Energy storage system operation state classification method

Abstract

The invention discloses an energy storage system operation state classification method, and relates to the technical field of energy storage batteries, and the method comprises the steps: collecting the performance data and operation data of the energy storage batteries; performing cleaning and normalization processing on the collected data; extracting preprocessed data features, and constructing feature vectors; defining a state space and an execution space; designing a reward mechanism, and calculating an expected reward; calculating a value function and a Q value; continuously updating the Q value according to the learned instant reward and future expected reward; estimating a Q value by using a neural network; updating parameters of the actor network based on the gradient of the Q value; updating parameters of the target network; dynamically adjusting the capacity of the experience pool; a priority sampling strategy is adopted for training; and generating and verifying a classification model, and classifying the energy storage batteries. According to the energy storage system operation state classification method provided by the invention, the accuracy and reliability of energy storage system operation state classification are improved, and meanwhile, the service life and the performance of the battery are optimized.

Description

Technical Field

[0001] The present invention relates to the technical field of energy storage batteries, and in particular to a method for classifying operating states of energy storage systems. Background Art

[0002] With the development of renewable energy and the popularity of electric vehicles, energy storage batteries are playing an increasingly important role in energy management and power systems. However, the state of energy storage batteries is complex and changeable, and is affected by multiple factors, such as voltage, capacity, internal resistance, temperature, cycle life, etc. These factors not only affect each other, but may also cause the battery's performance to gradually degrade, increase maintenance costs, and even affect the overall stability of the system.

[0003] In the existing technology, energy storage battery status monitoring mainly relies on simple physical parameter measurement and rule-based evaluation methods, which cannot effectively deal with the complexity and dynamics of battery status. In addition, these traditional methods usually rely on experience and subjective judgment, lack scientificity and accuracy, and are difficult to meet the needs of modern power systems for precise management of energy storage batteries. Summary of the invention

[0004] In view of the above-mentioned problems, the present invention is proposed.

[0005] Therefore, the technical problem solved by the present invention is: the existing energy storage battery status monitoring method has insufficient data processing capability and low classification accuracy, and how to effectively deal with the complexity and dynamic optimization of the battery status.

[0006] To solve the above technical problems, the present invention provides the following technical solutions: a method for classifying the operating status of an energy storage system, comprising: collecting performance data and operation data of an energy storage battery; cleaning and normalizing the collected data; extracting preprocessed data features and constructing a feature vector; defining a state space and an execution space; designing a reward mechanism and calculating expected rewards; calculating a value function and a Q value; continuously updating the Q value based on learned immediate rewards and future expected rewards; estimating the Q value using a neural network; updating the parameters of an actor network based on the gradient of the Q value; updating the parameters of a target network; dynamically adjusting the capacity of an experience pool; training using a priority sampling strategy; inputting the preprocessed feature vector into a deep reinforcement learning model; generating and verifying a classification model to classify energy storage batteries.

[0007] As a preferred solution of the energy storage system operating status classification method described in the present invention, wherein: the performance data includes voltage, capacity, internal resistance, temperature and cycle life;

[0008] The operation data includes charging time, discharging time, operation mode, load condition and ambient temperature.

[0009] As a preferred solution of the energy storage system operation status classification method described in the present invention, wherein: the data features include the slope of the voltage-capacity curve, the temperature-internal resistance change rate, the cycle life decay rate, the load adaptability characteristics, the influence of the ambient temperature on the charge-discharge efficiency, the comprehensive health index and the charge-discharge efficiency ratio;

[0010] The slope of the voltage-capacity curve is expressed as,

[0011]

[0012] Among them, V i represents the voltage value at the i-th time point, C i represents the capacity value at the i-th time point, Represents the average value of all voltage values, represents the average of all capacity values, and n represents the total number of time points measured;

[0013] The temperature-internal resistance change rate is expressed as,

[0014]

[0015] Wherein, R1 represents the internal resistance value measured at temperature T1, R2 represents the internal resistance value measured at temperature T2, T1 represents the first ambient temperature, and T2 represents the second ambient temperature;

[0016] The cycle life decay rate is expressed as,

[0017]

[0018] Among them, C N represents the capacity at the Nth cycle, C0 represents the initial capacity, and N represents the total number of cycles;

[0019] The load adaptability characteristic is expressed as,

[0020]

[0021] Among them, T dis Indicates the discharge time, L indicates the load condition;

[0022] The effect of the ambient temperature on the charge and discharge efficiency is expressed as:

[0023]

[0024] Among them, T chg Indicates the charging time at a specific ambient temperature, T dis Indicates the discharge time at a specific ambient temperature, T amb Indicates the ambient temperature;

[0025] The comprehensive health index is expressed as,

[0026]

[0027] Among them, V represents the battery voltage, R represents the battery internal resistance, C represents the current capacity of the battery, and C rated Indicates the rated capacity of the battery, N life Indicates the current number of cycles, N rated represents the rated number of cycles, α, β, γ, δ represent weight coefficients;

[0028] The charge-discharge efficiency ratio is expressed as,

[0029]

[0030] Among them, T dis,n represents the discharge time in operation mode n, T chg,n Indicates the charging time in operating mode n.

[0031] As a preferred solution of the energy storage system operating state classification method described in the present invention, the state space is represented as:

[0032]

[0033] The execution space includes adjusting the charging mode, adjusting the load current and adjusting the load power;

[0034] The expected reward is expressed as,

[0035]

[0036] Where R represents the expected reward, r(s,a) represents the immediate reward for performing operation a in state s, γ represents the decay coefficient, ranging from 0 to 1, representing the discount of future rewards, P(s′|s,a) represents the state transition probability, which represents the probability of transferring to the new state s′ after performing operation a in state s, and V(s) represents the value function of the new state s′;

[0037] The value function is expressed as,

[0038]

[0039] Among them, V(s) represents the value function of state s, which represents the maximum Q value that can be obtained in state s. Q(s,a) represents the value function of the state-action pair, which represents the long-term reward of performing action a in state s;

[0040] The Q value is expressed as,

[0041]

[0042] Among them, Q(s,a) represents the value function of the state-action pair, r(s,a) represents the immediate reward, γ represents the decay coefficient, P(s′|s,a) represents the state transition probability, and V(s′) represents the value function of the new state.

[0043] As a preferred solution of the energy storage system operating state classification method described in the present invention, the updated Q value is expressed as:

[0044]

[0045] Among them, Q(s,a) represents the value function of the state-action pair, α represents the learning rate, which controls the speed of Q value update, r represents the immediate reward, γ represents the decay coefficient, and Q(s′,a′) represents the maximum Q value in the new state;

[0046] The Q value estimated by neural network is expressed as:

[0047]

[0048] in, represents the Q value estimated by the neural network, θ represents the parameters of the neural network, Q(s,a) represents the actual Q value, min θ Indicates optimizing network parameters by minimizing the error;

[0049] The parameter update of the actor network based on the Q-value gradient is expressed as,

[0050]

[0051] Among them, θ actor represents the parameters of the actor network, α represents the learning rate, represents the gradient of Q value with respect to the actor network parameters;

[0052] The parameters of the updated targeting network are expressed as,

[0053] θ′=τθ+(1-τ)θ′

[0054] Among them, θ′ represents the parameters of the target network, τ represents the parameters that control the update speed, and θ represents the parameters of the current network.

[0055] As a preferred solution of the energy storage system operating status classification method described in the present invention, the dynamically adjusted experience pool capacity is expressed as:

[0056] T(n)=T(1)×n -k

[0057]

[0058] η=α×S

[0059] β=1-α

[0060] N=η×i

[0061] Among them, T(n) represents the time period of the nth event, T(1) represents the initial time period, n represents the number of current training times, k represents the parameter of learning effect, which is used to control the change of learning curve, α represents the change of learning rate, which reflects the improvement of learning efficiency, η represents learning efficiency, S represents the sample size in the experience pool, β represents the supplementary parameter, which is used to adjust the capacity of experience pool and learning efficiency, N represents the capacity of experience pool, and i represents the number of training times of the algorithm.

[0062] As a preferred solution of the energy storage system operating status classification method described in the present invention, the priority sampling strategy is expressed as:

[0063]

[0064] Among them, Δ represents the temporal difference error, which measures the importance of the experience data, r(s,a) represents the immediate reward, γ represents the attenuation coefficient, Q(s,a) represents the Q value of performing operation a in the current state s, Q(s′,a′) represents the Q value of performing operation a′ in the new state s′, P(sample) represents the probability that the experience data is sampled, Δ: temporal difference error, α represents the priority parameter, which controls the influence of TD value on sampling probability, π represents the priority parameter, which is used to calculate the priority of each experience data, ∈ represents a small positive value to prevent the TD value from being zero, and γ represents the weight parameter, which is used to control the distribution of the priority parameter.

[0065] As a preferred solution of the energy storage system operating state classification method described in the present invention, wherein: the classification of the energy storage battery includes generating a classification model through a trained deep reinforcement learning model, and classifying the state of the energy storage battery based on the input feature vector;

[0066] After the model is trained, an independent validation dataset is used to test the classification model to verify the classification accuracy and generalization ability of the model on the validation dataset.

[0067] Further adjust model parameters or training strategies based on the test results of the validation set;

[0068] The trained and validated classification model is used to classify the energy storage battery to generate specific classification labels, which include healthy and efficient operation, degraded and inefficient operation, high temperature risk, and excessive internal resistance.

[0069] A computer device comprises a memory and a processor, wherein the memory stores a computer program, and the processor implements the steps of the above-mentioned energy storage system operating status classification method when executing the computer program.

[0070] A computer-readable storage medium stores a computer program, which, when executed by a processor, implements the steps of the energy storage system operating status classification method as described above.

[0071] Beneficial effects of the invention: The energy storage system operating status classification method provided by the invention improves the accuracy and reliability of energy storage system operating status classification by introducing a deep reinforcement learning algorithm, while optimizing the battery life and performance. It reduces maintenance costs and operational risks, adapts to a variety of application scenarios, and realizes intelligent management and optimization of energy storage batteries. BRIEF DESCRIPTION OF THE DRAWINGS

[0072] In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the accompanying drawings required for use in the description of the embodiments will be briefly introduced below. Obviously, the accompanying drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other accompanying drawings can be obtained based on these accompanying drawings without paying creative work.

[0073] Figure 1 An overall flow chart of a method for classifying operating states of an energy storage system provided by an embodiment of the present invention. DETAILED DESCRIPTION

[0074] In order to make the above-mentioned purposes, features and advantages of the present invention more obvious and easy to understand, the specific implementation methods of the present invention are described in detail below in conjunction with the drawings of the specification. Obviously, the described embodiments are part of the embodiments of the present invention, but not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary persons in the art without creative work should fall within the scope of protection of the present invention.

[0075] In the following description, many specific details are set forth to facilitate a full understanding of the present invention, but the present invention may also be implemented in other ways different from those described herein, and those skilled in the art may make similar generalizations without violating the connotation of the present invention. Therefore, the present invention is not limited to the specific embodiments disclosed below.

[0076] Example 1

[0077] Reference Figure 1 , as an embodiment of the present invention, provides a method for classifying the operating status of an energy storage system, comprising:

[0078] Collect performance and operation data of energy storage batteries;

[0079] Clean and normalize the collected data;

[0080] Extract the preprocessed data features and construct feature vectors;

[0081] Define state space and execution space;

[0082] Design reward mechanisms and calculate expected rewards;

[0083] Calculate the value function and Q value;

[0084] Continuously update the Q value based on the learned immediate rewards and future expected rewards;

[0085] Estimate Q value using neural network;

[0086] Update the parameters of the actor network based on the gradient of the Q value;

[0087] Update the parameters of the targeting network;

[0088] Dynamically adjust the experience pool capacity;

[0089] Use priority sampling strategy for training;

[0090] Input the preprocessed feature vector into the deep reinforcement learning model;

[0091] Generate and validate a classification model to classify energy storage batteries.

[0092] The performance data include voltage, capacity, internal resistance, temperature and cycle life;

[0093] The operation data includes charging time, discharging time, operation mode, load condition and ambient temperature.

[0094] The data features include voltage-capacity curve slope, temperature-internal resistance change rate, cycle life attenuation rate, load adaptability characteristics, the impact of ambient temperature on charge-discharge efficiency, comprehensive health index, and charge-discharge efficiency ratio;

[0095] The slope of the voltage-capacity curve is expressed as,

[0096]

[0097] Among them, V i represents the voltage value at the i-th time point, C i represents the capacity value at the i-th time point, Represents the average value of all voltage values, represents the average of all capacity values, and n represents the total number of time points measured;

[0098] The temperature-internal resistance change rate is expressed as,

[0099]

[0100] Wherein, R1 represents the internal resistance value measured at temperature T1, R2 represents the internal resistance value measured at temperature T2, T1 represents the first ambient temperature, and T2 represents the second ambient temperature;

[0101] The cycle life decay rate is expressed as,

[0102]

[0103] Among them, C N represents the capacity at the Nth cycle, C0 represents the initial capacity, and N represents the total number of cycles;

[0104] The load adaptability characteristic is expressed as,

[0105]

[0106] Among them, T dis Indicates the discharge time, L indicates the load condition (current I or power P);

[0107] The effect of the ambient temperature on the charge and discharge efficiency is expressed as:

[0108]

[0109] Among them, T chg Indicates the charging time at a specific ambient temperature, T dis Indicates the discharge time at a specific ambient temperature, T amb Indicates the ambient temperature;

[0110] The comprehensive health index is expressed as,

[0111]

[0112] Among them, V represents the battery voltage, R represents the battery internal resistance, C represents the current capacity of the battery, and C rated Indicates the rated capacity of the battery, N life Indicates the current number of cycles, N rated represents the rated number of cycles, α, β, γ, δ represent weight coefficients;

[0113] The charge-discharge efficiency ratio is expressed as,

[0114]

[0115] Among them, T dis,n represents the discharge time in operation mode n, T chg,nIndicates the charging time in operating mode n.

[0116] The state space is represented as,

[0117]

[0118] The execution space includes adjusting the charging mode, adjusting the load current and adjusting the load power;

[0119] The expected reward is expressed as,

[0120]

[0121] Where R represents the expected reward, r(s,a) represents the immediate reward for performing operation a in state s, γ represents the decay coefficient, ranging from 0 to 1, representing the discount of future rewards, P(s′|s,a) represents the state transition probability, which represents the probability of transferring to the new state s′ after performing operation a in state s, and V(s′) represents the value function of the new state s′;

[0122] The value function is expressed as,

[0123]

[0124] Among them, V(s) represents the value function of state s, which represents the maximum Q value that can be obtained in state s. Q(s,a) represents the value function of the state-action pair, which represents the long-term reward of performing action a in state s;

[0125] The Q value is expressed as,

[0126]

[0127] Among them, Q(s,a) represents the value function of the state-action pair, r(s,a) represents the immediate reward, γ represents the decay coefficient, P(s′|s,a) represents the state transition probability, and V(s′) represents the value function of the new state.

[0128] The updated Q value is expressed as,

[0129]

[0130] Among them, Q(s,a) represents the value function of the state-action pair, α represents the learning rate, which controls the speed of Q value update, r represents the immediate reward, γ represents the decay coefficient, and Q(s′,a′) represents the maximum Q value in the new state;

[0131] The Q value estimated by neural network is expressed as:

[0132]

[0133] in, represents the Q value estimated by the neural network, θ represents the parameters of the neural network, Q(s,a) represents the actual Q value, min θ Indicates optimizing network parameters by minimizing the error;

[0134] The parameter update of the actor network based on the Q-value gradient is expressed as,

[0135]

[0136] Among them, θ actor represents the parameters of the actor network, α represents the learning rate, represents the gradient of Q value with respect to the actor network parameters;

[0137] The parameters of the updated targeting network are expressed as,

[0138] θ′=τθ+(1-τ)θ′

[0139] Among them, θ′ represents the parameters of the target network, τ represents the parameters that control the update speed, and θ represents the parameters of the current network.

[0140] The dynamically adjusted experience pool capacity is expressed as:

[0141] T(n)=T(1)×n -k

[0142]

[0143] η=α×S

[0144] β=1-α

[0145] N=η×i

[0146] Among them, T(n) represents the time period of the nth event, T(1) represents the initial time period, n represents the number of current training times, k represents the parameter of learning effect, which is used to control the change of learning curve, α represents the change of learning rate, which reflects the improvement of learning efficiency, η represents learning efficiency, S represents the sample size in the experience pool, β represents the supplementary parameter, which is used to adjust the capacity of experience pool and learning efficiency, N represents the capacity of experience pool, and i represents the number of training times of the algorithm.

[0147] The priority sampling strategy is expressed as,

[0148]

[0149] Among them, Δ represents the temporal difference error, which measures the importance of the experience data, r(s,a) represents the immediate reward, γ represents the attenuation coefficient, Q(s,a) represents the Q value of performing operation a in the current state s, Q(s′,a′) represents the Q value of performing operation a′ in the new state s′, P(sample) represents the probability that the experience data is sampled, Δ: temporal difference error, α represents the priority parameter, which controls the influence of TD value on sampling probability, π represents the priority parameter, which is used to calculate the priority of each experience data, ∈ represents a small positive value to prevent the TD value from being zero, and γ represents the weight parameter, which is used to control the distribution of the priority parameter.

[0150] The classifying of the energy storage battery includes generating a classification model through the trained deep reinforcement learning model, and classifying the state of the energy storage battery based on the input feature vector;

[0151] After the model is trained, an independent validation dataset is used to test the classification model to verify the classification accuracy and generalization ability of the model on the validation dataset.

[0152] Further adjust model parameters or training strategies based on the test results of the validation set;

[0153] The trained and validated classification model is used to classify the energy storage battery to generate specific classification labels, which include healthy and efficient operation, degraded and inefficient operation, high temperature risk, and excessive internal resistance.

[0154] Embodiment 2, an embodiment of the present invention, which is different from the previous embodiment in that:

[0155] If the functions are implemented in the form of software functional units and sold or used as independent products, they can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present invention, or the part that contributes to the prior art or the part of the technical solution, can be embodied in the form of a software product. The computer software product is stored in a storage medium, including several instructions for a computer device (which can be a personal computer, a server, or a network device, etc.) to perform all or part of the steps of the methods described in each embodiment of the present invention. The aforementioned storage medium includes: U disk, mobile hard disk, read-only memory (ROM, Read-Only Memory), random access memory (RAM, Random Access Memory), disk or optical disk, etc., which can store program codes.

[0156] The logic and / or steps represented in the flowchart or otherwise described herein, for example, can be considered as an ordered list of executable instructions for implementing logical functions, and can be embodied in any computer-readable medium for use by an instruction execution system, device or apparatus (such as a computer-based system, a system including a processor, or other system that can fetch instructions from an instruction execution system, device or apparatus and execute instructions), or in conjunction with such instruction execution systems, devices or apparatuses. For the purposes of this specification, "computer-readable medium" can be any device that can contain, store, communicate, propagate or transmit a program for use by an instruction execution system, device or apparatus, or in conjunction with such instruction execution systems, devices or apparatuses.

[0157] More specific examples of computer-readable media (a non-exhaustive list) include the following: an electrical connection with one or more wires (electronic device), a portable computer disk case (magnetic device), a random access memory (RAM), a read-only memory (ROM), an erasable and programmable read-only memory (EPROM or flash memory), an optical fiber device, and a portable compact disk read-only memory (CDROM). In addition, the computer-readable medium may even be a paper or other suitable medium on which the program is printed, since the program may be obtained electronically, for example, by optically scanning the paper or other medium, followed by editing, deciphering or, if necessary, processing in another suitable manner, and then stored in a computer memory.

[0158] It should be understood that the various parts of the present invention can be implemented by hardware, software, firmware or a combination thereof. In the above-mentioned embodiments, a plurality of steps or methods can be implemented by software or firmware stored in a memory and executed by a suitable instruction execution system. For example, if implemented by hardware, as in another embodiment, it can be implemented by any one of the following technologies known in the art or their combination: a discrete logic circuit having a logic gate circuit for implementing a logic function for a data signal, a dedicated integrated circuit having a suitable combination of logic gate circuits, a programmable gate array (PGA), a field programmable gate array (FPGA), etc.

[0159] Example 3 is an embodiment of the present invention, which provides a method for classifying the operating status of an energy storage system. In order to verify the beneficial effects of the present invention, a scientific demonstration is carried out through simulation experiments.

[0160] Use lithium-ion energy storage batteries of the same specifications. Collect performance data and operation data.

[0161] The traditional rule-based energy storage battery status monitoring method and the deep reinforcement learning classification method of the present invention.

[0162] Battery status monitoring and classification at different temperatures, different loads, and different operating modes.

[0163] 50 battery samples were used and divided into two groups, and classified using the traditional method and the method of the present invention respectively.

[0164] Collect performance and operation data of 50 battery samples. Clean and normalize the collected data to ensure data accuracy and consistency.

[0165] The basic characteristics of voltage, capacity and internal resistance are extracted using the traditional method; the characteristics are extracted using the method of the present invention.

[0166] The two sets of data were trained using the traditional method and the method of the present invention respectively. The model was tested using the same validation data set, and the classification accuracy and other performance indicators were recorded.

[0167] The classification accuracy, predicted battery life, and battery performance under different operation modes are recorded for each method.

[0168] The classification accuracy, predicted battery life, and battery performance optimization effects of the two methods are compared. The experimental results are shown in Table 1.

[0169] Table 1 Comparison of experimental results

[0170]

[0171] The method of the present invention improves classification accuracy by introducing a deep reinforcement learning algorithm. The algorithm can effectively utilize the multi-dimensional data of the battery and capture the complex relationship between different features. By extracting features and defining them in the state space, the model can more accurately reflect the health status of the battery. At the same time, the Q value is estimated using a neural network, and the network parameters are continuously optimized through the gradient descent method, so that the model can still maintain high-precision classification capabilities when facing different environments and operating conditions. Compared with traditional classification methods based on simple rules and experience, the advantage of the deep reinforcement learning algorithm lies in its adaptability and accuracy, which makes classification more reliable.

[0172] In terms of predicting battery life, the method of the present invention significantly improves the model's prediction accuracy for life by dynamically adjusting the experience pool capacity and adopting a priority sampling strategy. Specifically, the dynamic adjustment of the experience pool capacity enables the model to prioritize learning key data that has a significant impact on battery life during training, thereby enhancing the model's adaptability to future data. The priority sampling strategy further ensures that the model can learn the most representative and important samples from past experience. These technical means not only enable the model to more accurately predict battery life, but also reduce reliance on unimportant data, improving the efficiency and accuracy of overall predictions.

[0173] The significant improvement in performance optimization of the present invention is attributed to the introduction of an adaptive reward mechanism and Q-value update strategy. By designing a reward mechanism and updating the Q-value according to the immediate reward and future expected reward after each operation, the system can continuously optimize the battery's charge and discharge strategy. This dynamic adjustment allows the battery to maintain optimal performance under different operating modes and environmental conditions. Traditional methods usually rely on fixed rules and parameters and are difficult to adapt to real-time changes in battery status. The method of the present invention can be adjusted in real time according to changes in the environment and operation, thereby achieving more efficient performance management. This technical solution ensures the optimal operating state of the battery and significantly improves its overall operating efficiency.

[0174] Through the application of the above technical solutions, the method of the present invention has achieved significant advantages in energy storage system operation status classification, service life prediction and performance optimization. The combination of deep reinforcement learning algorithm, adaptive reward mechanism, dynamic experience pool management and priority sampling strategy not only improves the accuracy and reliability of the model, but also enables the system to maintain efficient operation in complex and dynamic environments.

[0175] It should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention rather than to limit it. Although the present invention has been described in detail with reference to the preferred embodiments, those skilled in the art should understand that the technical solutions of the present invention may be modified or replaced by equivalents without departing from the spirit and scope of the technical solutions of the present invention, which should all be included in the scope of the claims of the present invention.

Claims

1. A method for classifying the operating status of an energy storage system, characterized in that: include: Collect performance and operation data of energy storage batteries; Clean and normalize the collected data; Extract the preprocessed data features and construct feature vectors; Define state space and execution space; Design reward mechanisms and calculate expected rewards; Calculate the value function and Q value; Continuously update the Q value based on the learned immediate rewards and future expected rewards; Estimate Q value using neural network; Update the parameters of the actor network based on the gradient of the Q value; Update the parameters of the targeting network; Dynamically adjust the experience pool capacity; Use priority sampling strategy for training; Input the preprocessed feature vector into the deep reinforcement learning model; Generate and validate a classification model to classify energy storage batteries.

2. The energy storage system operating status classification method according to claim 1, characterized in that: The performance data include voltage, capacity, internal resistance, temperature and cycle life; The operation data includes charging time, discharging time, operation mode, load condition and ambient temperature.

3. The method for classifying the operating status of an energy storage system according to claim 2, characterized in that: The data features include voltage-capacity curve slope, temperature-internal resistance change rate, cycle life attenuation rate, load adaptability characteristics, the impact of ambient temperature on charge-discharge efficiency, comprehensive health index, and charge-discharge efficiency ratio; The slope of the voltage-capacity curve is expressed as, Among them, V i represents the voltage value at the i-th time point, C i represents the capacity value at the i-th time point, Represents the average value of all voltage values, represents the average of all capacity values, and n represents the total number of time points measured; The temperature-internal resistance change rate is expressed as, Wherein, R1 represents the internal resistance value measured at temperature T1, R2 represents the internal resistance value measured at temperature T2, T1 represents the first ambient temperature, and T2 represents the second ambient temperature; The cycle life decay rate is expressed as, Among them, C N represents the capacity at the Nth cycle, C0 represents the initial capacity, and N represents the total number of cycles; The load adaptability characteristic is expressed as, Among them, T dis Indicates the discharge time, L indicates the load condition; The effect of the ambient temperature on the charge and discharge efficiency is expressed as: Among them, T chg Indicates the charging time at a specific ambient temperature, T dis Indicates the discharge time at a specific ambient temperature, T amb Indicates the ambient temperature; The comprehensive health index is expressed as, Among them, V represents the battery voltage, R represents the battery internal resistance, C represents the current capacity of the battery, and C rated Indicates the rated capacity of the battery, N life Indicates the current number of cycles, N rated represents the rated number of cycles, α, β, γ, δ represent weight coefficients; The charge-discharge efficiency ratio is expressed as, Among them, T dis,n represents the discharge time in operation mode n, T chg,n Indicates the charging time in operating mode n.

4. The method for classifying the operating status of an energy storage system according to claim 3, characterized in that: The state space is represented as, The execution space includes adjusting the charging mode, adjusting the load current and adjusting the load power; The expected reward is expressed as, Where R represents the expected reward, r(s,a) represents the immediate reward for performing operation a in state s, γ represents the decay coefficient, ranging from 0 to 1, representing the discount of future rewards, P(s′|s,a) represents the state transition probability, which represents the probability of transferring to the new state s′ after performing operation a in state s, and V(s′) represents the value function of the new state s′; The value function is expressed as, Among them, V(s) represents the value function of state s, which represents the maximum Q value that can be obtained in state s. Q(s,a) represents the value function of the state-action pair, which represents the long-term reward of performing action a in state s; The Q value is expressed as, Among them, Q(s,a) represents the value function of the state-action pair, r(s,a) represents the immediate reward, γ represents the decay coefficient, P(s′|s,a) represents the state transition probability, and V(s′) represents the value function of the new state.

5. The method for classifying the operating status of an energy storage system according to claim 4, characterized in that: The updated Q value is expressed as, Among them, Q(s,a) represents the value function of the state-action pair, α represents the learning rate, which controls the speed of Q value update, r represents the immediate reward, γ represents the decay coefficient, and Q(s′,a′) represents the maximum Q value in the new state; The Q value estimated by neural network is expressed as: in, represents the Q value estimated by the neural network, θ represents the parameters of the neural network, Q(s,a) represents the actual Q value, min θ Indicates optimizing network parameters by minimizing the error; The parameter update of the actor network based on the Q-value gradient is expressed as, Among them, θ actor represents the parameters of the actor network, α represents the learning rate, represents the gradient of Q value with respect to the actor network parameters; The parameters of the updated targeting network are expressed as, θ′=τθ+(1-τ)θ′ Among them, θ′ represents the parameters of the target network, τ represents the parameters that control the update speed, and θ represents the parameters of the current network.

6. The method for classifying the operating status of an energy storage system according to claim 5, characterized in that: The dynamically adjusted experience pool capacity is expressed as: T(n)=T(1)×n -k η=α×S β＝1-α N=η×i Among them, T(n) represents the time period of the nth event, T(1) represents the initial time period, n represents the number of current training times, k represents the parameter of learning effect, which is used to control the change of learning curve, α represents the change of learning rate, which reflects the improvement of learning efficiency, η represents learning efficiency, S represents the sample size in the experience pool, β represents the supplementary parameter, which is used to adjust the capacity of experience pool and learning efficiency, N represents the capacity of experience pool, and i represents the number of training times of the algorithm.

7. The method for classifying the operating status of an energy storage system according to claim 6, characterized in that: The priority sampling strategy is expressed as, Among them, Δ represents the temporal difference error, which measures the importance of the experience data, r(s,a) represents the immediate reward, γ represents the attenuation coefficient, Q(s,a) represents the Q value of performing operation a in the current state s, Q(s′,a′) represents the Q value of performing operation a′ in the new state s′, P(sample) represents the probability that the experience data is sampled, Δ: temporal difference error, α represents the priority parameter, which controls the influence of TD value on sampling probability, π represents the priority parameter, which is used to calculate the priority of each experience data, ∈ represents a small positive value to prevent the TD value from being zero, and γ represents the weight parameter, which is used to control the distribution of the priority parameter.

8. The method for classifying the operating status of an energy storage system according to claim 7, characterized in that: The classifying of the energy storage battery includes generating a classification model through the trained deep reinforcement learning model, and classifying the state of the energy storage battery based on the input feature vector; After the model is trained, an independent validation dataset is used to test the classification model to verify the classification accuracy and generalization ability of the model on the validation dataset. Further adjust model parameters or training strategies based on the test results of the validation set; The trained and validated classification model is used to classify the energy storage battery to generate specific classification labels, which include healthy and efficient operation, degraded and inefficient operation, high temperature risk, and excessive internal resistance.

9. A computer device comprising a memory and a processor, wherein the memory stores a computer program, wherein: When the processor executes the computer program, the steps of the energy storage system operating status classification method described in any one of claims 1 to 8 are implemented.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the steps of the energy storage system operating status classification method described in any one of claims 1 to 8 are implemented.

Images