is determined based on the information entropy to highlight important features. Features with low information entropy have high weights in subsequent analysis, improving the accuracy and effectiveness of data processing. Then, the weight is optimized online through the Q reinforcement learning algorithm to adapt to the dynamic changes of the data. Based on the optimized weight, a third-order grading model is constructed and the data is divided into emergency level, important level and ordinary level, providing targeted processing and management strategies for data at different levels;

[0056] Finally, through the grading storage optimization module, according to the results of dynamic grading, a cost function is constructed to comprehensively consider the storage medium, data level, data volume and storage capacity, providing a quantitative basis for storage cost evaluation, assisting in storage resource allocation decisions, calculating the data heat index in combination with the cost sensitivity coefficient and time weight, comprehensively reflecting the changes in data heat, and making the storage decision more reasonable. BRIEF DESCRIPTION OF THE DRAWINGS

[0057] Figure 1 is a system block diagram of a hierarchical processing system for energy storage data proposed by the present invention. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0058] Next, in combination with the accompanying drawings in the embodiments of the present invention, the technical solutions in the embodiments of the present invention will be clearly and completely described. Obviously, the described embodiments are only a part of the embodiments of the present invention, rather than all the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those of ordinary skill in the art without creative efforts shall fall within the protection scope of the present invention.

[0059] Embodiment 1

[0060] As Figure 1 shown, a hierarchical processing system for energy storage data proposed by the present invention includes:

[0061] Optimized feature extraction module: Collect initial energy storage data, construct a unified time series for alignment sampling at a specified time interval in the time dimension of the initial energy storage data, introduce the aligned sampled initial energy storage data into an autoencoder for noise reduction processing, and perform feature extraction on the data after noise reduction processing to obtain optimized energy storage feature data; Implementation process of the optimized feature extraction module:

[0062] Collect initial energy storage data from different data sources, and these data sources cover all aspects during the operation of the energy storage system, specifically including:

[0063] Energy storage device operation data: Collect device operation status data such as the voltage, current, power, and temperature of the energy storage battery, as well as the power conversion parameters of the inverter;

[0064] Energy storage environment monitoring data: Obtain environmental parameters such as environmental temperature, humidity, air pressure, and light intensity;

[0065] User operation record data: Record the operation instructions of the user on the energy storage system, such as the time and power settings of charging and discharging operations;

[0066] Assume that the total number of data points collected from all data sources is

[0067] , and the data point set of the th data source collected is , and the data from these data sources are different in format, collection frequency, timestamp, etc., constituting multi-source heterogeneous initial energy storage data;

[0068] Align the multi-source heterogeneous data in the time dimension. For each data point in the data source , its corresponding timestamp is , and unify the data from all data sources to an optimal time interval . The determination process of is as follows:

[0069] Analyze the original acquisition frequencies of each data source. For example, the energy storage device operation sensors collect data multiple times per second, while the environmental monitoring devices may collect data once every few minutes. Collect the original acquisition frequencies of each data source. For the nth data source , its original acquisition frequency is denoted as , where is the number of data points in the nth data source . Each data point has a timestamp. Set the search range of a time scale, starting from a time interval of 0.1 and gradually increasing to a time interval of 10 seconds at a certain step size of 0.1 per second. Denote the set of time scales as , where is the minimum time scale, is the maximum time scale. Obtain the change indicators of the data at each time scale, analyze the change trend of these change indicators with the time scale, and find a time scale such that the data change indicators tend to be stable. The time interval corresponding to this time scale is the appropriate value ;

[0070] Specifically, based on the analysis of the change trend, finding a specific time scale such that the data change indicators tend to be stable is because when the time scale reaches this specific time scale, as the time scale further increases, the change amplitude of the data change indicators becomes very small, that is, tends to be stable. This means that after this time scale, continuing to increase the time interval, the fluctuation characteristics of the data will not change significantly. Therefore, the time interval corresponding to this time scale is determined as the appropriate unified time interval ;

[0071] Determine the unified time interval After that, construct the unified time series , is the starting sampling time point, is the last sampling time point. Align and sample the data of each data source according to the unified time series to obtain the initial energy storage data after alignment sampling , where is the total number of all data points, and the total number of data points remains unchanged after alignment;

[0072] After aligning and sampling the data of each data source through the unified time series, the data of different data sources have consistency on the time axis, which is convenient for subsequent analysis and processing;

[0073] The process of introducing the initial energy storage data after alignment sampling into the autoencoder for noise reduction is as follows:

[0074] The initial energy storage data after alignment sampling As the input of the autoencoder, the autoencoder maps the high-dimensional input data to a low-dimensional hidden layer representation through the encoder, and the decoder then restores the hidden layer representation to the reconstructed data, and adjusts the encoder and decoder by minimizing the reconstruction error value of the reconstructed data. The formula is expressed as:

[0075]

[0076] where, is the number of all input data (one input data corresponds to one initial energy storage data after aligned sampling, the same as the above ), is the -th data in the input data, is the -th element in the reconstructed data;

[0077] Then, through the backpropagation algorithm, the reconstruction error is passed back to each layer of the encoder and decoder, the gradients are calculated, and the weight matrix and bias vector are updated, so that the reconstruction error gradually decreases. After training, the autoencoder can effectively remove the noise in the input data, extract more pure and representative data features, and provide high-quality data for the subsequent energy storage data processing steps;

[0078] After training the autoencoder with the backpropagation algorithm, feature extraction is performed on the denoised data to obtain optimized energy storage feature data .

[0079] Dynamic hierarchical adjustment module: Standardize the optimized energy storage feature data, calculate the information entropy of the standardized features to obtain the initial weights, introduce the Q reinforcement learning algorithm to optimize the weights online, design a third-order hierarchical model based on the optimized initial weights, and perform data grading on different combinations of optimized energy storage feature data and dynamic weights according to the third-order hierarchical model;

[0080] Implementation process of the dynamic hierarchical adjustment module:

[0081] Suppose the optimized energy storage feature data contains different feature dimensions, through a optimized energy storage feature data matrix , where is the number of data points of the optimized energy storage feature data, is the number of feature dimensions of the optimized energy storage feature data. Each row corresponds to a data point, and each column corresponds to a feature dimension. Standardize the optimized energy storage feature data. For the -th feature dimension , the standardization formula is:

[0082]

[0083] Among them, represents the data point corresponding to the th feature dimension, is the mean of the th feature dimension, is the standard deviation of the th feature dimension;

[0084] Calculate the information entropy of each standardized feature. For the th feature dimension, the steps to obtain the information entropy are as follows:

[0085] For the th feature dimension, calculate the probability of each data point in the , divide the standardized feature value of each data point in the th dimension by the sum of the standardized feature values of all data points under this feature dimension , to obtain the probability of each data point in the th dimension;

[0086] According to the calculated probability , further calculate the information entropy of the th feature dimension. First, take the natural logarithm of the probability of each data point in the th feature dimension and multiply it by itself, then sum the above results for each data point, and finally multiply the sum result by , so as to obtain the information entropy of the th feature dimension, and the formula is expressed as:

[0087]

[0088] Then, based on the information entropy obtain the initial weight of each feature dimension. Divide the complementary value of each feature dimension by the sum of the complementary values of all feature dimensions , and the obtained result is the initial weight of this feature dimension;

[0089] Among them, the initial weight is assigned according to the information entropy of the feature dimension. For the feature dimension with low information entropy (i.e., high proportion of effective information), its value is relatively large, and the initial weight is also relatively high, which means that in subsequent data processing and analysis, these feature dimensions will be assigned greater weights and can play their roles more fully. For feature dimensions with high information entropy (low proportion of effective information), their values are relatively small, and the obtained initial weights are also relatively low, and their influence in subsequent operations is relatively small. Such a weight allocation method helps to highlight important features and improve the accuracy and effectiveness of data processing and analysis;

[0090] The implementation process of introducing the Q reinforcement learning algorithm to optimize the weights online is as follows:

[0091] Determine a state space according to the combination of the value ranges of feature dimensions and determine an action space according to the set of adjustment operations for each feature weight , create and initialize a of value matrix , where is the size of the state space, is the size of the action space. At each time step, determine the current state and select an action from the action space with a probability of , and select the action with the largest value in the current state with a probability of . After executing the selected action, adjust the weights and give an immediate reward according to the processing result . When the accuracy of the grading result improves, give a positive reward; when the accuracy decreases, give a negative reward, and then update the value. Repeat the above steps until the termination condition is met, and finally obtain the optimized initial weights ;

[0092] Among them, in the framework of the Q reinforcement learning algorithm, the action space contains various adjustment operations for feature weights. When an action is selected from the action space at a certain time step, this action needs to be executed and the feature weights need to be modified accordingly. Taking increasing the weight of a certain feature as an example, a preset adjustment amplitude (increasing the current weight of this feature by 10%) is set. Suppose the current weight of the feature is . When the selected action is to increase the weight of this feature, the adjusted weight value will increase by 10% of the current weight. When the selected action is to increase the weight of this feature, the adjusted weight value will decrease by 10% of the current weight;

[0093] Specifically, when selecting an action, The initial value can be set to 0.2 and adjusted dynamically according to the convergence situation during the learning process. The learning rate ranges from (0, 1), and the recommended initial value is set to 0.1, gradually decaying according to the learning effect. The discount factor ranges from (0, 1), and the recommended initial value is set to 0.9;

[0094] The termination conditions require: set to reach the preset maximum number of learning steps (1000 steps), the value matrix converges (the sum of the absolute values of the value changes in two adjacent iterations is less than or the data classification performance index reaches a satisfactory level (the classification accuracy rate reaches over 95%));

[0095] The implementation process of designing a third - order classification model based on the optimized initial weights is as follows:

[0096] Construct a third - order classification model, and the third - order classification of the model is emergency level, important level, and normal level;

[0097] It is known that the optimized energy storage feature data contains different feature dimensions. The optimized initial weight vector obtained through the previous steps is denoted as , where represents the optimized weight of the th feature dimension ;

[0098] Divide the data into three levels: emergency level , important level and normal level , and set two thresholds and to be used as the basis for classification determination;

[0099] Among them, the determination condition for the emergency level is: for the th data point of the th feature dimension( ), its optimized energy storage feature data is . When , the corresponding data point is determined to be at the emergency level. Among them, is the number of feature dimensions used for emergency - level determination , which is determined according to the actual business requirements and data characteristics;

[0100] The determination condition for the important level is: if the data point does not meet the emergency - level condition and meets , then the data point is determined to be at the important level;

[0101] General - level L3 determination condition: If a data point satisfies neither the emergency - level condition nor the important - level condition, it is determined to be of the general level;

[0102] After being processed by the three - level classification model, three groups of classified data are obtained, which are respectively , and ;

[0103] Specifically, the emergency level refers to data that is closely related to the key operating states and potential risks of the energy - storage system. For the emergency level, the data - characteristic dimensions are usually closely related to the key system states or emergency situations. For example, in an energy - storage system, core parameters such as battery voltage and current are incorporated into the characteristic dimensions for emergency - level determination and are given relatively high weights. When these key characteristic dimensions exceed the threshold through the classification model, it indicates that the system is in an emergency state, and the corresponding data points are determined to be of the emergency level;

[0104] The important level refers to data that reflects the important operating parameters and relevant information of the energy - storage system. For important - level determination, on the premise that the data point does not satisfy the emergency - level condition, it is considered whether the result obtained by the classification model for the remaining characteristic dimensions exceeds the threshold . Although these characteristic dimensions do not play a dominant role in emergency - situation judgment, they are crucial for reflecting the important operating parameters and relevant information of the system, such as characteristic dimensions like environmental humidity and charge - discharge power;

[0105] General - level L3 refers to general operating - record data. If a data point satisfies neither the emergency - level condition nor the important - level condition, it is determined to be of the general level, that is, when neither of the above two determination situations is satisfied, the data point is classified into the general level.

[0106] Hierarchical - storage optimization module: According to the classification results of dynamic classification, a cost function is constructed. By calculating the storage - cost change rate of data at each level in different time stages through the cost function and combining the cost - sensitivity coefficient and time weight, the data - heat - index is obtained. A storage threshold is set. When the data - heat - index exceeds the storage threshold , it is high - frequency - accessed data and is migrated to the SSD. When the data - heat - index is lower than or equal to the storage threshold , it is low - frequency - accessed data and is migrated to the HDD;

[0107] Implementation process of the hierarchical - storage optimization module:

[0108] According to the classification results of dynamic classification, a cost function is constructed. Let the set of storage media be , where Representing SSD, representing HDD. For the storage medium, when storing the th level ( = 1, 2, 3, and corresponding to 、 、 respectively), the unit storage cost of the data is , the storage capacity is , the data volume of each data level is . Define a cost function to calculate the storage cost of using the storage medium ( = 1, 2) to store the th level data: , represents the index of the storage medium set or ;

[0109] Calculate the storage cost change rate of each level of data in different time periods through the cost function, and calculate the relative change ratio of the comprehensive storage cost of the th level data in two adjacent time periods. By subtracting the storage cost of the th time period from the storage cost of the th time period, the change amount of the storage cost is obtained, and then this change amount is divided by the storage cost of the previous time period to obtain the relative change rate ,

[0110] At the same time, according to the level importance of different data, set a cost sensitivity coefficient for each level, , for example, the sensitivity coefficient of the emergency level is set to , the sensitivity coefficient of the important level is set to , the sensitivity coefficient of the normal level is set to . The larger the sensitivity coefficient, the more sensitive the high-level data is to cost changes;

[0111] Set a time weight according to different time periods, ( , represents the total number of divided time periods, ), each time period has a corresponding time weight , and the sum of the weights of all time periods is 1;

[0112] Finally, through the storage cost change rate , cost sensitivity coefficient and time weight Calculate the data heat index: Among them, represents the classification where the data point is located;

[0113] Set a storage threshold , when that is, when the data heat index exceeds the upper threshold , it is high-frequency access data and is migrated to the SSD. When that is, when the data heat index is lower than or equal to the storage threshold , it is low-frequency access data and is migrated to the HDD.

[0114] In this embodiment, in the face of a large amount of multi-source heterogeneous data generated by numerous energy storage devices and a complex operating environment, through this solution, the optimized feature extraction module collects sensor data of each device, environmental monitoring data, and user operation records of each device, unifies the time series and reduces noise, integrates the data of each energy storage node, and provides high-quality data for the power station operation and maintenance personnel after processing. The dynamic grading adjustment module grades the data, reflects the grading of the importance of the system state, and provides a basis for energy scheduling. Based on the grading storage optimization module, the operation and maintenance personnel can focus on the emergency-level data, discover and handle potential risks in a timely manner. The grading storage optimization module determines the data storage location based on the heat index and the threshold, comprehensively considers the storage medium, data level, data volume, and storage capacity, provides a quantitative basis for the storage cost assessment, and assists in the decision-making of the storage resource allocation.

[0115] In the application, several formulas involved are calculated by taking the numerical values after dimensionlessization, and the establishment of the formulas is a formula obtained by collecting a large amount of data for software simulation to approximate the real situation. Some coefficients or weights in the formulas are set by those skilled in the art according to the actual situation, so no more details will be elaborated here.

[0116] The above embodiments can be implemented in whole or in part by software, hardware, firmware, or any other combination. When implemented using software, the above embodiments can be implemented in whole or in part in the form of a computer program product. Those of ordinary skill in the art can realize that the units and algorithm steps of the examples described in conjunction with the embodiments disclosed herein can be implemented by electronic hardware, or a combination of computer software and electronic hardware. Whether these functions are executed in a hardware or software manner depends on the specific application and design constraints of the technical solution.

[0117] Although embodiments of the present invention have been shown and described, those of ordinary skill in the art will appreciate that various changes, modifications, substitutions and variations can be made to these embodiments without departing from the principles and spirit of the present invention. The scope of the present invention is defined by the appended claims and their equivalents.

Claims

1. A hierarchical processing system for energy storage data, characterized in that: include: Optimize feature extraction module: collect initial energy storage data, and use specified time intervals in the time dimension of the initial energy storage data Construct a unified time series for aligned sampling, introduce the initial energy storage data after aligned sampling into the autoencoder for noise reduction, perform feature extraction on the noise-reduced data, and obtain optimized energy storage feature data; Dynamic grading adjustment module: standardize the optimized energy storage feature data, calculate the information entropy of the standardized features to obtain the initial weights, introduce the Q reinforcement learning algorithm to optimize the weights online, design a three-order grading model based on the optimized initial weights, and perform data grading for different optimized energy storage feature data combinations and dynamic weights according to the three-order grading model; Among them, the optimized initial weight acquisition process is: Assume that the optimized energy storage characteristic data includes different feature dimensions, through a Optimized energy storage characteristic data matrix ,in To optimize the number of data points for energy storage characteristic data, In order to optimize the number of characteristic dimensions of energy storage characteristic data, each row corresponds to a data point, and each column corresponds to a characteristic dimension. The optimized energy storage characteristic data is standardized. feature dimensions, the standardized formula is: in, Indicates The data points corresponding to the feature dimensions are It is The mean of the feature dimensions, It is The standard deviation of the feature dimension; Calculate the information entropy of each standardized feature. feature dimensions, information entropy The steps to obtain are as follows: For feature dimensions, calculate each data point in Probability in Dimensions , each in the The standardized feature values ​​of the data points in the dimension Divide by the sum of the standardized feature values ​​of all data points in this feature dimension , get each data point in Probability in Dimensions ; According to the calculated probability , further calculate the Information entropy of feature dimensions , the formula is: Then based on information entropy Get the initial weight for each feature dimension , the complementary value of each feature dimension Divide by the sum of the complementary values ​​of all feature dimensions , the result is the initial weight of the feature dimension ; Then the Q reinforcement learning algorithm is introduced to optimize the weights online; Determine a state space based on the range of feature dimension values ; Determine an action space based on a set of adjustment operations on each feature weight ; Create and initialize a of Value Matrix ,in is the size of the state space, is the size of the action space, at each time step the current state is determined and The probability of moving from the action space Select an action from The probability of choosing The action with the largest value will adjust the weight of the selected action and give an immediate reward based on the processing result. , when the accuracy of the classification result improves, a positive reward is given, and when the accuracy decreases, a negative reward is given, and then the update value, repeat until the preset maximum number of learning steps is reached or The value matrix converges and finally obtains the optimized initial weights ; Hierarchical storage optimization module: Based on the hierarchical results of dynamic grading, a cost function is constructed. The storage cost change rate of each level of data at different time stages is calculated through the cost function, and the data heat index is calculated by combining the cost sensitivity coefficient and time weight, and a storage threshold is set. , when the data heat index exceeds the storage threshold When the data is frequently accessed, it is migrated to SSD. When the data heat index is lower than or equal to the storage threshold When the data is accessed infrequently, it is migrated to HDD.

2. A hierarchical processing system for energy storage data according to claim 1, characterized in that: The initial energy storage data includes energy storage equipment operation data, energy storage environment monitoring data and user operation record data; Assume that the total number of data points collected by all data sources is , collected The data point set of the data source is , forming multi-source heterogeneous initial energy storage data.

3. A hierarchical processing system for energy storage data according to claim 1, characterized in that: The process of aligning and sampling the data from each data source according to the unified time series is as follows: Align multi-source heterogeneous data from the time dimension, and Data points in , and its corresponding timestamp is , unify the data from all data sources into an optimal time interval superior, The determination process is: Analyze and collect the original collection frequency of each data source. Data sources , and its original acquisition frequency is recorded as ,in For the Data sources The number of data points in ; Each data point has a timestamp. A time scale search range is set, starting from a time interval of 0.1 seconds, and gradually increasing to a time interval of 10 seconds with a certain step size of 0.1 per second. The time scale set is recorded as ,in is the minimum time scale, As the maximum time scale, obtain the change index of the data at each time scale, analyze the change trend of these change indexes with the time scale, and find the time scale that makes the data change index tend to be stable. The time interval corresponding to this time scale is .

4. The hierarchical processing system for energy storage data according to claim 1, characterized in that: The process of obtaining the optimized energy storage characteristic data is as follows: Based on time interval Constructing unified time series , is the starting sampling time point, As the last sampling time point, the data of each data source are aligned and sampled according to the unified time series to obtain the initial energy storage number after aligned sampling. ,in, is the number of all data points. After alignment, the number of overall data points remains unchanged. The initial energy storage data after alignment sampling is introduced into the autoencoder for noise reduction processing; The initial energy storage number after alignment sampling As the input of the autoencoder, the autoencoder maps the high-dimensional input data to a low-dimensional hidden layer representation through the encoder, and the decoder restores the hidden layer representation to the reconstructed data, and adjusts the encoder and decoder by minimizing the reconstruction error value of the reconstructed data. The formula is expressed as: in, is the number of all input data, is the first data, is the first elements; The autoencoder is trained through the back-propagation algorithm, and the reconstruction error is passed back to each layer of the encoder and decoder. The gradient is calculated and the weight matrix and bias vector are updated so that the reconstruction error is gradually reduced. After training, the feature extraction of the denoised data is performed to obtain the optimized energy storage feature data. .

5. The hierarchical processing system for energy storage data according to claim 1, characterized in that: The implementation process of designing the three-order classification model based on the optimized initial weights is as follows: Construct a three-level classification model, the three levels of the model are emergency level, important level and ordinary level; Based on optimizing energy storage characteristic data Different feature dimensions and the initial weight vector optimized in the previous steps are recorded as ,in Indicates The optimized weights of feature dimensions ; The data is divided into three levels: emergency level , Importance and ordinary level , set two thresholds and , used as the basis for grading determination; Among them, emergency The judgment condition is: The feature dimension Data points ( ), and its optimized energy storage characteristic data is ,when When , the corresponding data point is judged as emergency level, where is the number of feature dimensions used for emergency level determination ; Importance The judgment conditions are: the emergency level conditions are not met and the , then the data point is judged as important; Normal level L3 judgment conditions: If the data point meets neither the emergency level conditions nor the important level conditions, it is judged as normal level; After the classification process by the three-order classification model, three groups of data were obtained, which are: , as well as .

6. The hierarchical processing system for energy storage data according to claim 1, characterized in that: The construction process of the cost function is: Suppose the storage medium set is ,in Represents SSD, Represents HDD, for storage media, storage Level, among which, =1, 2, 3, and correspond to , , , the unit storage cost of data is , the storage capacity is , the amount of data at each data level is , define a cost function to calculate the use of storage media Storage Storage cost of level data: in, Indicates that the storage medium set is or The index of .

7. The hierarchical processing system for energy storage data according to claim 1, characterized in that: The process of obtaining the data heat index is as follows: The cost function is used to calculate the storage cost change rate of each level of data in different time stages, and the storage cost change rate of each level of data in two adjacent time stages is calculated. The relative change ratio of the comprehensive storage cost of the first-level data is calculated by using The storage cost of the time period minus the storage cost of the The storage cost of a time period is calculated to obtain the change in storage cost, and then this change is divided by the storage cost of the previous time period to obtain the relative change rate. ; According to the importance of different data levels, set the cost sensitivity coefficient for each level ; Set a time weight according to different time periods , where each time stage There is a corresponding time weight , the sum of the weights of all time stages is 1; By storage cost change rate , cost sensitivity coefficient and time weight Calculate the data heat index: in, Indicates the bin to which the data point belongs.

Images