CN117406092A

CN117406092A - State estimation method, device and system for distributed modular battery energy storage system

Info

Publication number: CN117406092A
Application number: CN202311323846.4A
Authority: CN
Inventors: 沈瑜
Original assignee: Yangtze Delta Region Institute of Tsinghua University Zhejiang
Current assignee: Yangtze Delta Region Institute of Tsinghua University Zhejiang
Priority date: 2023-10-12
Filing date: 2023-10-12
Publication date: 2024-01-16

Abstract

The invention provides a state estimation method, device and system for a distributed modular battery energy storage system. The method comprises the following steps: the bad data discrimination module receives the non-electric quantity and electric quantity data at the moment t and judges whether bad data exists in the non-electric quantity and electric quantity data at the moment t; when the non-electric quantity and the electric quantity data at the time t have bad data, carrying out data correction according to a preset rule to obtain correction data at the time t; and determining the real-time state of charge of the battery pack to be tested based on the mutual influence factors among the battery units, the attenuation coefficient of battery energy storage and the correction data received by the state estimation calculation module. Therefore, the real-time state of charge of the distributed modular energy storage system can be obtained, the residual available capacity and the health state of the distributed modular energy storage system can be further monitored through indexes such as the real-time state of charge and attenuation, and the charge and discharge capacity of the energy storage system can be evaluated.

Description

State estimation method, device and system for distributed modular battery energy storage system

Technical Field

One or more embodiments of the present disclosure relate to the field of modular battery energy storage technologies, and in particular, to a method, an apparatus, and a system for estimating a state of a distributed modular battery energy storage system.

Background

In order to effectively reduce carbon emission, realize the utilization of distributed resources, improve the energy utilization rate of the distributed resources, and have to overcome the inherent characteristics of fluctuation, randomness and the like of the distributed resources, the configuration of the energy storage system is helpful for people to overcome the influence of the aspect of the distributed power supply, the energy storage system can realize time shifting of energy such as photovoltaic wind power and the like, charge energy storage in a period with high wind and light output, discharge load in a power consumption peak, not only be helpful for the utilization of the distributed energy sources, but also generate considerable economic benefits.

However, the existing energy storage technology cannot be widely adopted due to the defects of high technical cost, insufficient performance stability, increased harmonic pollution for a power distribution network and the like, and meanwhile, the traditional energy storage system solution of a small and medium industrial park has the defects of poor balance among different battery units, large battery scale, poor safety, high switching frequency loss, high output waveform with higher harmonic content and the like, and the defects restrict the whole operation performance of the energy storage system, and some of the defects even challenge the safety during operation, and have adverse effects on popularization and utilization of the energy storage system. Based on the traditional energy storage system solution of the middle and small industrial park, a distributed modularized energy storage system is proposed, compared with the traditional solution, the distributed modularized energy storage system controls smaller energy storage units, and the rectifying inversion links of the traditional energy storage system are reduced in a multi-level output mode, so that higher efficiency, lower cost, higher safety and the like are achieved.

Many performance indexes of the distributed modular energy storage system are better than those of the traditional energy storage system, but because the distributed modular energy storage system controls smaller energy storage units, the energy storage units in each battery box only have thirty to forty energy storage units, which presents challenges for the links of state estimation, integral control, active equalization, fault maintenance and the like of the energy storage system, in order to keep the safety during operation, real-time efficient and reasonable estimation and prediction are required for the state of each energy storage unit of each battery box of the distributed modular energy storage system, various indexes such as temperature, voltage and charge state during battery operation are reflected, and the charge state, health state, safety state, functional state, available energy state and the like of the energy storage units are accurately evaluated.

Disclosure of Invention

The invention describes a state estimation method, device and system for a distributed modular battery energy storage system, which can solve the technical problems.

According to a first aspect, a state estimation method of a distributed modular battery energy storage system is provided, which is characterized by being applied to at least one state estimation module, wherein the state estimation module comprises a plurality of measurement modules, a bad data discrimination module, a remote database and a state estimation calculation module; the state estimation calculation module comprises a convolutional neural network module and an improved ampere-hour integration method module; the state estimation module is connected with a battery box, the battery box comprises at least one battery module, the battery module comprises at least one battery unit, and the battery units are distributed; the measuring modules are used for measuring electrical quantity and non-electrical quantity data of the battery unit at the moment t; the bad data discrimination module receives the non-electric quantity and the electric quantity data at the time t, and judges whether bad data exists in the electric quantity and the non-electric quantity data at the time t through the historical data of the remote database and the attenuation coefficient of the battery energy storage obtained by the state estimation calculation module; when bad data exists in the electrical quantity and non-electrical quantity data at the time t, carrying out data correction according to a preset rule to obtain correction data at the time t; the attenuation coefficient of the battery energy storage is obtained through a convolutional neural network module, and the convolutional neural network module determines the attenuation coefficient by receiving the battery electric quantity and non-electric quantity data in a battery box and obtaining the relation between the electric quantity and non-electric quantity data and the residual capacity of a battery unit; the improved ampere-hour integration method module determines the mutual influence factors between the battery units according to the mutual influence factors related to the distance between the battery units, the mutual influence factors related to attenuation and the mutual influence factors related to the current flow direction; and determining the real-time charge state of the battery unit based on the mutual influence factors among the battery units, the attenuation coefficient of battery energy storage and the correction data received by the state estimation calculation module.

In one embodiment, the determining whether there is bad data in the electrical quantity and non-electrical quantity data at the time t; when bad data exists in the electrical quantity and non-electrical quantity data at the time t, carrying out data correction according to a preset rule to obtain correction data at the time t, wherein the method comprises the following steps:

taking the environmental temperature in the electrical quantity and non-electrical quantity data at the time t as a reference, and taking out the historical data which are different from the environmental temperature by a preset degree in the remote database module, wherein the historical data store the correction data before the time t;

summing corresponding voltage and current values in the historical data based on the attenuation coefficient to obtain a data measuring value c influenced by electrical or non-electrical quantity parameters ₁ ；

Obtaining a data measuring value c influenced by non-electrical parameters based on the voltage and current values at the time t-1 and the attenuation rate at the time t ₂ ；

Calculating a value c for the data affected by the electrical parameter ₁ And a data measurement value c affected by a non-electrical parameter ₂ Weighted summation to obtain a parameter theoretical value c _cal ；

Calculating theoretical value of parameter c _cal And the difference probability between the electrical quantity and the non-electrical quantity data at the time t, and judging that bad data exists in the electrical quantity and the non-electrical quantity data at the time t when the interpolation probability is larger than a second preset threshold value;

Based on the correction data at time t-1 and the theoretical value c of the parameter _cal Correcting the bad data to obtain corrected data c at t time _t ′。

In one embodiment, the measured value c of the data affected by the electrical or non-electrical parameters ₁ Respectively of the calculation formulas of (a)The method comprises the following steps:

wherein ε _j For the attenuation rate of the jth battery cell, U _i The positive electrode voltage of the battery unit at the moment i is t ₀ Indicating the starting time, I _i For the current of the battery cell at time i, T _surf·i For i time cell surface temperature, T _envr·i The ambient temperature at time i.

In one embodiment, the training process of the convolutional neural network module includes:

inputting a training data set, wherein the training data set comprises the capacities of the battery units at different moments and the following parameters: operating time, voltage, current, cell surface pressure, ambient temperature, battery surface temperature;

the data corresponding to each parameter in the training set is subjected to de-equalization and normalization to obtain a normalized value of each dimension;

inputting the normalized value and the capacities of the battery units at different moments into a convolution layer for calculation and extraction of characteristic parameters;

inputting the characteristic parameters into an activation layer to activate by adopting an activation function;

Screening the characteristic parameters at a pooling layer to obtain final pooling parameters;

sending the final pooling parameters into a full-connection layer for multi-layer convolution calculation to obtain final characteristic parameters;

and obtaining the attenuation rate of the battery unit based on the final characteristic parameters.

In one embodiment, the interaction factor expression between the battery cells is:

σ _ij ＝σ _dis·ij ·σ _atte·ij ·σ _dir·ij

wherein,σ _atte·ij representing an interaction factor related to attenuation; n is the total number of battery units contained in one battery module; epsilon _t·m The decay rate of the mth battery cell at time t. Sigma (sigma) _dir·ij Indicating a mutual influence factor related to the current flow direction, U _i 、U _j The positive electrode voltage and the negative electrode voltage of the i-th battery cell, respectively.

In one embodiment, the attenuation coefficient is: epsilon _t ＝ε(U ₁ …U _n I ₁ …I _n T _surf1 …T _surfn T _envr1 …T _envrn ) Where ε is the attenuation coefficient, U ₁ …U _n To be the voltage of the battery pack to be tested, I ₁ …I _n For the current of the battery to be tested, T _surf1 …T _surfn To be measured of the surface temperature of the battery cell, T _envr1 …T _envrn Is ambient temperature.

In one embodiment, the state estimation module further includes an external communication module, and the external communication module receives the real-time state of charge of the battery unit and the correction data at the time t and outputs the information.

In one embodiment, the modified data is sent to the remote database, the state estimate calculation module, and the external communication module.

In one embodiment, the remote database is divided into intervals of the ambient temperature and the surface temperature of the battery cell according to a preset degree; the method comprises the steps of storing the voltage, the current and the surface pressure of a battery unit by taking the running time as a label, wherein the specific storage process is as follows:

receiving correction data at the time t-1 and storing the correction data in a storage unit corresponding to the same environment temperature and the surface temperature of the battery cell, wherein the data structure of the correction data is that the time t-1 is taken as a label, and voltage, current and pressure parameters corresponding to the time t-1 are stored;

the correction data at time t is received, the overlay time t-1 corresponds to the stored data of the storage unit corresponding to the same ambient temperature and the cell surface temperature, and the correction data at time t-1 is stored.

In one embodiment, the data includes: electrical quantity data and non-electrical quantity data, wherein the electrical quantity data comprise the voltage and the current of a battery to be tested; the non-electrical quantity data comprise ambient temperature, cell surface temperature of the battery to be tested, running time and cell surface pressure of the battery to be tested.

In one embodiment, the remote database constructs a two-dimensional coordinate system with the ambient temperature and the surface temperature of the battery cell, divides the two-dimensional coordinate system with the interval of 0.2 ℃ and records at least the following information: voltage, current, cell surface pressure; the recorded information is in a format that the running time is taken as a label, and at least the following data under the label are stored: voltage, current, cell surface pressure.

In one embodiment, the operating data of the temperature interval determined by the ambient temperature and the surface temperature of the battery cell only retains the data which is closest to the current time from the currently stored data tag in the temperature interval.

According to a second aspect, a distributed modular battery energy storage system state estimation apparatus is provided. The device comprises: the system comprises at least one state estimation module, a remote database and a state estimation calculation module, wherein the state estimation module comprises a plurality of measurement modules, a bad data screening module and a bad data identification module; the state estimation calculation module comprises a convolutional neural network module and an improved ampere-hour integration method module; the battery box comprises at least one battery module, wherein the battery module comprises at least one battery unit, and the battery units are distributed;

a measurement module configured to measure electrical quantity and non-electrical quantity data at a time t of the battery cell;

the bad data discrimination module is configured to receive the electrical quantity and the non-electrical quantity data at the time t, and judge whether bad data exists in the electrical quantity and the non-electrical quantity data at the time t through the historical data of the remote database and the attenuation coefficient of the battery energy storage obtained by the state estimation calculation module; when bad data exists in the electrical quantity and non-electrical quantity data at the time t, carrying out data correction according to a preset rule to obtain correction data at the time t; the attenuation coefficient of the battery energy storage is obtained through a convolutional neural network module, and the convolutional neural network module determines the attenuation coefficient by receiving the electric quantity and non-electric quantity data of a battery unit in a battery box and obtaining the relation between the electric quantity and non-electric quantity data and the residual capacity of the battery;

A state estimation calculation module configured to determine, by the improved ampere-hour integration module, an interaction factor between the battery cells according to the interaction factor related to the distance between the battery cells, the interaction factor related to the attenuation, and the interaction factor related to the current flow direction; and determining the real-time charge state of the battery unit based on the mutual influence factors among the battery units, the attenuation coefficient of battery energy storage and the correction data received by the state estimation calculation module.

In one embodiment, the apparatus further comprises: and the external communication module is configured to receive the correction data at the time t and the state of charge at the time t and send the correction data and the state of charge at the time t to the corresponding module for data analysis.

According to a third aspect, there is provided a decentralized modular energy storage system, comprising at least one battery box, at least one H-bridge power module and filter inductance, at least one state estimation module and a fault decision module; the at least one battery box is connected with a filter inductor after being connected in series, and the filter inductor is connected into a low-voltage distribution network; the at least one battery box is connected through the at least one H-bridge power module; the at least one H-bridge power module comprises an H-bridge converter circuit and a direct current power supply; the battery box comprises an energy storage sub-module, at least one battery unit and a corresponding liquid cooling device; the state estimation module is connected with the battery box and is responsible for monitoring the energy storage sub-module to obtain the evaluation of the running state and real-time running data; and the fault decision module receives the evaluation of the running state and real-time running data and responds to the fault.

In the method and the device provided by the embodiment of the specification, based on the convolutional neural network and the improved ampere-hour integration method, the attenuation rate of the battery is deduced by training the convolutional neural network in advance to correlate the residual capacity of the battery with the measured parameters; the improved ampere-hour integration method increases the impact on the ambient temperature, the self-decay conditions of the battery cells, and the operation of other battery cells compared to conventional ampere-hour integration methods. By means of the remote database and the neural network training result, the bad data is firstly identified and screened, the bad data is corrected in real time, the corrected operation data and the battery attenuation rate are substituted into an improved ampere-hour integration method for calculation, and the real-time state of charge (SOC) of the battery unit is calculated through the ampere-hour integration method. The problems that when the distributed modular energy storage system controls a plurality of energy storage units, state estimation is tedious and low-efficiency due to a plurality of controlled individuals, bad data cannot be effectively screened by data redundancy, and various states of the energy storage units cannot be estimated are solved.

Drawings

In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the drawings that are needed in the description of the embodiments below are briefly introduced, and it is obvious that the drawings in the following description are only some embodiments of the present invention, and that other drawings can be obtained according to these drawings without inventive effort for a person skilled in the art.

Fig. 1 shows a schematic structural diagram of a distributed modular energy storage system according to an embodiment of the present invention;

fig. 2 shows a basic structure of a battery box and a schematic structure of an H-bridge equalization module according to an embodiment of the present invention;

fig. 3 is a schematic flow chart of a state estimation method of a distributed modular battery energy storage system according to an embodiment of the present invention;

FIG. 4 is a schematic diagram of a state estimation module according to an embodiment of the present invention;

FIG. 5 is a schematic diagram illustrating the working principle of a remote database according to an embodiment of the present invention

FIG. 6 shows a process flow diagram of a convolutional neural network module provided by an embodiment of the present invention;

FIG. 7 shows a block diagram of a convolutional neural network provided by an embodiment of the present invention;

fig. 8 is a schematic structural diagram of a state estimation device for a distributed modular battery energy storage system according to an embodiment of the present invention.

Detailed Description

The following describes the scheme provided in the present specification with reference to the drawings.

In order to make the objects, technical solutions and advantages of the embodiments of the present application more clear, the technical solutions of the embodiments of the present application will be described below with reference to the accompanying drawings.

In the description of embodiments of the present application, words such as "exemplary," "such as" or "for example," are used to indicate by way of example, illustration, or description. Any embodiment or design described herein as "exemplary," "such as" or "for example" is not necessarily to be construed as preferred or advantageous over other embodiments or designs. Rather, the use of words such as "exemplary," "such as" or "for example," etc., is intended to present related concepts in a concrete fashion.

In the description of the embodiments of the present application, the term "and/or" is merely an association relationship describing an association object, and indicates that three relationships may exist, for example, a and/or B may indicate: a alone, B alone, and both A and B. In addition, unless otherwise indicated, the term "plurality" means two or more.

Furthermore, the terms "first," "second," and the like, are used for descriptive purposes only and are not to be construed as indicating or implying relative importance or implicitly indicating an indicated technical feature. Thus, a feature defining "a first" or "a second" may explicitly or implicitly include one or more such feature. The terms "comprising," "including," "having," and variations thereof mean "including but not limited to," unless expressly specified otherwise.

The main structure of the distributed modular energy storage system is provided with a battery box, an H-bridge power module and a filter inductor, wherein the battery boxes which are uniformly distributed in three phases are connected in series and then are connected into the filter inductor at the tail end of a circuit to be connected into a low-voltage distribution network, and the distributed modular energy storage system is different from the traditional battery energy storage system in that the distributed modular energy storage system controls battery units with lower granularity to realize multi-level output, so that an AC-DC conversion link of converting the traditional battery energy storage into direct current is omitted.

However, the distributed modular battery energy storage system manages more battery units, so that the collected data are numerous, different battery units can be mutually influenced, redundancy of the collected data is caused, bad data cannot be effectively distinguished, efficiency is reduced, and difficulty is caused in state estimation of the battery energy storage system. In order to solve the problem of state estimation of a distributed modular battery energy storage system, the embodiment of the invention is based on a convolutional neural network and an improved ampere-hour integration method, and the attenuation rate of a battery is deduced by carrying out associated training on the residual capacity and the measurement parameters of the battery on the convolutional neural network in advance; the improved ampere-hour integration method increases the impact on the ambient temperature, the self-decay conditions of the battery cells, and the operation of other battery cells compared to conventional ampere-hour integration methods. By means of historical data in a remote database and the training result of a neural network, firstly bad data are identified and screened, the bad data are corrected in real time, corrected operation data and battery attenuation rate are substituted into an improved ampere-hour integrating method to calculate, and the real-time state of Charge (SOC) of a battery unit is calculated through the ampere-hour integrating method.

Fig. 1 shows a schematic structural diagram of a distributed modular energy storage system provided by an embodiment of the present invention, as shown in fig. 1, taking an access to a 10kV power distribution network as an example, the energy storage system includes a battery box, an H-bridge power module, and a filter inductor, where the battery box includes a battery module and an energy storage sub-module. The main topological structure is that battery boxes with different phases are connected in series through H-bridge power modules, the tail ends of the battery boxes are connected to a three-phase low-voltage 10kV distribution network through filter inductors, the filter inductors configured at the tail ends of the lines further limit harmonic waves of electric energy finally connected to the low-voltage distribution network, the voltage passing through the filter inductors is closer to sine waves in waveform, the electric energy quality conveyed by the lines is improved, and the damage caused by potential harmonic waves is reduced. The body structure of the energy storage system is that three-phase evenly distributed battery boxes are connected in series, and then the tail end of a circuit is connected with a filter inductor to be directly connected into a low-voltage distribution network, and different battery boxes of the same phase are connected with each other through an H-bridge power module. The filter inductance at the end of the line further processes the electric energy during output, so that the electric energy flowing into the power distribution network has lower harmonic content and higher electric energy quality.

Fig. 2 shows a basic structure of a battery box and a structural schematic diagram of an H-bridge equalization module provided by an embodiment of the present invention, as shown in fig. 2 (a), the battery box includes an energy storage sub-module, a battery module (battery pack) and other modules, the battery module generally includes about several (for example, 24 or 52) lithium ion batteries arranged at the same distance apart in a vertical state in a battery placement area, and liquid cooling devices are distributed between adjacent lithium ion batteries.

The battery box comprises a single battery module, a corresponding water cooling device, a state estimation module and a fault decision module, wherein the single battery module comprises a plurality of battery units, and the energy storage submodule is generally divided into a power electronic switching circuit and a power control submodule.

In one embodiment, the state estimation module and the fault decision module are located outside the battery box, and the state estimation module is connected to the battery cells in the battery box.

As shown in fig. 2 (b), the H-bridge power module is composed of an H-bridge converter circuit and a dc power supply, the dc power supply supplies power to the H-bridge converter circuit, the H-bridge converter circuit generally adopts square wave pulse modulation to eliminate harmonics of specific frequency, and has the functions of electrical isolation, reactive compensation, voltage superposition and multi-level output, and the multi-level output of the H-bridge power module can effectively realize the equalization of output voltages of battery boxes with different phases under abnormal operation conditions, so that the stability of the overall operation of the system is improved, and the system has a strong electrical isolation function.

In addition, the H-bridge power module is used as a multi-level output, the superposition of a plurality of modules can simulate the wanted sine waveform, and compared with the traditional three-level converter, the harmonic content of the output voltage is low. In the daily operation process, the H-bridge power module is used as a rectifier bridge, so that the balance of power among the battery boxes can be realized, and the operation reliability of the energy storage system is improved.

Fig. 3 shows a flow chart of a state estimation method of a distributed modular battery energy storage system according to an embodiment of the present invention, and fig. 4 shows a structure and a data processing schematic of a state estimation module according to an embodiment of the present invention, where, as shown in fig. 3 and fig. 4, the method includes the following steps:

in step S310, several measurement modules are used to measure the electrical quantity and non-electrical quantity data of the battery unit at the time t.

In the embodiment of the application, the state estimation method of the distributed modular battery energy storage system is applied to at least one state estimation module in at least a battery box, wherein the state estimation module comprises a plurality of measurement modules, a bad data discrimination module, a remote database and a state estimation calculation module; the state estimation calculation module comprises a convolutional neural network module and an improved ampere-hour integration method module; the battery box comprises at least one battery module, the battery module comprises at least one battery unit, and the battery units are distributed.

The measurement data of the battery cell to be tested includes: electrical quantity data and non-electrical quantity data, wherein the electrical quantity data comprise the voltage and the current of a battery to be tested; the non-electrical quantity data comprise ambient temperature, cell surface temperature of the battery to be tested, running time and cell surface pressure of the battery to be tested.

In one embodiment, after the voltages of each battery unit in the battery pack at each measurement time are measured one by one, the voltages are summarized as the sum of voltages of the battery packs to be measured, as shown in the following formula (1);

[U ₁ …U _n I ₁ …I _n T _surf1 …T _surfn T _envr1 …T _envrn p ₁ …p _n ∑U](1)

wherein U is ₁ …U _n For the voltages of the n cells contained in the battery cell, I ₁ …I _n T is the current of n cells contained in the battery cell _surf1 …T _surfn For the cell surface temperature, T, of the n cells contained in the cell _envr1 …T _envrn N ambient temperatures.

In one embodiment, the original data at time t is stored in an array, specifically in the form of formula (2):

[U ₁ …U _n I ₁ …I _n T _surf1 …T _surfn T _envr1 …T _envrn ](2)

wherein [ U ] ₁ …U _n ]、[I ₁ …I _n ]、[T _surf1 …T _surfn ]、[T _envr1 …T _envrn ]The battery module is in an array form and finally combined into a new array, wherein the array comprises data of the voltage, the current, the ambient temperature and the surface temperature of the battery cells, and n is the total number of the battery cells contained in the battery module.

In one embodiment, one battery module includes a plurality of battery cells.

In one embodiment, a measurement module performs data acquisition on at least one battery cell.

Step S320, the bad data discrimination module receives the electrical quantity and non-electrical quantity data at the time t, and judges whether bad data exists in the electrical quantity and non-electrical quantity data at the time t through the historical data of the remote database and the attenuation coefficient of the battery energy storage obtained by the state estimation calculation module; when bad data exists in the electrical quantity and non-electrical quantity data at the time t, carrying out data correction according to a preset rule to obtain correction data at the time t; the attenuation coefficient of the battery energy storage is obtained through a convolutional neural network module, and the convolutional neural network module determines the attenuation coefficient by receiving the battery electric quantity and non-electric quantity data in the battery box and obtaining the relation between the electric quantity and non-electric quantity data and the residual capacity of the battery.

The at least one state estimation module sends information to the fault decision module.

Bad data may exist in the original data at the time t due to the influence of factors such as temperature, current, voltage and the like, so that the bad data needs to be screened.

In one embodiment, the voltage and current values are used for judging, and the historical correction data which is different from the temperature interval in the original data at the moment t by a preset degree is taken out for summation calculation. Specifically, the following formula (3)

Wherein ε _j For the attenuation rate of the jth battery cell, U _i The positive electrode voltage of the battery unit at the moment i is t ₀ Indicating the starting time, I _i For the current of the battery cell at time i, T _surf·i For i time cell surface temperature, T _envr·i The ambient temperature at time i. The electrical quantity data comprise the voltage and the current of the battery unit to be tested; the non-electrical quantity data comprise ambient temperature, cell surface temperature of the battery cell to be tested, running time and cell surface pressure of the battery cell to be tested.

Calculating c using equation (4) ₂ ：

c ₂ ＝c _t (1-η _t ) (4)

And finally, calculating theoretical values of different parameters by using a formula (5).

c _cal ＝ω ₁ c ₁ +ω ₂ c ₂ (5)

Wherein c ₁ C is a data measurement value affected by the electric parameter ₂ Measuring and calculating a value for data affected by the non-electrical parameter; omega ₁ And omega ₂ C is ₁ And c ₂ The corresponding weights, preferably, can be taken to be 0.3 and 0.7, respectively.

After obtaining the theoretical value corresponding to the parameter, judging whether the measured value at the time t is bad data according to the formula (6),

wherein c _meas Representing measured values of different parameters, K _set For the set criterion, preferably, K _set Set to 30%.

In order to improve efficiency, the measured parameters are allowed to have a small error due to environmental factors and meter self factors, and the data are not considered to be bad data, so that the overall operation of the system is not adversely affected, and bad data which are far from practical can be efficiently detected.

When judging that the voltage and current values at the time t have bad data according to the formula (6), correcting the original data through the following formula (7):

c _t ′＝ω ₃ c _t-1 +ω ₄ c _cal (7)

wherein c _t ' is corrected data; c _t-1 C is t-1 time data _cal Is a parameter theoretical value; omega ₃ And omega ₄ The weights are preferably 0.3 and 0.7, respectively.

In a specific embodiment, the predetermined number of degrees is plus or minus 0.5 ℃.

In another embodiment, the correction data at time t-1 is taken out and summed using the battery surface temperature, ambient temperature determination. Specifically, the following formula (8)

As with the method of judging by using the voltage and current values, the bad data of the battery surface temperature and the ambient temperature in the data at the time t are determined and corrected, and the details are not repeated here.

In one embodiment, if one of the voltage, the current and the temperature in the data at the time t is judged to be bad data, the corresponding data at the time t needs to be corrected to obtain corrected data at the time t.

In one embodiment, the remote database is updated with data from when the energy storage system is put into operation.

In one embodiment, the history correction data is stored in a remote database.

In one embodiment, when no bad data is detected, the bad data discrimination module sends the data at time t to the remote database, the external communication module, and the state estimation calculation module without modifying the data at time t.

And acquiring measured real-time operation data, wherein the data is correction data after bad data identification, and the storage form of the correction data in a remote database is to store the voltage, the current, the ambient temperature and the surface temperature of the battery cell at the current moment by taking the operation time as a label.

Wherein the historical data is compared with the data collected at the moment of a single sign. After the data collected at the current moment is calculated for one time, the data collected at the current moment becomes the historical data of the next calculation. Colleagues, in turn, store these data in the history database at the same time.

Fig. 5 is a schematic diagram illustrating an operation principle of a remote database according to an embodiment of the present invention, where the remote database stores environmental temperature, surface temperature of a battery, surface pressure of a battery, voltage and current data at each moment, and records voltage, current and surface pressure of the battery of a single decentralized modularized battery energy storage system with running time as a tag, as shown in fig. 5.

In one embodiment, the remote database stores the ambient temperature and the cell surface temperature divided into intervals of a preset degree, preferably 0.2 degrees celsius.

It should be noted that, in order to prevent the data redundancy caused by the excessive collected data, the bad data screening module needs to refer to too much related data, so as to improve the efficiency of bad data screening, the database only keeps one for the data in a single temperature interval, and the running time in the temperature interval is a string of data which is closest to the current running time, so that the array in the database takes the running time as a label.

Fig. 6 shows a process flow chart of a convolutional neural network module provided by the embodiment of the present invention, and fig. 7 shows a structure diagram of the convolutional neural network provided by the embodiment of the present invention, where, as shown in fig. 6 and fig. 7, the convolutional neural network includes a data input layer, a convolutional calculation layer, a ReLU excitation layer, a pooling layer and a full connection layer, and by inputting a data set to perform training continuously, the characteristics of the correlation between different data can be obtained by calculation of a convolutional kernel.

In the embodiment of the invention, the training set required by the convolutional neural network training is parameters such as voltage, current, ambient temperature, surface temperature and the like obtained in advance when the convolutional neural network is operated by a battery. The data in the training set extracts the array of raw data, as shown in equation (2) above.

In one embodiment, the remaining capacity of the battery pack is calculated for each measurement instant, whereby the correlations of voltage, current, temperature and remaining capacity are derived by convolving the neural network.

In one embodiment, the data of the battery attenuation rate eta arranged in time sequence is produced according to the final residual capacity at the measuring moment by the obtained correlation function of the voltage, the current, the temperature and the residual capacity.

Inputting a training set into a convolutional neural network for calculation, and firstly, performing de-equalization and normalization in a data input layer, wherein the de-equalization formula (9) and the normalization formula (10) are used for performing de-equalization;

wherein x represents the voltage, current, ambient temperature and cell surface temperature in the training set, n represents the number of data in the data set, x _min For the minimum value of each parameter, x _max For the maximum value of each parameter, x ₀ For each parameter t ₀ A value of the time of day.

Thus, the minimum normalized value of each parameter is calculated by the formula (9) and the formula (10).

In one embodiment, the maximum normalized value for each parameter is taken according to different situations, see equation (11).

After the data in the training set is initially processed, various characteristics of the finally obtained data are calculated through a full-connection layer through the combination of a plurality of convolution layers, an excitation layer and a pooling layer, and then the interrelationship among voltage, current, temperature and attenuation coefficients is obtained.

After the training set data are subjected to mean removal and normalization in the data processing layer, the processed data are input into the convolution layer, and the convolution layer is responsible for extracting characteristic parameters through calculation of convolution kernels.

In one embodiment, the operation is performed using a 4 x 4 convolution kernel, the four rows of which represent voltage, current, ambient temperature, and surface temperature, respectively. The diagonal elements represent their own effects on attenuation, and the off-diagonal elements represent interactions between the two dimensions corresponding to the line.

In one embodiment, the convolution layers are 4 to 5 layers, and after convolution operation, a matrix with characteristic parameters of 4×4 is obtained, which correspond to the respective influences and mutual influences of the voltage current at the specific running time and the surface temperature of the ambient temperature.

In one embodiment, the activation layer activates the computational element using a relu activation function.

The pooling layer simplifies the parameters obtained by the convolution layer and screens the characteristic parameters.

Training the convolutional neural network to finally obtain a functional relation corresponding to the formula (12):

S _rest·t ＝S(U ₁ …U _n I ₁ …I _n T _surf1 …T _surfn T _envr1 …T _envrn ) (12)

wherein S is _rest Residual capacity of single cell, S _rest·t Is the remaining capacity of the battery energy storage at time t.

Calculating the decay rate over time according to equation (13):

ε _t ＝ ε(U ₁ …U _n I ₁ …I _n T _surf1 …T _surfn T _envr1 …T _envrn ) (14)

where ε is the attenuation coefficient, U ₁ …U _n To be the voltage of the battery pack to be tested, I ₁ …I _n For the current of the battery to be tested, T _surf1 …T _surfn To be measured of the surface temperature of the battery cell, T _envr1 …T _envrn Is ambient temperature.

Step S230, the improved ampere-hour integration method module determines the mutual influence factors between the battery units according to the mutual influence factors related to the distance between the battery units, the mutual influence factors related to attenuation and the mutual influence factors related to the current flow direction; and determining the real-time state of charge of the battery unit to be tested based on the mutual influence factors among the battery units, the attenuation coefficient of battery energy storage and the correction data received by the state estimation calculation module.

After the attenuation condition of the battery energy storage is analyzed through the convolutional neural network, the final calculation is completed by adopting an ampere-hour integration method, wherein the general ampere-hour integration method is shown as a formula (15), and the SOC is shown in the formula ₀ The SOC is respectively the initial charge state and the end charge state of the energy storage in a single time period; t is t ₀ T is the starting time and the ending time respectively; s is S _N Rated capacity for storing energy of the battery; i is the current stored by the battery; η is the calibrated charge-discharge efficiency, and charge is negative and discharge is positive.

But the ampere-hour integration method is directly used, and the environment temperature during charge and discharge and the attenuation condition of the battery during multiple charge and discharge cannot be reflected. In a decentralized modular battery energy storage system, the effect of other battery packs on one battery pack should also be considered, embodied as other battery pack charge-discharge and accumulation effects.

The modified ampere-hour integration method is formula (16),

wherein SOC is _i 、SOC _i-1 The states of charge at the moment i and the moment i-1 of battery energy storage are respectively; t is t ₀ T respectively correspond to the moment i-1 and the moment i; s is S _N Rated capacity for storing energy of the battery; epsilon _i·m The attenuation rate of the mth battery unit in the ith period; t (T) _i And T _i-1 The time corresponding to the moment i and the moment i-1 respectively; n is the total number of battery units contained in one battery module; i is the current stored by the battery; sigma (sigma) _mj Is the interaction factor between cell m and cell j; eta is the calibrated chargeDischarge efficiency, charge to negative discharge to positive.

Further, will sigma _mj M is replaced by i, sigma _ij Indicating the factors of interaction between the battery cells.

The following is sigma _ij The attenuation of the energy storage system in a single battery box is related to all the single battery units connected in series and parallel, and for the battery units with slightly higher capacity attenuation, the influence of the attenuation on other battery packs is larger at the same distance, and sigma _ij The distance of different battery units is considered, because the charge and discharge process with a short distance can be more influenced than the charge and discharge process with a long distance, and then the battery attenuation at a specific moment is considered, the distance is calibrated by the residual available capacity, and the distance is expressed by the following formula (17):

further, the attenuation-related expression (18) is expressed as:

in sigma _atte·ij Representing an interaction factor related to attenuation; n is the total number of battery units contained in one battery module; epsilon _t·m The decay rate of the mth battery cell at time t.

Further, since the current flow is unidirectional, the influence of the current flowing first is larger than that of the current flowing later, the factor of the current is represented by the voltage, the higher the voltage is, the current flows out from it, and the coefficient representing the direction is represented by the formula (19)

In sigma _dir·ij Indicating a mutual influence factor related to the current flow direction, U _i 、U _j Respectively isPositive and negative voltages of the i-th battery cell.

Further, a final sigma is obtained _ij As shown in formulas (20) - (21):

σ _ij ＝σ _dis·ij ·σ _atte·ij ·σ _dir·ij (20)

by adopting the improved ampere-hour integration method, the real-time state of charge (SOC) of the distributed modular energy storage system can be obtained, and the residual available capacity and health state of the distributed modular energy storage system are further monitored through indexes such as SOC, attenuation and the like, and the charge and discharge capacity of the energy storage system is evaluated. The improved ampere-hour integration method is used for calculating, so that the high efficiency is ensured, and the effects of the environmental temperature, the attenuation of the stored energy and the attenuation of other battery units which cannot be considered by the traditional ampere-hour integration method are considered, so that the state evaluation method is more accurate and reasonable.

The present invention will be further described with reference to fig. 2-4.

The state estimation module and the fault decision module are responsible for monitoring and controlling the energy storage system, continuously sending the running state of the energy storage system to the outside, the state estimation module is connected with the battery box and reflects the physical characteristics of the battery box running in real time, the fault decision module receives the evaluation and real-time running data of the state estimation module on the running state, early warning the fault and timely responding to the fault when the fault occurs, and the safety and stability of the running of the energy storage system are maintained through the combined action of the state estimation module and the fault decision module.

The technical scheme of the embodiment of the invention is applied to the state estimation module, the state estimation module can obtain the real-time running state of the energy storage subsystem and evaluate the running state, and the state estimation module is placed in the battery box and comprises a measurement module, a state estimation calculation module, a bad data discrimination module, a remote database and an external communication module. The method comprises the following steps:

(1) The measurement module mainly monitors non-electric quantity and electric quantity in a double-layer mode, wherein the non-electric quantity mainly comprises ambient temperature, battery surface temperature, pressure and operation time, the electric quantity mainly comprises voltage and current of a battery unit, and each time node of the measurement module sends the measured non-electric quantity and electric quantity to the bad data screening module for analysis in an array mode.

(2) The state estimation calculation module is responsible for the main function of the state estimation module, is responsible for carrying out specific operation on the state estimation of the distributed modular energy storage system, adopts compound state estimation, adopts a mode of combining a convolutional neural network and an improved ampere-hour integration method, and is trained by the convolutional neural network to obtain the attenuation conditions of battery energy storage under the condition of different environmental temperatures in the length of running time, wherein the attenuation condition data are main evidence for screening by the bad data screening module, and the ampere-hour integration method is simple and quick in solving, but the self algorithm is not regulated when the conditions such as the temperature change due to the fact that the attenuation conditions and the temperature conditions of the battery are insensitive, so that the accuracy of the scheme is lower due to poor adaptability although the method is simple and convenient to calculate, besides the optimization and improvement on the ampere-hour integration method based on the neural network, the method is modularized energy storage, a plurality of individuals are controlled, and the interaction influence among different individuals including influence on the state of charge SOC and the attenuation degree of the self-body is considered, and the accuracy of the state estimation result can be ensured to the greatest extent.

(3) The bad data discrimination module receives the arrays sent by different measurement modules, sorts the arrays into a matrix, judges the authenticity of the data by comparing the existing data with the historical data in the remote database, and judges the attenuation condition of the battery energy storage and the data in the remote database obtained by the state estimation calculation module as criteria, if the bad data are considered as bad data, the bad data discrimination module corrects the bad data according to a pre-specified principle, and in order to ensure the efficiency, the bad data discrimination module allows the measured non-electric quantity and the electric quantity to float within an acceptable range, so as to remove the data with obvious abnormality.

(4) The remote database stores the operation parameters, various electric quantities and non-electric quantities of the battery energy storage system in a specific form, and adopts a strategy of timing update, and is also used as a reference of a bad data discrimination link. The external communication module is responsible for transferring information, sending the operation information of the energy storage system to the fault decision module or a remote dispatching center, and further referring and deciding by the power supply energy storage system or an external mechanism. The fault decision module is connected with the state estimation module, receives information output from at least one state estimation module and responds to faults

The invention also provides a device corresponding to the method provided by the invention. Fig. 8 is a schematic structural diagram of a state estimation device of a distributed modular battery energy storage system according to an embodiment of the present disclosure. As shown in fig. 8, the apparatus 800 includes:

the measurement module 810 is configured to measure non-charge and charge data at time t of the battery cell. The bad data discrimination module 820 is configured to receive the non-electric quantity and electric quantity data at the time t, and determine whether bad data exists in the non-electric quantity and electric quantity data at the time t according to the historical data of the remote database and the attenuation coefficient of the battery energy storage obtained by the state estimation calculation module; when bad data exists in the non-electric quantity and electric quantity data at the time t, carrying out data correction according to a preset rule to obtain correction data at the time t; the attenuation coefficient of the battery energy storage is obtained through a convolutional neural network module, and the convolutional neural network module determines the attenuation coefficient by receiving the battery electric quantity and non-electric quantity data in the battery box and obtaining the relation between the electric quantity and non-electric quantity data and the residual capacity of the battery.

A state estimation calculation module 830 configured to determine, by the modified ampere-hour integration module, a mutual influence factor between the battery cells according to a mutual influence factor related to a distance between the battery cells, a mutual influence factor related to attenuation, and a mutual influence factor related to a current flow direction; and determining the real-time charge state of the battery unit based on the mutual influence factors among the battery units, the attenuation coefficient of battery energy storage and the correction data received by the state estimation calculation module.

In one embodiment, the apparatus 800 further includes an external communication module 840 for receiving the real-time state of charge of the battery cells and the corrected data and sending the same to the fault decision module for analysis.

It should be noted that, for the description of the apparatus in fig. 8, reference may also be made to the description of the foregoing method.

The invention also provides a distributed modular energy storage system, as shown in fig. 1 and 2, comprising at least one battery box, at least one H-bridge power module, a filter inductor, at least one state estimation module and a fault decision module; the at least one battery box is connected with a filter inductor after being connected in series, and the filter inductor is connected into a low-voltage distribution network; the at least one battery box is connected through the at least one H-bridge power module; the at least one H-bridge power module comprises an H-bridge converter circuit and a direct current power supply; the battery box comprises an energy storage sub-module, at least one battery unit and a corresponding liquid cooling device; the state estimation module is connected with the battery box and is responsible for monitoring the energy storage sub-module to obtain the evaluation of the running state and real-time running data; and the fault decision module receives the evaluation of the running state and real-time running data and responds to the fault.

According to an embodiment of another aspect, there is also provided a computer-readable storage medium having stored thereon a computer program which, when executed in a computer, causes the computer to perform the method described in connection with fig. 3.

According to an embodiment of yet another aspect, there is also provided a computing device including a memory having executable code stored therein and a processor that, when executing the executable code, implements the method described in connection with fig. 3. Those skilled in the art will appreciate that in one or more of the examples described above, the functions described in the present invention may be implemented in hardware, software, firmware, or any combination thereof. When implemented in software, these functions may be stored on or transmitted over as one or more instructions or code on a computer-readable medium.

The foregoing embodiments have been provided for the purpose of illustrating the general principles of the present invention in further detail, and are not to be construed as limiting the scope of the invention, but are merely intended to cover any modifications, equivalents, improvements, etc. based on the teachings of the invention.

Claims

1. The state estimation method of the distributed modular battery energy storage system is characterized by being applied to at least one state estimation module, wherein the state estimation module comprises a plurality of measurement modules, a bad data discrimination module, a remote database and a state estimation calculation module; the state estimation calculation module comprises a convolutional neural network module and an improved ampere-hour integration method module; the state estimation module is connected with a battery box, the battery box comprises at least one battery module, the battery module comprises at least one battery unit, and the battery units are distributed;

the measuring modules are used for measuring electrical quantity and non-electrical quantity data of the battery unit at the moment t;

the bad data discrimination module receives the non-electric quantity and the electric quantity data at the time t, and judges whether bad data exists in the electric quantity and the non-electric quantity data at the time t through the historical data of the remote database and the attenuation coefficient of the battery energy storage obtained by the state estimation calculation module; when bad data exists in the electrical quantity and non-electrical quantity data at the time t, carrying out data correction according to a preset rule to obtain correction data at the time t; the attenuation coefficient of the battery energy storage is obtained through a convolutional neural network module, and the convolutional neural network module determines the attenuation coefficient by receiving the battery electric quantity and non-electric quantity data in a battery box and obtaining the relation between the electric quantity and non-electric quantity data and the residual capacity of a battery unit;

The improved ampere-hour integration method module determines the mutual influence factors between the battery units according to the mutual influence factors related to the distance between the battery units, the mutual influence factors related to attenuation and the mutual influence factors related to the current flow direction; and determining the real-time charge state of the battery unit based on the mutual influence factors among the battery units, the attenuation coefficient of battery energy storage and the correction data received by the state estimation calculation module.

2. The method of claim 1, wherein said determining if there is bad data in the electrical quantity and non-electrical quantity data at time t; when bad data exists in the electrical quantity and non-electrical quantity data at the time t, carrying out data correction according to a preset rule to obtain correction data at the time t, wherein the method comprises the following steps:

Calculating a value c for the data affected by the electrical parameter or the non-electrical parameter ₁ And a data measurement value c affected by a non-electrical parameter ₂ Weighted summation to obtain a parameter theoretical value c _cal ；

based on the correction data at time t-1 and the theoretical value c of the parameter _cal Correcting the bad data to obtain the correction at the time tData c _t ′。

3. The method according to claim 2, characterized in that the data measured value c affected by electrical or non-electrical parameters ₁ The calculation formulas of (a) are respectively as follows:

wherein ε _j For the attenuation rate of the jth battery cell, U _i The positive electrode voltage of the battery unit at the moment i is t ₀ Indicating the starting time, I _i For the current of the battery cell at time i, T _surf·i For i time cell surface temperature, T _envr·i The ambient temperature at the moment i; the electrical quantity data comprise the voltage and the current of the battery to be tested; the non-electrical quantity data comprise ambient temperature, cell surface temperature of the battery to be tested, running time and cell surface pressure of the battery to be tested.

4. The method of claim 1, wherein the training process of the convolutional neural network module comprises:

the data corresponding to each parameter in the training set is subjected to de-equalization and normalization to obtain a normalized value of each parameter;

5. The method of claim 1, wherein the interaction factor expression between the battery cells is:

σ _ij ＝σ _dis·ij ·σ _atte·ij ·σ _dir·ij

6. The method of claim 1, wherein the attenuation coefficient is: epsilon _t ＝ε(U ₁ …U _n I ₁ …I _n T _surf1 …T _surfn T _envr1 …T _envrn ) Where ε is the attenuation coefficient, U ₁ …U _n For the voltage of the battery cell to be measured, I ₁ …I _n For the current of the battery cell to be measured, T _surf1 …T _surfn To be measured of the surface temperature of the battery cell, T _envr1 …T _envrn Is ambient temperature.

7. The method of claim 1, wherein the remote database is partitioned by ambient temperature and cell surface temperature by a predetermined number of degrees; the method comprises the steps of storing the voltage, the current and the surface pressure of a battery unit by taking the running time as a label, wherein the specific storage process is as follows:

and receiving correction data at the time t, wherein the covering time t-1 corresponds to storage data of a storage unit corresponding to the same environment temperature and the cell surface temperature, and storing the correction data at the time t-1 into a history database.

8. The method of claim 1, wherein the remote database builds a two-dimensional coordinate system with ambient temperature and cell surface temperature, is partitioned with 0.2 degrees celsius as a zone, and records the at least the following information: voltage, current, cell surface pressure; the recorded information is in a format that the running time is taken as a label, and at least the following data under the label are stored: voltage, current, cell surface pressure.

9. The state estimation device of the distributed modular battery energy storage system is characterized by comprising at least one state estimation module, wherein the state estimation module comprises a plurality of measurement modules, a bad data discrimination module, a remote database and a state estimation calculation module; the state estimation calculation module comprises a convolutional neural network module and an improved ampere-hour integration method module; the battery box comprises at least one battery module, wherein the battery module comprises at least one battery unit, and the battery units are distributed;

10. The distributed modular energy storage system is characterized by comprising at least one battery box, at least one H-bridge power module, a filter inductor, at least one state estimation module and a fault decision module;

the at least one battery box is connected with a filter inductor after being connected in series, and the filter inductor is connected into a low-voltage distribution network;

the at least one battery box is connected through the at least one H-bridge power module; the at least one H-bridge power module comprises an H-bridge converter circuit and a direct current power supply;

the battery box comprises an energy storage sub-module, at least one battery unit and a corresponding liquid cooling device;

the state estimation module is connected with the battery box and is responsible for monitoring the energy storage sub-module to obtain the evaluation of the running state and real-time running data;

And the fault decision module receives the evaluation of the running state and real-time running data and responds to the fault.