A novel distributed storage method and system for energy storage data of intelligent energy storage station

Abstract

The present invention relates to the field of energy storage data processing technology, and specifically to a new type of intelligent energy storage station energy storage data distributed storage method and system, the method comprising: determining the energy storage data of each battery per minute in the new type of intelligent energy storage station; determining the fit reward value when performing each storage action on the energy storage data of each battery per minute; based on the predetermined spatial distance between each battery and each storage node, weighting the fit reward value to obtain the final reward value for storing the energy storage data of each battery per minute in each storage node; based on the final reward value, determining the optimal distributed storage strategy through a reinforcement learning algorithm; storing the energy storage data of each battery per minute according to the optimal distributed storage strategy. The solution provided by the present invention can effectively aggregate the relevant energy storage data of the same battery into closer storage nodes, so that the subsequent data access and retrieval efficiency are significantly improved.

Description

Technical Field

[0001] The present invention relates to the technical field of energy storage data processing, and in particular to a novel distributed storage method and system for energy storage data of an intelligent energy storage station. Background Art

[0002] New intelligent energy storage stations usually refer to energy storage facilities that integrate advanced battery energy storage technology, intelligent management systems, and data analysis platforms. Since there are a large number of battery packs and corresponding SOC (State of Charge, also known as remaining power) data in the energy storage station, these energy storage data can be shared and processed among multiple storage nodes through distributed storage, which can reduce the data storage pressure of a single storage device and thus improve data processing efficiency.

[0003] In the related art, when the SOC data of each battery in the energy storage station is distributedly stored every minute, the corresponding storage nodes are usually allocated to the energy storage data to be stored only according to the remaining storage capacity of each storage node. However, after the data is stored in the above storage method, it is difficult to quickly locate the relevant energy storage data of the same battery during the data access and retrieval process, resulting in low efficiency of subsequent data access and retrieval.

[0004] It is not difficult to find that the traditional distributed storage solution for energy storage data has the problem of disordered data storage. Summary of the invention

[0005] In order to solve the technical problem of disordered data storage in traditional distributed storage solutions for energy storage data, the purpose of the present invention is to provide a new type of distributed storage method and system for energy storage data in intelligent energy storage stations. The technical solutions adopted are as follows:

[0006] A novel distributed storage method for energy storage data of an intelligent energy storage station, the method comprising:

[0007] Determine the energy storage data of each battery in the new smart energy storage station every minute;

[0008] Determine a compatibility reward value when each storage action is performed on the energy storage data of each battery per minute;

[0009] Based on the predetermined spatial distance between each battery and each storage node, the compatibility reward value is weighted to obtain a final reward value for storing the energy storage data of each battery per minute in each storage node;

[0010] Based on the final reward value, determining an optimal distributed storage strategy through a reinforcement learning algorithm;

[0011] The energy storage data of each battery every minute is stored according to the optimal distributed storage strategy.

[0012] According to a novel distributed storage method for energy storage data of an intelligent energy storage station provided by the present invention, the energy storage data of each battery in the novel intelligent energy storage station is determined every minute, including:

[0013] Construct a cycle number-battery rated capacity model;

[0014] Based on the cycle number-battery rated capacity model, determining the theoretical rated capacity of each battery in the new intelligent energy storage station at each cycle;

[0015] Based on the theoretical rated capacity of each battery in each cycle, the energy storage data of each battery per minute is determined.

[0016] According to a novel distributed storage method for energy storage data of an intelligent energy storage station provided by the present invention, a cycle number-battery rated capacity model is constructed, including:

[0017] Construct an initial model containing unknown parameters;

[0018] Determine the true usability of each charging cycle;

[0019] Establishing a data error square sum function based on the initial model and the actual availability of each charging cycle;

[0020] Determining the optimal solution of the unknown parameters in the initial model based on the data error square sum function;

[0021] Based on the initial model and the optimal solution of the unknown parameters, a cycle number-battery rated capacity model is constructed.

[0022] According to a novel distributed storage method for energy storage data of an intelligent energy storage station provided by the present invention, the real availability of each charging cycle is determined, including:

[0023] Determine the turning point and temperature data of each charging cycle;

[0024] Determining the duration of each charging cycle based on the turning point of each charging cycle;

[0025] Based on the turning point, the time length and the temperature data of each charging cycle, the actual availability of each charging cycle is calculated.

[0026] According to a novel distributed storage method for energy storage data of an intelligent energy storage station provided by the present invention, the turning point of each charging cycle is determined, including:

[0027] Obtain the charging current value corresponding to each charging moment during each charging cycle;

[0028] For any charging moment, based on the charging current values ​​of a preset number of adjacent moments before the charging moment and the charging current values ​​of a preset number of adjacent moments after the charging moment, a probability value of the charging moment being a turning point is calculated;

[0029] Based on the probability value, a turning point of each charging cycle is determined.

[0030] According to a novel distributed storage method for energy storage data of an intelligent energy storage station provided by the present invention, based on the probability value, the turning point of each charging cycle is determined, including:

[0031] Extract the target moment when the probability value is higher than the preset probability threshold to obtain the initial screening result;

[0032] For any target time in the preliminary screening result, determine whether there is an abnormal time point with a probability value higher than a preset probability threshold within a preset period before the target time, and obtain a determination result;

[0033] The target time when the judgment result in the preliminary screening result is negative is used as the turning point of each charging cycle.

[0034] According to a novel distributed storage method for energy storage data of an intelligent energy storage station provided by the present invention, based on the theoretical rated capacity of each battery in each cycle, the energy storage data of each battery per minute is determined, including:

[0035] Obtain the charge change of each battery from the previous minute to the current minute, and obtain the energy storage data of each battery in the previous minute;

[0036] Divide the charge change by the theoretical rated capacity of each battery in each cycle to obtain a correction amount;

[0037] The energy storage data of the previous minute is summed with the correction amount to calculate the energy storage data of each battery in the energy storage station per minute.

[0038] According to a novel distributed storage method for energy storage data of an intelligent energy storage station provided by the present invention, determining a compatibility reward value when performing each storage action on the energy storage data of each battery per minute includes:

[0039] Subtract the energy storage data of each battery per minute from the preset value to obtain the power difference;

[0040] Input the power difference into a preset maximum and minimum value normalization function to calculate the data attention of each battery per minute;

[0041] For each battery, obtaining historical data of the battery in each storage node;

[0042] Based on the data attention and the historical data, a compatibility reward value when each storage action is performed on the energy storage data of each battery per minute is calculated.

[0043] According to a novel distributed storage method for energy storage data of an intelligent energy storage station provided by the present invention, based on the final reward value, an optimal distributed storage strategy is determined by a reinforcement learning algorithm, including:

[0044] Determine the state space based on the energy storage data of the battery at different times and the data status of each storage node;

[0045] Determine the action space based on storing the energy storage data of the battery at different times to any storage node;

[0046] In the state space and the action space, the storage node with the highest final reward value is determined for the energy storage data of each battery per minute through a reinforcement learning algorithm to obtain an optimal distributed storage strategy.

[0047] On the other hand, the present invention also provides a novel intelligent energy storage station energy storage data distributed storage system, the system comprising:

[0048] A determination module, used to determine the energy storage data of each battery in the new intelligent energy storage station every minute;

[0049] A calculation module, used for determining a compatibility reward value when each storage action is performed on the energy storage data of each battery per minute;

[0050] A weighting module, configured to perform weighted processing on the compatibility reward value based on a predetermined spatial distance between each battery and each storage node, to obtain a final reward value for storing the energy storage data of each battery per minute in each storage node;

[0051] An optimization module, used for determining an optimal distributed storage strategy through a reinforcement learning algorithm based on the final reward value;

[0052] A storage module is used to store the energy storage data of each battery every minute according to the optimal distributed storage strategy.

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

[0054] By determining the energy storage data of each battery per minute in the new intelligent energy storage station, and determining the fit reward value when performing each storage action on the energy storage data of each battery per minute, based on the predetermined spatial distance between each battery and each storage node, the fit reward value is weighted to obtain the final reward value for storing the energy storage data of each battery per minute in each storage node, and then based on the final reward value, the optimal distributed storage strategy is determined through the reinforcement learning algorithm, and finally the energy storage data of each battery per minute is stored according to the optimal distributed storage strategy. Since the fit reward value when performing each storage action on the energy storage data of each battery per minute and the spatial distance between each battery and each storage node are comprehensively considered in the process of determining the optimal distributed storage strategy, the relevant energy storage data of the same battery can be effectively aggregated into closer storage nodes, so that the subsequent data access and retrieval efficiency are significantly improved. BRIEF DESCRIPTION OF THE DRAWINGS

[0055] In order to more clearly illustrate the technical solutions and advantages in the embodiments of the present invention or the prior art, the drawings required for use in the embodiments or the prior art descriptions are briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.

[0056] Figure 1 A schematic diagram of a method flow of a novel distributed storage method for energy storage data of an intelligent energy storage station provided by an embodiment of the present invention;

[0057] Figure 2 A scatter plot corresponding to the current data collected in the time series for a certain battery;

[0058] Figure 3 A scatter plot of the temperature data collected over a certain battery time series;

[0059] Figure 4 A schematic diagram of the system structure of a new type of intelligent energy storage station energy storage data distributed storage system provided by an embodiment of the present invention. DETAILED DESCRIPTION

[0060] In order to further explain the technical means and effects adopted by the present invention to achieve the predetermined invention purpose, the following is a detailed description of the specific implementation method, structure, features and effects of a new type of intelligent energy storage station energy storage data distributed storage method and system proposed by the present invention in combination with the accompanying drawings and preferred embodiments. In the following description, different "one embodiment" or "another embodiment" does not necessarily refer to the same embodiment. In addition, specific features, structures or characteristics in one or more embodiments may be combined in any suitable form.

[0061] Unless defined otherwise, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this invention belongs.

[0062] The specific scheme of a novel intelligent energy storage station energy storage data distributed storage method and system provided by the present invention is described in detail below with reference to the accompanying drawings.

[0063] See also Figure 1 , which shows a method flow chart of a novel distributed storage method for energy storage data of an intelligent energy storage station provided by an embodiment of the present invention, such as Figure 1 As shown, the above-mentioned novel intelligent energy storage station energy storage data distributed storage method specifically includes the following steps:

[0064] Step 110: Determine the energy storage data of each battery in the new intelligent energy storage station every minute.

[0065] In this embodiment, it is assumed that there are N batteries in the new intelligent energy storage station. First, the specific location of each battery is collected. Then, in order to calculate the energy storage data generated by each battery every minute, a current sensor with a collection frequency of 1Hz needs to be installed at the interface where each battery is connected to the external circuit, and a temperature sensor with a collection frequency of 1Hz needs to be installed on the surface of the battery. The current data of the battery is collected by the current sensor, and the temperature data is collected by the temperature sensor, so as to realize the collection of battery status data, and thus provide an effective data basis for the determination of energy storage data.

[0066] Figure 2 and Figure 3 The changes of current data and temperature data in a battery collection sequence are shown as examples. By analyzing the current data and temperature data related to the battery operation, the statistics of the battery charge change over a period of time can be realized, thereby providing an effective data source for determining energy storage data.

[0067] It should be noted that the energy storage data mentioned in this embodiment mainly refers to the battery SOC data and other data that can characterize the battery energy storage state.

[0068] Step 120: Determine the compatibility reward value when each storage action is performed on the energy storage data of each battery per minute.

[0069] It can be understood that the compatibility reward value can represent the reward of the corresponding storage action when the energy storage data of the current battery is placed in a certain storage node. The higher the compatibility reward value, the more the energy storage data matches the corresponding storage node, that is, the more reasonable the storage action is, and the subsequent data retrieval and access links are more efficient.

[0070] Step 130: Based on the predetermined spatial distance between each battery and each storage node, the compatibility reward value is weighted to obtain a final reward value for storing the energy storage data of each battery per minute in each storage node.

[0071] It is understandable that the final reward value can represent the reward situation of putting the current energy storage data into a certain storage node from the perspective of data relevance and spatial distance rationality. The higher the final reward value, the more reasonable it is to put the current energy storage data into the storage node from the perspective of data relevance and spatial distance rationality.

[0072] This embodiment uses the spatial distance between the battery and the storage node as the basis for determining the weighting coefficient. By storing in a storage node closer to the data source, the communication volume with the remote server can be reduced, thereby avoiding the problem of response lag caused by network delays, and further improving the rationality of energy storage data storage.

[0073] Step 140: Based on the final reward value, determine the optimal distributed storage strategy through a reinforcement learning algorithm.

[0074] In practical applications, through reinforcement learning algorithms, the storage node with the highest final reward value can be determined for the energy storage data to be stored, thereby obtaining the optimal distributed storage strategy.

[0075] Step 150: Store the energy storage data of each battery every minute according to the optimal distributed storage strategy.

[0076] The solution provided in this embodiment first determines the compatibility reward value of each battery's energy storage data per minute when executing each storage action, and finally obtains the final reward value of each battery's storage data per minute when executing each storage action through the battery's spatial position. In this way, the relevant energy storage data of the same battery can be effectively aggregated into closer storage nodes, so that higher efficiency is presented in subsequent data access and retrieval.

[0077] In one embodiment, determining the energy storage data of each battery in the new intelligent energy storage station every minute specifically includes:

[0078] First, a cycle number-battery rated capacity model is constructed.

[0079] In a specific implementation, a cycle number-battery rated capacity model is constructed, specifically including:

[0080] In the first step, an initial model including unknown parameters is constructed.

[0081] In actual applications, during long-term use, the battery will gradually experience a capacity decay process due to factors such as internal chemical reactions, decomposition of the electrolyte, structural changes in electrode materials, and battery crystallization. These decay behaviors are mainly caused by the battery's charge and discharge process. In particular, lithium batteries and other common types of batteries usually show slow initial decay during long-term use, but as the number of cycles increases, the decay rate will gradually accelerate, and eventually tend to exponential decay characteristics. This is because as the number of charge and discharge cycles increases, the internal losses of the battery gradually accumulate, resulting in a nonlinear decay process.

[0082] Based on the above degradation law, the following initial model containing unknown parameters can be constructed to describe the change of rated capacity of the battery during use. The initial model is as follows:

[0083] (1)

[0084] in, Indicates the number of cycles, It represents the theoretical rated capacity of the battery at the cycle number N. All represent unknown parameters.

[0085] The second step is to determine the true availability of each charging cycle.

[0086] In this embodiment, determining the actual availability of each charging cycle specifically includes:

[0087] First, determine the turning point and temperature data for each charging cycle.

[0088] In this embodiment, the turning point can be the charging start time and the charging end time, which can be determined according to the change of the charging current value during the charging cycle. The temperature data can be collected by a temperature sensor installed on the battery.

[0089] Specifically, the temperature data may include temperature values ​​at different charging moments in each charging cycle, an average temperature value, and an average temperature change.

[0090] Then, based on the turning point of each charging cycle, the length of each charging cycle is determined.

[0091] It can be understood that the length of each charging cycle can be determined by subtracting the charging end time from the charging start time during a charging cycle.

[0092] Finally, based on the turning point, time length and temperature data of each charging cycle, the actual availability of each charging cycle is calculated.

[0093] It should be noted that if the charge and discharge process is too long within one charging cycle, that is, the length of each charging cycle is long, it indicates that the battery is under high load for a long time, which may lead to incomplete chemical reactions inside the battery. Therefore, the shorter the charge and discharge process, the more useful it is. On this basis, if the temperature changes greatly during the charge and discharge process, it indicates that the material inside the battery may undergo microstructural changes due to thermal expansion and contraction, which may lead to performance failures. Therefore, the first The actual availability of charge cycles can be expressed as follows:

[0094] (2)

[0095] in, Indicates The actual availability of charging cycles, represents the exponential function with base e, Indicates At the end of the charging cycle, Indicates The starting time of the charging cycle, Indicates In the first charging cycle The temperature value at the moment, Indicates The average temperature value within the charging cycle, Indicates The average temperature change within a charging cycle, Indicates The length of time for a charging cycle, Represents a hyperparameter.

[0096] At this point, the true usability of the model of each charging cycle versus cycle number-battery rated capacity can be obtained.

[0097] In the third step, a data error sum-of-squares function is established based on the initial model and the actual availability of each charging cycle.

[0098] This embodiment mainly uses the least squares method to determine the two unknown parameters in the initial model based on the sum of square errors of the battery rated capacity under all historical charging cycles. However, if a charging cycle with a low real availability appears, the credibility of the unknown parameters will be extremely low. Therefore, it is necessary to make the charging cycle with a high real availability contribute more to the error sum of squares, and the charging cycle with a low real availability contribute less to the error sum of squares. Based on this, the following data error sum of squares function can be constructed:

[0099] (3)

[0100] in, represents the sum of squared errors of all data points after correction, Indicates the number of predetermined charging cycles, Indicates The actual availability of charging cycles, Indicates The actual battery capacity of the charge cycle, Indicates the theoretical rated capacity of the battery after correction.

[0101] The fourth step is to determine the optimal solution for the unknown parameters in the initial model based on the data error sum of squares function.

[0102] Based on the constructed data error sum of square function, the unknown parameters in the above data error sum of square function can be obtained using the least squares method. The optimal solution of .

[0103] The fifth step is to construct a cycle number-battery rated capacity model based on the initial model and the optimal solution of the unknown parameters.

[0104] It can be understood that by substituting the optimal solution of the unknown parameters into the initial model that has been constructed, a cycle number-battery rated capacity model can be obtained.

[0105] Then, based on the cycle number-battery rated capacity model, the theoretical rated capacity of each battery in the new intelligent energy storage station in each cycle is determined.

[0106] Finally, the energy storage data of each battery per minute is determined based on the theoretical rated capacity of each battery at each cycle.

[0107] This embodiment first analyzes the collected current data to determine the time length corresponding to each charging cycle, and then calculates the availability of each historical charging cycle to the cycle number-battery rated capacity model through the temperature change under the corresponding time length, so as to construct the cycle number-battery rated capacity model, so as to obtain the theoretical rated capacity corresponding to each charging cycle. Compared with the traditional solution of using the rated capacity of the battery as a fixed value to calculate the energy storage data, this embodiment can realize the correction of the energy storage data of each battery every minute.

[0108] In one embodiment, determining the turning point of each charging cycle specifically includes:

[0109] The first step is to obtain the charging current value corresponding to each charging moment during each charging cycle.

[0110] In this embodiment, the charging current value can be collected by a pre-installed current sensor.

[0111] In the second step, for any charging moment, based on the charging current values ​​of a preset number of adjacent moments before the charging moment and the charging current values ​​of a preset number of adjacent moments after the charging moment, the probability value of each charging moment being a turning point is calculated.

[0112] In this embodiment, the preset number of adjacent time points can be reasonably set according to actual computing requirements. For example, the preset number can be set to 10.

[0113] In practical applications, the charging current value of the battery during the charging process is used as the basis for judging whether the battery is fully charged. When the battery is close to being fully charged, the voltage difference inside the battery will decrease, causing the charging current value to gradually decrease. When the charging current value drops to a certain preset threshold, it can usually be considered that the battery is fully charged. For lithium batteries, when the charging current value is reduced to a set trickle current (for example, 10% of the battery capacity), the battery can be considered to be fully charged. Therefore, the first The probability value of a charging moment belonging to the turning point of the end of a charge can be expressed as follows:

[0114] (4)

[0115] in, Indicates The probability value of a charging moment being the turning point of the end of a charge, represents the exponential function with base e, Indicates The charging current value at each charging moment, Indicates The sum of the charging current values ​​of the 10 adjacent moments after the charging moment, Indicates The charging current value at each charging moment, Indicates The sum of the charging current values ​​of the 10 adjacent moments before the charging moment.

[0116] The third step is to determine the turning point of each charging cycle based on the probability value.

[0117] In a specific implementation, determining the turning point of each charging cycle based on the probability value includes:

[0118] First, the target moments with probability values ​​higher than the preset probability threshold are extracted to obtain the initial screening results.

[0119] In this embodiment, the preset probability threshold can be reasonably set according to actual computing requirements. For example, the preset probability threshold can be set to 0.8. At this time, the target moment with a probability value higher than 0.8 can be preliminarily determined as the turning point.

[0120] Then, for any target moment in the initial screening result, it is determined whether there is an abnormal time point with a probability value higher than a preset probability threshold within a preset period before the target moment to obtain a determination result.

[0121] Considering that there may still be some interference points near the turning point after the initial screening, the present embodiment sets a secondary screening link to determine whether there is an abnormal time point with a probability value higher than a preset probability threshold within a preset period before the target moment. If there is an abnormal time point, the target moment is eliminated. In practical applications, the preset period can be set to ten seconds, that is, if there is an abnormal time point with a probability value greater than 0.8 within ten seconds before the target moment, the target moment can be eliminated.

[0122] Finally, the target moment when the judgment result in the initial screening is negative is used as the turning point of each charging cycle.

[0123] After two screenings, the time points corresponding to all turning points can be obtained , ,… at this time Represents one charging cycle, Indicates another charging cycle.

[0124] In one embodiment, based on the theoretical rated capacity of each battery in each cycle, the energy storage data of each battery per minute is determined, specifically including:

[0125] The first step is to obtain the charge change of each battery from the previous minute to the current minute, and obtain the energy storage data of each battery in the previous minute.

[0126] In the second step, the charge change is divided by the theoretical rated capacity of each battery in each cycle to obtain the correction value.

[0127] The third step is to sum the energy storage data of the previous minute with the correction value to calculate the energy storage data of each battery in the energy storage station per minute.

[0128] In this embodiment, the energy storage data of the battery at the ath minute is determined as an example. Assume that at the ath minute, The theoretical rated capacity of the battery at the ath minute can be obtained through the cycle number-battery rated capacity model. So far, the following formula can be constructed to represent the energy storage data of the battery at the ath minute:

[0129] (5)

[0130] in, Represents the energy storage data of the battery at minute a, Indicates the energy storage data of the battery at minute a-1, It represents the change in battery charge from minute a-1 to minute a. Indicates that the battery is performing the first Theoretical rated capacity at the time of charge cycle.

[0131] In one embodiment, determining the compatibility reward value when performing each storage action on the energy storage data of each battery per minute specifically includes:

[0132] The first step is to subtract the energy storage data of each battery per minute from the preset value to obtain the power difference.

[0133] In the second step, the power difference is input into the preset maximum and minimum value normalization function to calculate the data attention of each battery per minute.

[0134] In actual applications, overcharge may cause problems such as battery heating, battery expansion or leakage, battery material damage, and shortened battery life. Over-discharge may cause problems such as battery capacity reduction, internal battery short circuit, battery failure to charge, and significantly shortened battery life. Therefore, more attention should be paid to the energy storage data of the battery every minute in the overcharge or over-discharge state. In other words, the farther the battery's energy storage data is from 50% (i.e., the preset value), the more attention it deserves. Therefore, the following formula can be constructed to represent the data attention of the energy storage data of the Ath battery in the ath minute:

[0135] (6)

[0136] in, Indicates the data attention of the energy storage data of the Ath battery at the ath minute, represents the normalized function of the maximum and minimum values ​​(compared with the energy storage data obtained from all historical minutes of the Ath battery), Represents the energy storage data of the Ath battery at the ath minute.

[0137] Similarly, the data attention of each battery per minute can be determined.

[0138] The third step is to obtain the historical data of the battery in each storage node for each battery.

[0139] The fourth step is to calculate the compatibility reward value for each storage action performed on the energy storage data of each battery every minute based on data attention and historical data.

[0140] In practical applications, for the energy storage data of each battery per minute, storing it together with the data with higher compatibility about this battery in a storage node can greatly reduce the time required for retrieval. When it is necessary to query the status of a specific battery, centrally storing data with high compatibility can obtain multiple related information through a single data access, thereby reducing the number and delay of I / O operations and facilitating unified analysis and processing. Data mining, trend prediction and other tasks can be performed based on the energy storage data of a specific battery in the same storage node.

[0141] However, for all historical data of this battery stored in the same storage node, the data contribution at different times is different. In order to make the calculation result more timely, so that the closer the time is, the higher the contribution is, this embodiment constructs the following formula to represent the compatibility reward value of the energy storage data of the Ath battery obtained at the ath minute and the historical data of the Ath battery stored in the Qth storage node:

[0142] (7)

[0143] in, It represents the reward value of the compatibility between the energy storage data of the Ath battery obtained at the ath minute and the historical data of the Ath battery stored in the Qth storage node. represents the exponential function with base e, represents the amount of historical data of the Ath battery stored in the Qth storage node, represents the wth corresponding time of the historical data of the Ath battery stored in the Qth storage node, It represents the data attention of the energy storage data of the Ath battery obtained at the ath minute. It represents the data attention of the historical data of the Ath battery stored in the Qth storage node at the wth corresponding time.

[0144] Similarly, the compatibility reward value for each storage action performed on the energy storage data of each battery per minute can be obtained.

[0145] It is understandable that based on the fit reward value when performing each storage action on the energy storage data of each battery every minute, since energy storage stations may usually be distributed in remote areas, if all energy storage data is transmitted to a remote data center, it will be limited by factors such as network delays and bandwidth bottlenecks. Therefore, storing it in a storage node closer to the data source can reduce the communication volume with the remote server and avoid response delays caused by network delays.

[0146] Therefore, this embodiment uses spatial distance to perform weighted processing on the obtained compatibility reward value, and specifically constructs the following formula to represent the final reward value of putting the energy storage data of the Ath battery obtained at the ath minute into the Qth storage node:

[0147] (8)

[0148] in, It represents the final reward value of putting the energy storage data of the Ath battery in the ath minute into the Qth storage node. represents the spatial distance between the Ath battery and the Qth storage node, represents the maximum and minimum normalized function, It represents the reward value of the compatibility between the energy storage data of the Ath battery obtained at the ath minute and the historical data of the Ath battery stored in the Qth storage node.

[0149] At this point, the final reward value for performing each storage action on the energy storage data of each battery every minute can be obtained.

[0150] In one embodiment, based on the final reward value, an optimal distributed storage strategy is determined by a reinforcement learning algorithm, specifically including:

[0151] The first step is to determine the state space based on the energy storage data of the battery at different times and the data status of each storage node.

[0152] In the second step, the action space is determined based on storing the energy storage data of the battery at different times to any storage node.

[0153] The third step is to determine the storage node with the highest final reward value for each battery’s energy storage data per minute in the state space and action space through a reinforcement learning algorithm to obtain the optimal distributed storage strategy.

[0154] In this embodiment, the reinforcement learning algorithm can adopt the Q learning algorithm. Since the main purpose of this embodiment is to select the optimal storage node for the energy storage data of each battery per minute through the Q learning algorithm, it is necessary to define the state space and action space first. Among them, the state space is defined as the current energy storage data of the battery and the data state of each storage node. The action space is defined as the selection of which storage node to store the energy storage data of the current battery. Assuming there are M storage nodes, the size of the action space of each battery per minute is M, that is, one storage node is selected to store the energy storage data of the battery.

[0155] In a specific implementation, in the state space and action space, the storage node with the highest final reward value is determined by the reinforcement learning algorithm for the energy storage data of each battery every minute, and the optimal distributed storage strategy is obtained, which specifically includes:

[0156] The first step is to initialize the Q table: initialize the Q value for each pair of state-action data, usually the initialized Q value is zero.

[0157] The second step is to select an action: At each time step, the energy storage data of each battery per minute uses the ε-greedy strategy to select a storage action according to the current state, that is, to determine a storage node.

[0158] The third step is to execute the action: after storing the battery’s energy storage data in the selected storage node, transfer to the next state.

[0159] Step 4: Update the Q value: Update the Q value of the current state-action data pair based on the reward value obtained by the determined execution action and the maximum Q value of the next state.

[0160] In the fifth step, the training process is repeated until the Q value converges, and the optimal distributed storage strategy is extracted from the Q table, that is, the storage action with the highest Q value is selected for each state.

[0161] At this point, the Q-learning algorithm maximizes the final reward value by continuously adjusting the action selection (i.e., which storage node to store the energy storage data of each battery), thereby achieving the optimal distributed storage strategy.

[0162] Based on the same general inventive concept, the present invention also protects a novel distributed storage system for energy storage data of an intelligent energy storage station. The novel distributed storage system for energy storage data of an intelligent energy storage station provided by the present invention is described below. The novel distributed storage system for energy storage data of an intelligent energy storage station described below and the novel distributed storage method for energy storage data of an intelligent energy storage station described above can be referred to each other.

[0163] See also Figure 4 , which shows a system structure diagram of a new type of intelligent energy storage station energy storage data distributed storage system provided by an embodiment of the present invention, such as Figure 4 As shown, the above-mentioned new intelligent energy storage station energy storage data distributed storage system specifically includes:

[0164] The determination module 210 is used to determine the energy storage data of each battery in the new intelligent energy storage station every minute.

[0165] The calculation module 220 is used to determine the compatibility reward value when each storage action is performed on the energy storage data of each battery per minute.

[0166] The weighting module 230 is used to perform weighted processing on the compatibility reward value based on the predetermined spatial distance between each battery and each storage node, so as to obtain the final reward value for storing the energy storage data of each battery per minute in each storage node.

[0167] The optimization module 240 is used to determine the optimal distributed storage strategy based on the final reward value through a reinforcement learning algorithm.

[0168] The storage module 250 is used to store the energy storage data of each battery every minute according to the optimal distributed storage strategy.

[0169] The novel intelligent energy storage station energy storage data distributed storage system provided by the embodiment of the present invention can effectively aggregate the relevant energy storage data of the same battery into closer storage nodes, because it comprehensively considers the compatibility reward value of each battery's energy storage data per minute when executing each storage action, as well as the spatial distance between each battery and each storage node in the process of determining the optimal distributed storage strategy, so that the subsequent data access and retrieval efficiency are significantly improved.

[0170] Regarding the system in the above embodiment, the specific manner in which each module performs operations has been described in detail in the embodiment of the method, and will not be elaborated again here.

[0171] It should be noted that the sequence of the above embodiments of the present invention is only for description and does not represent the advantages and disadvantages of the embodiments. The processes depicted in the accompanying drawings do not necessarily require the specific order or continuous order shown to achieve the desired results. In some embodiments, multitasking and parallel processing are also possible or may be advantageous.

[0172] The various embodiments in this specification are described in a progressive manner, and the same or similar parts between the various embodiments can be referenced to each other, and each embodiment focuses on the differences from other embodiments.

Claims

1. A novel distributed storage method for energy storage data of an intelligent energy storage station, characterized in that: The method comprises: Determine the energy storage data of each battery in the new smart energy storage station every minute; Determine a compatibility reward value when each storage action is performed on the energy storage data of each battery per minute; Based on the predetermined spatial distance between each battery and each storage node, the compatibility reward value is weighted to obtain a final reward value for storing the energy storage data of each battery per minute in each storage node; Based on the final reward value, determining an optimal distributed storage strategy through a reinforcement learning algorithm; Storing the energy storage data of each battery every minute according to the optimal distributed storage strategy; Determining a compatibility reward value when performing each storage action on the energy storage data of each battery per minute, including: Subtract the energy storage data of each battery per minute from the preset value to obtain the power difference; Input the power difference into a preset maximum and minimum value normalization function to calculate the data attention of each battery per minute; For each battery, obtaining historical data of the battery in each storage node; Based on the data attention and the historical data, a compatibility reward value when each storage action is performed on the energy storage data of each battery per minute is calculated; The determination of the energy storage data of each battery per minute in the new intelligent energy storage station includes: Construct a cycle number-battery rated capacity model; Based on the cycle number-battery rated capacity model, determining the theoretical rated capacity of each battery in the new intelligent energy storage station at each cycle; Determine the energy storage data of each battery per minute based on the theoretical rated capacity of each battery in each cycle; The construction of the cycle number-battery rated capacity model includes: Construct an initial model including unknown parameters; Determine the true usability of each charging cycle; Establishing a data error square sum function based on the initial model and the actual availability of each charging cycle; Determining the optimal solution of the unknown parameters in the initial model based on the data error square sum function; Based on the initial model and the optimal solution of the unknown parameters, a cycle number-battery rated capacity model is constructed; Determining the actual availability of each charging cycle includes: Determine the turning point and temperature data of each charging cycle; Determining the duration of each charging cycle based on the turning point of each charging cycle; Based on the turning point, the time length and the temperature data of each charging cycle, the actual availability of each charging cycle is calculated.

2. According to claim 1, a novel distributed storage method for energy storage data of an intelligent energy storage station is characterized in that: Identify the turning points in each charging cycle, including: Obtain the charging current value corresponding to each charging moment during each charging cycle; For any charging moment, based on the charging current values ​​of a preset number of adjacent moments before the charging moment and the charging current values ​​of a preset number of adjacent moments after the charging moment, a probability value of the charging moment being a turning point is calculated; Based on the probability value, a turning point of each charging cycle is determined.

3. A novel distributed storage method for energy storage data of an intelligent energy storage station according to claim 2, characterized in that: Based on the probability value, determining the turning point of each charging cycle includes: Extract the target moment when the probability value is higher than the preset probability threshold to obtain the initial screening result; For any target time in the preliminary screening result, determine whether there is an abnormal time point with a probability value higher than a preset probability threshold within a preset period before the target time, and obtain a determination result; The target time when the judgment result in the preliminary screening result is negative is used as the turning point of each charging cycle.

4. A novel distributed storage method for energy storage data of an intelligent energy storage station according to claim 1, characterized in that: Based on the theoretical rated capacity of each battery in each cycle, the energy storage data of each battery per minute is determined, including: Obtain the charge change of each battery from the previous minute to the current minute, and obtain the energy storage data of each battery in the previous minute; Divide the charge change by the theoretical rated capacity of each battery in each cycle to obtain a correction amount; The energy storage data of the previous minute is summed with the correction amount to calculate the energy storage data of each battery in the energy storage station per minute.

5. A novel distributed storage method for energy storage data of an intelligent energy storage station according to claim 1, characterized in that: Based on the final reward value, an optimal distributed storage strategy is determined by a reinforcement learning algorithm, including: Determine the state space based on the energy storage data of the battery at different times and the data status of each storage node; Determine the action space based on storing the energy storage data of the battery at different times to any storage node; In the state space and the action space, the storage node with the highest final reward value is determined for the energy storage data of each battery per minute through a reinforcement learning algorithm to obtain an optimal distributed storage strategy.

6. A new type of intelligent energy storage station energy storage data distributed storage system, characterized in that: The system comprises: A determination module, used to determine the energy storage data of each battery in the new intelligent energy storage station every minute; A calculation module, used for determining a compatibility reward value when each storage action is performed on the energy storage data of each battery per minute; Determining a compatibility reward value when performing each storage action on the energy storage data of each battery per minute, including: Subtract the energy storage data of each battery per minute from the preset value to obtain the power difference; Input the power difference into a preset maximum and minimum value normalization function to calculate the data attention of each battery per minute; For each battery, obtaining historical data of the battery in each storage node; Based on the data attention and the historical data, a compatibility reward value when each storage action is performed on the energy storage data of each battery per minute is calculated; The determination of the energy storage data of each battery per minute in the new intelligent energy storage station includes: Construct a cycle number-battery rated capacity model; Based on the cycle number-battery rated capacity model, determining the theoretical rated capacity of each battery in the new intelligent energy storage station at each cycle; Determine the energy storage data of each battery per minute based on the theoretical rated capacity of each battery in each cycle; The construction of the cycle number-battery rated capacity model includes: Construct an initial model including unknown parameters; Determine the true usability of each charging cycle; Establishing a data error square sum function based on the initial model and the actual availability of each charging cycle; Determining the optimal solution of the unknown parameters in the initial model based on the data error square sum function; Based on the initial model and the optimal solution of the unknown parameters, a cycle number-battery rated capacity model is constructed; Determining the actual availability of each charging cycle includes: Determine the turning point and temperature data of each charging cycle; Determining the duration of each charging cycle based on the turning point of each charging cycle; Based on the turning point, the time length and the temperature data of each charging cycle, the actual availability of each charging cycle is calculated; A weighting module, configured to perform weighted processing on the compatibility reward value based on a predetermined spatial distance between each battery and each storage node, to obtain a final reward value for storing the energy storage data of each battery per minute in each storage node; An optimization module, used for determining an optimal distributed storage strategy through a reinforcement learning algorithm based on the final reward value; A storage module is used to store the energy storage data of each battery every minute according to the optimal distributed storage strategy.

Images