Abnormal voltage recovery method and system of lithium battery energy storage system

Abstract

The invention discloses an abnormal voltage recovery method for a lithium battery energy storage system. The method comprises the following steps: acquiring historical data information of a target lithium battery energy storage power station; determining an early warning threshold value; real-time data information of the target lithium battery energy storage power station is obtained in real time; abnormal battery positioning is carried out; constructing a Markov decision process; and performing abnormal voltage recovery of the target lithium battery energy storage power station by adopting a Markov decision process. The invention also discloses a system for realizing the abnormal voltage recovery method of the lithium battery energy storage system. According to the method, the data information of the target lithium battery energy storage system is acquired, and the established Markov decision process is adopted to realize monitoring and decision of the target lithium battery energy storage system, so that abnormal voltage recovery of the lithium battery energy storage system can be realized, the reliability is higher, the accuracy is better, and the method is more objective and scientific.

Description

Technical Field

[0001] The present invention belongs to the field of electrical automation, and in particular relates to an abnormal voltage recovery method and system for a lithium battery energy storage system. Background Art

[0002] With the development of economy and technology and the improvement of people's living standards, electricity has become an indispensable secondary energy source in people's production and life, bringing endless convenience to people's production and life. Therefore, ensuring the stable and reliable supply of electricity has become one of the most important tasks of the power system.

[0003] As environmental problems become more and more serious, more and more new energy power generation systems are being integrated into the power system to generate electricity. However, the randomness and discontinuity of the output of new energy power generation systems have brought great operating pressure to the power system. Electrochemical energy storage systems have flexible charging and discharging characteristics, which can smooth power fluctuations, reduce peak loads, improve power quality, and participate in system frequency regulation; therefore, power systems have begun to widely deploy electrochemical energy storage systems to solve various problems caused by the grid connection of new energy power generation systems.

[0004] Electrochemical energy storage systems are generally composed of a large number of single lithium batteries connected in series and parallel. Therefore, it is particularly important to accurately and reliably monitor abnormalities and restore voltage for electrochemical energy storage systems. At present, the abnormal voltage monitoring and recovery scheme of traditional lithium battery energy storage systems is still a scheme of manual monitoring and manual setting of the recovery process. However, this type of manual scheme not only has poor reliability and accuracy, but also has a large number of subjective factors in the processing process and is less scientific. Summary of the invention

[0005] One of the purposes of the present invention is to provide an abnormal voltage recovery method for a lithium battery energy storage system which is highly reliable, accurate, objective and scientific.

[0006] A second object of the present invention is to provide a system for realizing the abnormal voltage recovery method of the lithium battery energy storage system.

[0007] The abnormal voltage recovery method of the lithium battery energy storage system provided by the present invention comprises the following steps:

[0008] S1. Obtain historical data information of the target lithium battery energy storage power station;

[0009] S2. Determine the warning threshold according to the data information obtained in step S1;

[0010] S3. Obtain real-time data information of the target lithium battery energy storage power station in real time;

[0011] S4. According to the data information obtained in step S3, locate the abnormal battery;

[0012] S5. Construct a Markov decision process;

[0013] S6. Using the Markov decision process constructed in step S5, the abnormal voltage of the target lithium battery energy storage power station is restored.

[0014] Step S2, according to the data information obtained in step S1, determines the warning threshold, specifically including the following steps:

[0015] According to the data information obtained in step S1, the abnormal voltage warning values ​​are determined: the first level warning threshold TT1, the second level warning threshold TT2 and the third level warning threshold TT3;

[0016] When the voltage of a single lithium battery is less than TT1, the single lithium battery operates normally;

[0017] When the voltage of a single lithium battery is not less than TT1 and less than TT2, the single lithium battery operates normally and alarms, and is defined as a first-level abnormal working state;

[0018] When the voltage of a single lithium battery is not less than TT2 and less than TT3, the single lithium battery operates at reduced power and is defined as a secondary abnormal working state;

[0019] When the voltage of a single lithium battery is not less than TT3, the lithium battery energy storage system will shut down the corresponding single lithium battery and define it as a level 3 abnormal working state.

[0020] The real-time acquisition of the real-time data information of the target lithium battery energy storage power station described in step S3 specifically includes the following steps:

[0021] Through the battery cell management unit, the real-time voltage data of each lithium battery in the target lithium battery energy storage power station is obtained in real time, and the abnormal voltage is reported to the battery group management unit; the abnormal voltage U M,N Expressed as Among them U M,N The abnormal voltage of the single lithium battery numbered N when running at time M;

[0022] The battery pack management unit forms a voltage matrix with the received abnormal voltage data, which is expressed as Among them U i,M,N It indicates the abnormal voltage value of the battery numbered N in the i-th battery group when it is running at time M.

[0023] The Markov decision process described in step S5 specifically includes the following steps:

[0024] The process of the battery cluster management unit deciding on abnormal voltage recovery is constructed as a Markov decision process and defined as a five-tuple (S, A, P, R, Done):

[0025] S is the state space: define the state space S of the single lithium battery at time t t For S t =(S0, S1, S2, S3, N), where S0 represents the normal working state, S1 represents the first level abnormal working state, S2 represents the second level abnormal working state, S3 represents the third level abnormal working state, and N represents the number of abnormalities. When the nth abnormal voltage information is uploaded to the battery cluster management unit, the working state S of the lithium battery at this time n Indicated as S n =(S i ,n-1), S n Indicates the abnormal working state of the lithium battery and the number of abnormalities when the battery cluster management unit receives the nth abnormal signal;

[0026] A is the decision space: define the decision space A of the battery cluster management unit at time t t A t =(a0,a1,a2,a3,T1,T2,T3), where a0 indicates that the decision is normal operation, a1 indicates that the decision is level one abnormal operation, a2 indicates that the decision is level two abnormal operation, a3 indicates that the decision is level three abnormal operation, T1 indicates the normal operation time from the last abnormal voltage recovery to the next abnormal voltage, T2 is the time from the abnormal voltage to the decision action, and T3 is the abnormal voltage recovery time; when the abnormal voltage information is uploaded to the battery cluster management unit, the battery cluster management unit decides the recovery action a n Represented as a n =(a i ,T i,1 ,T i,2 ,T i,3 ), where a i represents the recovery action of the decision, T i,1 Represents normal working hours, T i,2 represents the decision time, T i,3 Indicates voltage recovery time;

[0027] P is the state transition probability, P is expressed as P(S t+1 |S t =S t ,A t =a t ), indicating that the battery cluster management unit takes decision a at time t t Change the state from S t Transformed to S t+1 The probability of

[0028] R is the reward function, which follows the probability distribution P(S t+1 |St =S t ,A t =a t ); The battery cluster management unit adopts decision a at time t t The reward R t , expressed as

[0029] R t =r t +γr t+1 +γ 2 r t+2 +…+γ H-1 r t+H-1

[0030] Where r t is the timely reward function; γ is the future time reward attenuation coefficient; H is the decision prediction cycle;

[0031] Done is the decision-making completion signal;

[0032] π i The output strategy space is parameterized by several decision actions, denoted as π i =[a i ,T i,1 ,T i,2 ,T i,3 ; a i+1 ,T i+1,1 ,T i+1,2 ,T i+1,3 ;...;a i+n ,T i+n,1 ,T i+n,2 ,T i+n,3 ].

[0033] The instant reward function specifically includes the following steps:

[0034] The abnormal action strategy continuity reward r1 is calculated using the following formula:

[0035]

[0036] Where T at+1 Indicates strategy a t+1 execution time; T1 is the acquisition time between two consecutive acquisitions of data information; ε is the error caused by communication delay;

[0037] Calculate the abnormal action strategy correctness reward r2 as

[0038] The first reward r3 for the reliability of energy storage power station operation is calculated using the following formula:

[0039]

[0040] Where MTBF T The average time without abnormality in a decision cycle; MTBF S is the expected average abnormal-free working time in a decision cycle;

[0041] The second reward r4 for the reliability of energy storage power station operation is calculated using the following formula:

[0042]

[0043] Where MTTR T MTTR is the average repair time within a decision cycle. S is the expected average repair time within a decision cycle;

[0044] The third reward r5 for the reliability of energy storage power station operation is calculated using the following formula:

[0045]

[0046] Where RTOP T is the relative number of abnormalities in a decision cycle, and FTOP T BPN is the number of abnormal lithium battery times in a decision cycle. T is the total number of single lithium batteries in a decision cycle; RTOP S is the expected number of relative anomalies in a decision cycle;

[0047] The following formula is used to calculate the fourth reward r6 of the energy storage power station operation reliability:

[0048]

[0049] IRB T is the battery failure rate within a decision cycle, and NIB T is the number of failed battery cells in a decision cycle, NB T is the total number of battery cells in a decision cycle; IRB S is the expected battery failure rate within a decision cycle;

[0050] The timely reward function r is calculated using the following formula: t :

[0051] r t =[λ1r1+λ2r2+λ3(r3+r4+r5+r6)] T

[0052] in[] Trepresents the value in the Tth decision cycle; λ1 is the first weight coefficient, λ2 is the second weight coefficient, λ3 is the third weight coefficient, and λ1+λ2+λ3=1.

[0053] The Markov decision process constructed in step S5 described in step S6 is used to restore the abnormal voltage of the target lithium battery energy storage power station, which specifically includes the following steps:

[0054] Use the Markov decision process constructed in step S5 to find the optimal strategy π t * , maps the battery working state to the decision action to maximize the expected reward within the decision cycle, expressed as

[0055]

[0056] In the Markov decision process, the next state is only related to the current state and action. Therefore, the decision process of maximizing the total reward is divided into the cumulative sum of the maximum rewards for each decision action. Each step optimizes the current decision action only through the reward function value under the current action, which is expressed as

[0057]

[0058] Where V * (s t ,a t ) is the reward function value obtained by taking the current decision action; R(s t ,a t ) is the immediate reward function value obtained when taking the current decision action; Q * (s t+1 ,a t+1 ) is the delayed reward function value obtained by taking the current decision action;

[0059] When the nth abnormal voltage signal is uploaded to the battery cluster management unit, the battery cluster management unit decides a recovery action, calculates the reward value corresponding to the decision, and when the next abnormal signal comes, optimizes the decision of the next abnormal voltage according to the reward value obtained from the previous decision. By continuously iteratively updating the recovery strategy until the maximum reward is reached, the optimal recovery action decision for the battery under abnormal state is finally obtained, and the abnormal voltage recovery of the target lithium battery energy storage power station is completed.

[0060] The present invention also provides a system for realizing the abnormal voltage recovery method of the lithium battery energy storage system, comprising a history acquisition module, a threshold determination module, a real-time acquisition module, an abnormal positioning module, a decision building module and a voltage recovery module; the history acquisition module, the threshold determination module, the real-time acquisition module, the abnormal positioning module, the decision building module and the voltage recovery module are connected in series in sequence; the history acquisition module is used to acquire the historical data information of the target lithium battery energy storage power station, and upload the data information to the threshold determination module; the threshold determination module is used to determine the warning threshold according to the received data information and the acquired data information, and upload the data information to the real-time acquisition module; the real-time acquisition module is used to acquire the real-time data information of the target lithium battery energy storage power station in real time according to the received data information, and upload the data information to the abnormal positioning module; the abnormal positioning module is used to locate the abnormal battery according to the received data information and the acquired data information, and upload the data information to the decision building module; the decision building module is used to construct a Markov decision process according to the received data information, and upload the data information to the voltage recovery module; the voltage recovery module is used to recover the abnormal voltage of the target lithium battery energy storage power station by using the constructed Markov decision process according to the received data information.

[0061] The abnormal voltage recovery method and system of the lithium battery energy storage system provided by the present invention acquires data information of the target lithium battery energy storage system and adopts a constructed Markov decision process to realize monitoring and decision-making of the target lithium battery energy storage system. Therefore, the present invention can not only realize abnormal voltage recovery of the lithium battery energy storage system, but also has higher reliability, better accuracy, and is more objective and scientific. BRIEF DESCRIPTION OF THE DRAWINGS

[0062] Figure 1 The figure is a schematic diagram of the method flow of the present invention.

[0063] Figure 2 Schematic diagram of the functional modules of the system of the present invention. DETAILED DESCRIPTION

[0064] like Figure 1 The method flow chart of the method of the present invention is shown as follows: The abnormal voltage recovery method of the lithium battery energy storage system disclosed in the present invention comprises the following steps:

[0065] S1. Obtain historical data information of the target lithium battery energy storage power station;

[0066] S2. Determine the warning threshold according to the data information obtained in step S1; specifically, the steps include:

[0067] According to the data information obtained in step S1, the abnormal voltage warning values ​​are determined: the first level warning threshold TT1, the second level warning threshold TT2 and the third level warning threshold TT3;

[0068] When the voltage of a single lithium battery is less than TT1, the single lithium battery operates normally;

[0069] When the voltage of a single lithium battery is not less than TT1 and less than TT2, the single lithium battery operates normally and alarms, and is defined as a first-level abnormal working state;

[0070] When the voltage of a single lithium battery is not less than TT2 and less than TT3, the single lithium battery operates at reduced power and is defined as a secondary abnormal working state;

[0071] When the voltage of a single lithium battery is not less than TT3, the lithium battery energy storage system shuts down the corresponding single lithium battery and defines it as a level 3 abnormal working state;

[0072] In specific implementation, the values ​​of the first-level warning threshold TT1, the second-level warning threshold TT2, and the third-level warning threshold TT3 are deliberately determined based on experience identification, machine learning, information fusion and other technologies, by extracting voltage characteristic signals from the historical abnormal data set of the energy storage power station, and comprehensively considering the construction period and operation standards of the energy storage power station;

[0073] S3. Real-time acquisition of real-time data information of the target lithium battery energy storage power station; specifically includes the following steps:

[0074] Through the battery cell management unit, the real-time voltage data of each lithium battery in the target lithium battery energy storage power station is obtained in real time, and the abnormal voltage is reported to the battery group management unit; the abnormal voltage U M,N Expressed as Among them U M,N The abnormal voltage of the single lithium battery numbered N when running at time M;

[0075] The battery pack management unit forms a voltage matrix with the received abnormal voltage data, which is expressed as Among them U i,M,N It represents the abnormal voltage value of the battery numbered N in the i-th battery group when it is running at time M;

[0076] S4. According to the data information obtained in step S3, locate the abnormal battery;

[0077] S5. Construct a Markov decision process; specifically, the following steps are included:

[0078] The process of the battery cluster management unit deciding on abnormal voltage recovery is constructed as a Markov decision process and defined as a five-tuple (S, A, P, R, Done):

[0079] S is the state space: define the state space S of the single lithium battery at time t t For S t=(S0, S1, S2, S3, N), where S0 represents the normal working state, S1 represents the first level abnormal working state, S2 represents the second level abnormal working state, S3 represents the third level abnormal working state, and N represents the number of abnormalities. When the nth abnormal voltage information is uploaded to the battery cluster management unit, the working state S of the lithium battery at this time n Indicated as S n =(S i ,n-1), S n Indicates the abnormal working state of the lithium battery and the number of abnormalities when the battery cluster management unit receives the nth abnormal signal;

[0080] A is the decision space: define the decision space A of the battery cluster management unit at time t t A t =(a0,a1,a2,a3,T1,T2,T3), where a0 indicates that the decision is normal operation, a1 indicates that the decision is level one abnormal operation, a2 indicates that the decision is level two abnormal operation, a3 indicates that the decision is level three abnormal operation, T1 indicates the normal operation time from the last abnormal voltage recovery to the next abnormal voltage, T2 is the time from the abnormal voltage to the decision action, and T3 is the abnormal voltage recovery time; when the abnormal voltage information is uploaded to the battery cluster management unit, the battery cluster management unit decides the recovery action a n Represented as a n =(a i ,T i,1 ,T i,2 ,T i,3 ), where a i represents the recovery action of the decision, T i,1 Represents normal working hours, T i,2 represents the decision time, T i,3 Indicates voltage recovery time;

[0081] P is the state transition probability, P is expressed as P(S t+1 |S t =S t ,A t =a t ), indicating that the battery cluster management unit takes decision a at time t t Change the state from S t Transformed to S t+1 The probability of

[0082] R is the reward function, which follows the probability distribution P(S t+1 |S t =S t ,A t =a t ); The battery cluster management unit adopts decision a at time t tThe reward R t , expressed as

[0083] R t =r t +γr t+1 +γ 2 r t+2 +…+γ H-1 r t+H-1

[0084] Where r t is the timely reward function; γ is the future time reward attenuation coefficient; H is the decision prediction cycle;

[0085] Done is the decision-making completion signal;

[0086] π i The output strategy space is parameterized by several decision actions, denoted as π i =[a i ,T i,1 ,T i,2 ,T i,3 ; a i+1 ,T i+1,1 ,T i+1,2 ,T i+1,3 ;...;a i+n ,T i+n,1 ,T i+n,2 ,T i+n,3 ];

[0087] Among them, the timely reward function specifically includes the following steps:

[0088] The reward function consists of abnormal action strategy continuity reward, abnormal action strategy correctness reward, and energy storage power station operation reliability reward. The energy storage power station operation reliability reward consists of indicators that describe the operation reliability of the energy storage power station, including average abnormal working time, average repair time, relative abnormal number of batteries, and battery failure probability.

[0089] The abnormal action strategy continuity reward r1 is calculated using the following formula. The purpose of this reward is to improve the continuity of the battery management system's decision-making voltage recovery process and prevent the harm caused by missed processing:

[0090]

[0091] Where T at+1 Indicates strategy a t+1 execution time; T1 is the acquisition time between two consecutive acquisitions of data information; ε is the error caused by communication delay;

[0092] Calculate the abnormal action strategy correctness reward r2 as The purpose of this reward is to improve the accuracy of the battery management system's decision-making voltage recovery process and prevent the harm caused by mishandling;

[0093] The purpose of the following rewards is to improve the reliability of the operation of energy storage power stations. Evaluation indicators that describe the reliability of the operation of energy storage power stations are selected, including average abnormal working time, average repair time, relative abnormal number of batteries, and battery failure probability;

[0094] The first reward r3 for the reliability of energy storage power station operation is calculated using the following formula:

[0095]

[0096] Where MTBF T The average time without abnormality in a decision cycle; MTBF S is the expected average abnormal-free working time in a decision cycle;

[0097] The second reward r4 for the reliability of energy storage power station operation is calculated using the following formula:

[0098]

[0099] Where MTTR T MTTR is the average repair time within a decision cycle. S is the expected average repair time within a decision cycle;

[0100] The third reward r5 for the reliability of energy storage power station operation is calculated using the following formula:

[0101]

[0102] Where RTOP T is the relative number of abnormalities in a decision cycle, and FTOP T BPN is the number of abnormal lithium battery times in a decision cycle. T is the total number of single lithium batteries in a decision cycle; RTOP S is the expected number of relative anomalies in a decision cycle;

[0103] The following formula is used to calculate the fourth reward r6 of the energy storage power station operation reliability:

[0104]

[0105] IRB T is the battery failure rate within a decision cycle, and NIB T is the number of failed battery cells in a decision cycle, NB Tis the total number of battery cells in a decision cycle; IRB S is the expected battery failure rate within a decision cycle;

[0106] The timely reward function r is calculated using the following formula: t :

[0107] r t =[λ1r1+λ2r2+λ3(r3+r4+r5+r6)] T

[0108] in[] T represents the value in the Tth decision cycle; λ1 is the first weight coefficient, λ2 is the second weight coefficient, λ3 is the third weight coefficient, and λ1+λ2+λ3=1;

[0109] The above reward function is set mainly to take into account the accuracy, continuity and reliability of the abnormal voltage recovery action decided by the battery cluster management unit of the energy storage power station. The battery cluster management unit constantly adjusts its decision-making actions to maximize the reward. After training with a large amount of data, the external representation is that when the battery group management unit locates the abnormality, the abnormal voltage signal is uploaded to the battery cluster management unit. The battery cluster management unit will quickly and accurately decide on the abnormal voltage recovery action while taking into account the reliability index of the energy storage power station operation, thereby quickly completing the abnormal recovery process of the battery and improving the safety prevention and control level of the energy storage power station;

[0110] S6. Using the Markov decision process constructed in step S5 to restore the abnormal voltage of the target lithium battery energy storage power station; specifically comprising the following steps:

[0111] Use the Markov decision process constructed in step S5 to find the optimal strategy The battery working state is mapped to the decision action to maximize the expected reward within the decision cycle, expressed as

[0112]

[0113] In the Markov decision process, the next state is only related to the current state and action. Therefore, the decision process of maximizing the total reward is divided into the cumulative sum of the maximum rewards for each decision action. Each step optimizes the current decision action only through the reward function value under the current action, which is expressed as

[0114] V*(st,at)=maxaE{R(st,at)+γ·P(St +1 | St = st, At = at) · Q * (st +1 ,at +1 )}

[0115] Where V* (s t ,a t ) is the reward function value obtained under the current decision action; R(s t ,a t ) is the immediate reward function value obtained when taking the current decision action; Q * (s t+1 ,a t+1 ) is the delayed reward function value obtained by taking the current decision action;

[0116] When the nth abnormal voltage signal is uploaded to the battery cluster management unit, the battery cluster management unit decides a recovery action, calculates the reward value corresponding to the decision, and when the next abnormal signal comes, optimizes the decision of the next abnormal voltage according to the reward value obtained from the previous decision. By continuously iteratively updating the recovery strategy until the maximum reward is reached, the optimal recovery action decision for the battery under abnormal state is finally obtained, and the abnormal voltage recovery of the target lithium battery energy storage power station is completed.

[0117] like Figure 2 The functional module schematic diagram of the system of the present invention is shown as follows: the system for realizing the abnormal voltage recovery method of the lithium battery energy storage system disclosed in the present invention comprises a history acquisition module, a threshold determination module, a real-time acquisition module, an abnormal location module, a decision building module and a voltage recovery module; the history acquisition module, the threshold determination module, the real-time acquisition module, the abnormal location module, the decision building module and the voltage recovery module are connected in series in sequence; the history acquisition module is used to acquire the historical data information of the target lithium battery energy storage power station, and upload the data information to the threshold determination module; the threshold determination module is used to determine the warning threshold according to the received data information and the acquired data information, and upload the data information to the real-time acquisition module; the real-time acquisition module is used to acquire the real-time data information of the target lithium battery energy storage power station in real time according to the received data information, and upload the data information to the abnormal location module; the abnormal location module is used to locate the abnormal battery according to the received data information and the acquired data information, and upload the data information to the decision building module; the decision building module is used to construct a Markov decision process according to the received data information, and upload the data information to the voltage recovery module; the voltage recovery module is used to recover the abnormal voltage of the target lithium battery energy storage power station by using the constructed Markov decision process according to the received data information.

Claims

1. A method for recovering abnormal voltage of a lithium battery energy storage system, comprising the following steps: S1. Obtain historical data information of the target lithium battery energy storage power station; S2. Determine the warning threshold according to the data information obtained in step S1; S3. Obtain real-time data information of the target lithium battery energy storage power station in real time; S4. According to the data information obtained in step S3, locate the abnormal battery; S5. Construct a Markov decision process; S6. Using the Markov decision process constructed in step S5, the abnormal voltage of the target lithium battery energy storage power station is restored.

2. The abnormal voltage recovery method of the lithium battery energy storage system according to claim 1, characterized in that Step S2, according to the data information obtained in step S1, determines the warning threshold, specifically including the following steps: According to the data information obtained in step S1, the abnormal voltage warning values ​​are determined: the first level warning threshold TT1, the second level warning threshold TT2 and the third level warning threshold TT3; When the voltage of a single lithium battery is less than TT1, the single lithium battery operates normally; When the voltage of a single lithium battery is not less than TT1 and less than TT2, the single lithium battery operates normally and alarms, and is defined as a first-level abnormal working state; When the voltage of a single lithium battery is not less than TT2 and less than TT3, the single lithium battery operates at reduced power and is defined as a secondary abnormal working state; When the voltage of a single lithium battery is not less than TT3, the lithium battery energy storage system will shut down the corresponding single lithium battery and define it as a level 3 abnormal working state.

3. The abnormal voltage recovery method of the lithium battery energy storage system according to claim 2 is characterized in that The real-time acquisition of the real-time data information of the target lithium battery energy storage power station described in step S3 specifically includes the following steps: Through the battery cell management unit, the real-time voltage data of each lithium battery in the target lithium battery energy storage power station is obtained in real time, and the abnormal voltage is reported to the battery group management unit; the abnormal voltage U M,N Expressed as Among them U M,N The abnormal voltage of the single lithium battery numbered N when running at time M; The battery pack management unit forms a voltage matrix with the received abnormal voltage data, which is expressed as Among them U i,M,N It indicates the abnormal voltage value of the battery numbered N in the i-th battery group when it is running at time M.

4. The abnormal voltage recovery method of the lithium battery energy storage system according to claim 3 is characterized in that The Markov decision process described in step S5 specifically includes the following steps: The process of the battery cluster management unit deciding on abnormal voltage recovery is constructed as a Markov decision process and defined as a five-tuple (S, A, P, R, Done): S is the state space: define the state space S of the single lithium battery at time t t For S t =(S0, S1, S2, S3, N), where S0 represents the normal working state, S1 represents the first level abnormal working state, S2 represents the second level abnormal working state, S3 represents the third level abnormal working state, and N represents the number of abnormalities; When the nth abnormal voltage information is uploaded to the battery cluster management unit, the working state S of the lithium battery is n Indicated as S n =(S i ,n-1), S n Indicates the abnormal working state of the lithium battery and the number of abnormalities when the battery cluster management unit receives the nth abnormal signal; A is the decision space: define the decision space A of the battery cluster management unit at time t t A t =(a0,a1,a2,a3,T1,T2,T3), where a0 indicates that the decision is normal operation, a1 indicates that the decision is level 1 abnormal operation, a2 indicates that the decision is level 2 abnormal operation, a3 indicates that the decision is level 3 abnormal operation, T1 indicates the normal operation time from the last abnormal voltage recovery to the next abnormal voltage, T2 is the time from the abnormal voltage to the decision action, and T3 is the abnormal voltage recovery time; when the abnormal voltage information is uploaded to the battery cluster management unit, the battery cluster management unit decides the recovery action a n Represented as a n =(a i ,T i,1 ,T i,2 ,T i,3 ), where a i represents the recovery action of the decision, T i,1 Represents normal working hours, T i,2 represents the decision time, T i,3 Indicates voltage recovery time; P is the state transition probability, P is expressed as P(S t+1 |S t =S t ,A t =a t ), indicating that the battery cluster management unit takes decision a at time t t Change the state from S t Transformed to S t+1 probability; R is the reward function, which follows the probability distribution P(S t+1 |S t =S t ,A t =a t ); The battery cluster management unit adopts decision a at time t t The reward R t , expressed as R t =r t +γr t+1 +g 2 r t+2 +…+c H-1 r t+H-1 Where r t is the timely reward function; γ is the future time reward attenuation coefficient; H is the decision prediction cycle; Done is the decision-making completion signal; π i The output strategy space is parameterized by several decision actions, denoted as π i =[a i ,T i,1 ,T i,2 ,T i,3 ; a i+1 ,T i+1,1 ,T i+1,2 ,T i+1,3 ;...;a i+n ,T i+n,1 ,T i+n,2 ,T i+n,3 ].

5. The abnormal voltage recovery method of the lithium battery energy storage system according to claim 4, characterized in that The instant reward function specifically includes the following steps: The abnormal action strategy continuity reward r1 is calculated using the following formula: Where T at+1 Indicates strategy a t+1 Time of execution; T1 is the acquisition time between two consecutive acquisitions of data information; ε is the error caused by communication delay; Calculate the abnormal action strategy correctness reward r2 as The first reward r3 for the reliability of energy storage power station operation is calculated using the following formula: Where MTBF T The average time without abnormality in a decision cycle; MTBF S is the expected average abnormal-free working time in a decision cycle; The second reward r4 for the reliability of energy storage power station operation is calculated using the following formula: Where MTTR T MTTR is the average repair time within a decision cycle. S is the expected average repair time within a decision cycle; The third reward r5 for the reliability of energy storage power station operation is calculated using the following formula: Where RTOP T is the relative number of abnormalities in a decision cycle, and FTOP T BPN is the number of abnormal lithium battery times in a decision cycle. T is the total number of single lithium batteries in a decision cycle; RTOP S is the expected number of relative anomalies in a decision cycle; The following formula is used to calculate the fourth reward r6 of the energy storage power station operation reliability: IRB T is the battery failure rate within a decision cycle, and NIB T is the number of failed battery cells in a decision cycle, NB T is the total number of battery cells in a decision cycle; IRB S is the expected battery failure rate within a decision cycle; The timely reward function r is calculated using the following formula: t : r t =[λ1r1+λ2r2+λ3(r3+r4+r5+r6)] T in[] T represents the value in the Tth decision cycle; λ1 is the first weight coefficient, λ2 is the second weight coefficient, λ3 is the third weight coefficient, and λ1+λ2+λ3=1.

6. The abnormal voltage recovery method of the lithium battery energy storage system according to claim 5, characterized in that The Markov decision process constructed in step S5 described in step S6 is used to restore the abnormal voltage of the target lithium battery energy storage power station, which specifically includes the following steps: Use the Markov decision process constructed in step S5 to find the optimal strategy The battery working state is mapped to the decision action to maximize the expected reward within the decision cycle, expressed as In the Markov decision process, the next state is only related to the current state and action. Therefore, the decision process of maximizing the total reward is divided into the cumulative sum of the maximum rewards for each decision action. Each step optimizes the current decision action only through the reward function value under the current action, which is expressed as V*(st,at)=maxaE{R(st,at)+γ·P(St +1 |St=st,At=at)·Q*(st +1 ,at +1 )} Where V * (s t ,a t ) is the reward function value obtained under the current decision action; R(s t ,a t ) is the immediate reward function value obtained when taking the current decision action; Q * (s t+1 ,a t+1 ) is the delayed reward function value obtained by taking the current decision action; When the nth abnormal voltage signal is uploaded to the battery cluster management unit, the battery cluster management unit decides a recovery action, calculates the reward value corresponding to the decision, and when the next abnormal signal comes, optimizes the decision of the next abnormal voltage according to the reward value obtained from the previous decision. By continuously iteratively updating the recovery strategy until the maximum reward is reached, the optimal recovery action decision for the battery under abnormal state is finally obtained, and the abnormal voltage recovery of the target lithium battery energy storage power station is completed.

7. A system for realizing the abnormal voltage recovery method of the lithium battery energy storage system according to any one of claims 1 to 6, characterized in that It includes a history acquisition module, a threshold determination module, a real-time acquisition module, an abnormality positioning module, a decision-making construction module and a voltage recovery module; the history acquisition module, the threshold determination module, the real-time acquisition module, the abnormality positioning module, the decision-making construction module and the voltage recovery module are connected in series in sequence; the history acquisition module is used to obtain the historical data information of the target lithium battery energy storage power station, and upload the data information to the threshold determination module; the threshold determination module is used to determine the warning threshold according to the received data information and the acquired data information, and upload the data information to the real-time acquisition module; The real-time acquisition module is used to acquire the real-time data information of the target lithium battery energy storage power station in real time according to the received data information, and upload the data information to the abnormal positioning module; the abnormal positioning module is used to locate the abnormal battery according to the received data information and the acquired data information, and upload the data information to the decision-making construction module; the decision-making construction module is used to construct a Markov decision process according to the received data information, and upload the data information to the voltage recovery module; the voltage recovery module is used to recover the abnormal voltage of the target lithium battery energy storage power station by using the constructed Markov decision process according to the received data information.

Images