DE102020129136A1 - Method and system for reducing the runtime-relevant complexity of a real-time control of a modular multilevel converter - Google Patents

Abstract

The invention relates to a method for reducing runtime-relevant complexity of a control algorithm of an AC battery, in which the AC battery comprises a central controller with a scheduler and at least one line with a plurality of battery modules, in which the scheduler is divided into a real-time and an offline part respective overall switching state (140) of the plurality of battery modules is specified for each switching cycle, in which in the real-time part for each switching cycle a switching state table, which contains all overall switching states together with a value assigned by a cost function for a conversion of the respective overall switching state, is accessed and from this the respective overall switching state with the smallest value of the cost function is formed, the cost function for all overall switching states of the least ns of a string is calculated by considering a respective string of the at least one string with a plurality of battery modules as divided into smaller sub-string sections with a respective sub-switching state (110, 120, 130), the cost function is calculated for all sub-switching states, the respective sub-switching states together are stored in a directory with the respective value of the cost function, and the value of the cost function of the respective overall switching state is composed of the values of the cost function for sub-switching states of the sub-strand sections forming the respective strand of the at least one strand. A circuit corresponding to the method is also claimed.

Description

The present invention relates to a reduction in runtime-relevant complexity in a calculation of switching positions in real-time control of a modular multilevel converter. A system is also claimed on which the method is implemented.

In order to obtain an AC voltage of a predetermined frequency from a DC voltage, input and output voltages are switched between a few levels, usually two to three, in conventional power electronics with a few power switches, with a desired value then being obtained on average over time. In contrast to this, modern multilevel converters work according to a scheme that generates the alternating voltage through a dynamically changeable configuration of energy storage devices, such as capacitors or energy cells, connected via a large number of electronic switches. Considerably more electronic switches, e.g. power semiconductor switches, are used here than in conventional power electronics, from which a large number of realizable overall switching states and associated output voltages of the multilevel converter can be formed in very fine gradations or levels. At the same time, there is a large number of degrees of freedom for realizing the same voltage level with different overall switching states, with a controller of the multilevel converter having to allocate a clear switching state to all switches at all times.

In principle, a respective design of a multilevel converter can be defined on the basis of an individual module, which is lined up in multiples for a respective strand per phase, for example shown in the publication DE 10 2015 112 512 A1 . Depending on the circuit arrangement of the switching elements included in an individual module, all switching states relating to the connection of the energy storage device that is also included can be achieved independently of additional connected, structurally identical individual modules. A central multilevel converter in this sense is a modular multilevel converter with serial and parallel connectivity (MMSPC), e.g. described in SM Goetz, AV Peterchev and T. Weyh, "Modular Multilevel Converter With Series and Parallel Module Connectivity: Topology and Control," in IEEE Transactions on Power Electronics, vol. 30, no. 1, pp. 203-215, Jan. 2015. The MMSPC meets high standards for voltage requirements for a variety of applications, e.g. as a traction battery to power an electric motor in vehicles. An additional degree of freedom of connecting individual modules of the MMSPC in parallel compared with a conventional modular multilevel converter (MMC) advantageously reduces source impedance losses, but significantly increases the complexity of connection options.

Simple control algorithms of the MMSPC try to reduce the complexity as much as possible in order to ensure real-time control, but largely omit optimization criteria for circuit specifications such as output voltage, charge balancing, temperature management or losses in the system. On the other hand, optimizing control algorithms can take various optimization criteria into account, which is described, for example, in Stefan M. Goetz; Zhongxi Li; Xinyu Liang; Chengduo Zhang; Srdjan M. Lukic; Angel V. Peterchev, "Control of Modular Multilevel Converter With Parallel Connectivity - Application to Battery Systems", IEEE Transactions on Power Electronics, vol. 32, no. 11, pp. 8381-8392. In this case, a number of the optimization criteria corresponds to a complexity of the optimizing control algorithm, which is noticeable in the computing effort and is therefore relevant to the runtime. Therefore, as complexity increases, the control algorithm loses real-time control capability.

Recently proposed hybrid algorithms represent a compromise between simple control algorithms and optimizing control algorithms. On the one hand, hybrid algorithms separate the computing effort for switching specifications into a relatively slow optimizing offline part, which carries out a thorough trawl through all possible overall switching states, evaluates them using a cost function and then updates a switching table with it , on the other hand in a real-time part, which only accesses the switching table in order to determine an overall switching state.

So is from Zhongxi Li; Ricardo Lizana; Angel V Peterchev; Stefan M. Goetz, "Predictive control of modular multilevel series/parallel converter for battery systems" in 2017 IEEE Energy Conversion Congress and Exposition (ECCE); conference papers; 2017, a multi-level predictive control method for MMSPCs is known. Instead of switching signals, the control method uses transistor currents as basic variables. The multilevel converter is thus mapped to a linear time-invariant system whose associated cost function has a linear quadratic form. This allows several steps to be calculated in advance.

Also in print DE 10 2018 125 728.7 a scheduling for the provision or calculation of switching positions is separated into offline and real-time parts. Both parts are executed in parallel on the controller's processing units, while they update each other (overall switching states completed by the real-time part affect the charge states of individual modules and thus change the cost function in the offline part).

A range of such hybrid algorithms depends on the computing effort of the optimizing offline part. A rather sporadic update of the switching table caused by this can be sufficient for a controlled charge equalization in the long run, but causes an overload of individual modules again and again in the short term. In addition, the computing effort increases exponentially with the number of modules for the same complexity.

Against this background, it is an object of the present invention to present a method for determining optimized overall switching states of modular multilevel converters, in which an overloading of individual modules due to belated updating of the switching table is avoided. In connection with this, the runtime-relevant computing effort of the optimizing offline part for the provision of switching positions evaluated according to optimization criteria should be reduced. Furthermore, a system is to be provided with which the method can be implemented.

A method for reducing the complexity of a control algorithm of an AC battery that is relevant to the runtime is proposed in order to solve the above-mentioned task, in which the AC battery comprises a central controller with a scheduler and at least one line with a plurality of battery modules. A respective battery module of the plurality of battery modules has at least one energy store and a plurality of power semiconductor switches, which connect the respective battery module in series or in parallel or in bypass to another battery module. A phase voltage is specified for the scheduler by the central controller, which leads to a desired phase current at an output of the AC battery. To realize the phase voltage, the scheduler specifies a respective overall switching state of the plurality of battery modules for each switching cycle, with the scheduler being divided into a real-time part and an offline part. In the real-time part, a switching state table is accessed for each switching cycle, which contains all the overall switching states of the at least one line together with a value assigned by a cost function for converting the respective overall switching state, and from this the respective overall switching state with a smallest one is calculated for the at least one line Value of the cost function formed. The overall switching state with the smallest value of the cost function corresponds to an optimal choice of switching states in the at least one line according to criteria for a respective state of the respective battery modules, which form the basis of the cost function. In the offline part, the cost function for all overall switching states of the at least one string is calculated in a continuous sequence using a memoization algorithm by considering a respective string of the at least one string with a plurality of battery modules as divided into smaller substring sections with a respective subswitching state, the cost function to all sub-switching states are calculated, the respective sub-switching states are stored in a directory together with the respective value of the cost function, and the value of the cost function of the respective overall switching state is composed of the values of the cost function for sub-switching states of the sub-strand sections forming the respective strand of the at least one strand.

A respective combination of the respective sub-switching states stored in the directory thus leads to a respective value of the cost function of the respective overall switching state. If one considers all overall switching states in a vector space spanned by these overall switching states, then the calculation of the cost function corresponds to a transformation of an element of this vector space into a space of state costs. In this way of looking at it, the memoization algorithm of the method according to the invention finds a basis in the vector space of the total switching states and transforms this basis into the space of the state costs. Accordingly, by simply combining the transformed basis, the state cost space can be spanned.

The scheduler sets the phase voltage on the at least one strand or the at least one phase, and the phase current is then set according to a connected load. At a higher control level (software in the microcontroller), current control actually takes place with the respective phase voltage as the manipulated variable. The switching state of the respective battery modules is controlled by means of scheduling and the respective phase voltage is thereby directly regulated, but the phase current is only indirectly regulated.

In one embodiment of the method according to the invention, the cost function is formed using a criterion from the following list: temperature of the respective battery module, ripple of respective battery module currents.

Verbunden mit der Berechnung von Gl. (1) ist damit eine Kenntnis der jeweiligen Batteriemodulströme zu dem jeweiligen Gesamtschaltzustand, bezeichnet auch als eine jeweilige Zustandsstromverteilung. Obwohl die jeweilige Zustandsstromverteilung zu jedem Gesamtschaltzustand völlig individuell ist, zeigen sich doch strenge Regelmäßigkeiten, welche in nachfolgenden Ausführungsformen vorteilhaft zur Vereinfachung in der Berechnung der Kostenfunktion Gl. (1) verwendet werden.In another embodiment of the method according to the invention, the cost function is formed as a sum over respective battery module currents i _k , which are weighted by a respective deviation Δq _k = (SoC _k - SOC _AV ) of a respective battery module state of charge SoC _k compared to an average battery state of charge SOC _AV . For an example with six battery modules (N=6) in a respective string of at least one string, the cost function S _Bsp is: $$S_{Bsp}={\mathrm{\Sigma }}_{k=1}^{N=6}{i}_k\cdot \Delta q_k=i_1\cdot \Delta q_1+i_2\cdot \Delta q_2+i_3\cdot \Delta q_3+i_4\cdot \Delta q_4+i_5\cdot \Delta q_5+i_6\cdot \Delta q_6.$$

Associated with the calculation of Eq. (1) is thus knowledge of the respective battery module currents for the respective overall switching state, also referred to as a respective state current distribution. Although the respective state current distribution for each overall switching state is completely individual, there are strict regularities which are advantageous in the following embodiments to simplify the calculation of the cost function Eq. (1) are used.

In a further embodiment of the method according to the invention, a respective substring section is formed by battery modules connected in parallel with one another. In this case, the same phase current flows through all sub-strand sections of the respective strand of the at least one strand, since the sub-strand sections are connected to one another in series within the respective strand. Thus, while the respective phase current flowing through the respective string flows through all the respective substring sections connected to one another in series, it is distributed within the respective substring sections to the battery modules respectively comprised by them.

Within a respective string, a respective sub-string section is subjected to the same load, with the respective load corresponding to the respective phase current in the respective string. Each substring section with battery modules connected in parallel carries the entire proportion of the respective phase current and then distributes this to its battery modules connected in parallel.

In a further developed embodiment of the method according to the invention, a distribution of the phase current divided among battery modules that are respectively connected in parallel with one another in the respective substring section is described by respective distribution coefficients.

The distribution of the phase current distributed in a respective sub-string section to the battery modules connected in parallel with each other is completely independent of battery modules of other sub-string sections or independent of the respective distributions of the respective phase current in other sub-string sections of the respective string of the at least one string, since all sub-string sections of the respective strand of the at least one strand of the same respective phase current flows. The independence means that the distribution of the respective phase current does not have to be dealt with at the level of the respective branch of the at least one branch, but that the respective distribution can be formed individually for each sub-strand section. According to the invention, this is achieved by introducing distribution coefficients for the respective substring section, the distribution coefficients describing the respective proportion of the phase current flowing through the respective battery module of the battery modules connected in parallel.

The method according to the invention thus makes advantageous use of the fact that, in order to calculate the cost function, the respective overall switching state can be broken down into mutually independent sub-switching states. A number of possible sub-switching states is significantly smaller than a number of possible overall switching states.

mit dem Batteriemodulstrom i_k=I_phase·d_k (mit k=3,4,5) und I_phase als der in dem jeweiligen Strang des mindestens einen Stranges fließende Phasenstrom.If, in the above example with six battery modules, for example the third, fourth and fifth battery module are connected in parallel to form a sub-string section, a cost function S _3-5 for this sub-string section with the respective distribution coefficients d _k can be written as: $$s_{3-5}=i_3\cdot \Delta q_3+i_4\cdot \Delta q_4+i_5\cdot \Delta q_5=I_{pHase}\cdot \left({\mathrm{i.e}}_3+\Delta q_3+{\mathrm{i.e}}_4+\Delta q_4+{\mathrm{i.e}}_5+\Delta q_5\right).$$

with the battery module current i _k =I _phase *d _k (with k=3,4,5) and I _phase as the phase current flowing in the respective phase of the at least one phase.

The distribution coefficients d _k are purely scalar quantities with a value range between zero and one. A value d _k =0 would mean that the kth battery module is bypassed, while a value d _k =1 flows through the kth battery module through the full portion of the phase current in the respective string. Within the respective sub-string section, the distribution coefficients add up to one (d ₃ +d ₄ +d ₅ =1 in the example above), which corresponds to the fact that the full respective phase current flows through the sub-string section under consideration. A functional description of a distribution coefficient is determined by a number of parameters, such as the respective battery module voltage, the internal resistance of a respective battery module, or a ratio between the internal resistance and the module connection resistance.

wobei zusätzlich der Phasenstrom I_phase normalisiert wurde.In an even further developed embodiment of the method according to the invention, in the case of battery modules of the respective string section which have complete charge equalization, the phase current is evenly divided between the respective battery modules which are connected in parallel with one another. As a result, the respective distribution coefficients of the respective substring section assume the same constant value, the value being inversely proportional to the number of battery modules of the respective substring section. In the example above, this means d _k = I _phase /N, so that the cost function s _3-5 for the substring section of the example can be reduced to (N=3): $$s_{3-5}=I_{pHase}\cdot \frac{1}{N}\left(\Delta q_3+\Delta q_4+\Delta q_5\right)\approx \frac{1}{3}\left(\Delta q_3+\Delta q_4+\Delta q_5\right).$$

where the phase current I _phase was also normalized.

The cost function S _Bsp for a respective overall switching state then results from the sum of the values of the cost functions of the substring sections forming the respective overall switching state. If, in the example above, a substring section with cost function s _1-2 is formed with the first and second battery module and a substring section with cost function s _6-6 is formed with the sixth battery module, the cost function S _Bsp results in: $$S_{Bsp}=s_{1-2}+s_{3-5}+s_{6-6}.$$

The independence of the substring sections forms the decisive advantage of the method according to the invention, namely reducing the complexity of the control algorithm that is relevant to runtime. This advantageously makes it possible to list all parallel connection options of the sub-string sections together with their value of the associated cost function in a list, the number of which is advantageously smaller than a number of all connection options of a respective string of the at least one string. Access to values already calculated in the directory, which can be carried out as often as required, creates a significant time advantage for the memoization algorithm according to the invention.

A complexity T of the control algorithm describes a computing time or a computing effort for an execution of the control algorithm. In this case, the computing time is proportional to a number of computing operations for calculating the cost function. In a naive view of an undivided string, the complexity T _string is proportional to a number of terms in the calculation of the cost function, which corresponds to a number N of battery modules in the string, and the number of possible switching states: $$T_{St\mathrm{right}anG}=2N\cdot {\mathrm{\Sigma }}_{k=0}^N\left(\begin{array}{c}N\\ k\end{array}\right)=2N\cdot 2^N\approx 2^N.$$

The complexity T _strand of an undivided strand thus grows approximately exponentially.

In contrast, the complexity Tmem for the memoization algorithm of the method according to the invention advantageously shows a merely cubic increase with a number N of battery modules within a respective string of the at least one string of the AC battery: $$T_{mem}={\mathrm{\Sigma }}_{k=0}^N\left(k\cdot \left(N+1-k\right)\right)=\frac{N^3+3N^2+2N}{6}\approx N^3.$$

Since the memoization algorithm lists all substring sections with battery modules connected in parallel together with their respective value of the cost function, the computational effort is proportional to the number of possible substring sections and a number of terms when calculating the costs function for the respective substring section, with the number of terms corresponding to the number of battery modules in the substring section.

Due to the only cubic increase in complexity and thus the computational effort, the method according to the invention can advantageously also be used to control larger strings of battery modules in real time, while the naive approach here fails due to the computational effort. The method according to the invention also enables further investigations into proposed overall switching states on the basis of precalculations of various switching profiles. Such "prediction base scheduling" is particularly time-critical, since system parameters (e.g. state of charge) change with each switching cycle and thus a calculation basis for the cost function is constantly created.

In a continued further embodiment of the method according to the invention, a total of all possible sub-switching states is restricted to a subset of battery modules that can be connected in parallel.

In a continued yet further embodiment of the method according to the invention, the memoization algorithm is used cascadingly on several levels, in that the values of the cost function for the respective sub-switching states are assembled from at least one list of values of the cost function for an individual interconnection of the respective battery module.

Furthermore, a system for reducing runtime-relevant complexity of a control algorithm of an AC battery is claimed, in which the system comprises an AC battery, the AC battery having a central controller with a scheduler and at least one string with a plurality of battery modules. A respective battery module of the plurality of battery modules has at least one energy store and a plurality of power semiconductor switches, which connect the respective battery module in series or in parallel or in bypass to another battery module. The central controller is configured to provide the scheduler with a phase voltage that results in a desired phase current at an output of the AC battery. In order to implement the phase voltage, the scheduler is configured to specify a respective overall switching state of the plurality of battery modules for each switching cycle, with the scheduler being divided into a real-time part and an offline part. The real-time part is configured to access a switching state table for each switching cycle, which contains all the overall switching states together with a value assigned by a cost function for a conversion of the respective overall switching state, and from this, for the at least one phase, the respective overall switching state with the smallest value to form the cost function. The offline part is configured to use a memoization algorithm to calculate the cost function for all overall switching states of the at least one string in a continuous sequence, with the memoization algorithm being configured to divide a respective string of the at least one string with a plurality of battery modules into smaller substring sections a respective sub-switching state, to calculate the cost function for all sub-switching states, to store the respective sub-switching states together with the respective value of the cost function in a directory, and the value of the cost function of the respective overall switching state from the values of the cost function for sub-switching states of the respective branch of the substrand sections forming at least one strand.

In one embodiment of the system according to the invention, the cost function is formed with a criterion from the following list: sum over respective battery module currents, which are weighted with a respective deviation of a respective battery module state of charge compared to an average battery state of charge, temperature of the respective battery module, ripple of respective battery module currents.

In a further refinement of the system according to the invention, a respective substring section is formed by battery modules which are respectively connected in parallel with one another. The same phase current flows through all sub-strand sections of the respective strand of the at least one strand.

In a further refinement of the system according to the invention, the respective distribution coefficients describe a distribution of the phase current in the respective substring section, which is distributed among battery modules that are respectively connected in parallel with one another.

In an even further developed embodiment of the system according to the invention, in the case of battery modules of the respective substring section which have complete charge equalization, the phase current is evenly divided between the respective battery modules which are connected in parallel with one another. As a result, the respective distribution coefficients of the respective substring section assume the same constant value, the value being inversely proportional to the number of battery modules of the respective substring section.

In a continued further embodiment of the system according to the invention, the scheduler is additionally configured to use the memoization algorithm in a cascading manner on several levels by assembling the values of the cost function for the respective sub-switching states from at least one directory of values of the cost function for an individual interconnection of the respective battery module.

Further advantages and refinements of the invention result from the description and the attached drawing.

• 1 zeigt schematisch ein erstes Verschaltungsbeispiel in einer Ausführungsform des erfindungsgemäßen Verfahrens.
• 2 zeigt schematisch ein zweites Verschaltungsbeispiel in einer weiteren Ausführungsform des erfindungsgemäßen Verfahrens.

It goes without saying that the features mentioned above and those still to be explained below can be used not only in the combination specified in each case, but also in other combinations or on their own, without departing from the scope of the present invention.

• 1 shows schematically a first interconnection example in an embodiment of the method according to the invention.
• 2 shows schematically a second interconnection example in a further embodiment of the method according to the invention.

In 1 a first interconnection example 100 is shown schematically in an embodiment of the method according to the invention. The interconnection example 100 includes six battery modules N=6 in a respective string of the at least one string of an AC battery. A first overall switching state 140 of the train under consideration is formed by a first sub-switching state 110 of a first sub-chain section, a second sub-switching state 120 of a second sub-chain section and a third sub-switching state 130 of a third sub-chain section. Since the three substring sections shown are connected in series, the same phase current 101, 102, 103, 104 flows through all three _substring sections a single battery module is involved, the phase current corresponds to 101, 102, 103, 104. Within the second substring section with the second subswitching state 120, three battery modules are connected in parallel. There should be charge equalization between the respective energy stores of the three battery modules. Following this, the phase current 101, 102, 103, 104 is divided one third into a second current i ₂ 121, a third current i ₃ 122 and a fourth current i ₄ 123 through the respective battery module in the second substring section. Finally, within the third substring section with the third subswitching state 130, two battery modules are connected in parallel. There should be charge equalization between the respective energy stores of the two battery modules. Following this, the phase current 101, 102, 103, 104 is divided equally between a fifth current i ₅ 131 and a sixth current i ₆ 132 through the respective battery module in the third substring section. The cost function S _VB1 for the first overall switching state 140 is thus: $$S_{VB1}=\left(i_1\cdot \Delta q_1\right)+\left(i_2\cdot \Delta q_2+i_3\cdot \Delta q_3+i_4\cdot \Delta q_4\right)+\left(i_5\cdot \Delta q_5+i_6\cdot \Delta q_6\right).$$

mit s_1-1=Δq₁, s_2-4=(Δq₂+Δq₃+Δq₄)/3 und s_5-6=(Δq₅+Δq₆)/2.According to the invention, the cost function in Eq. (7) can be represented as the sum of the respective cost functions of the substrand sections: $$S_{VB1}=s_{1-2}+s_{2-4}+s_{5-6}.$$

with s _1-1 =Δq ₁ , s _2-4 =(Δq ₂ +Δq ₃ +Δq ₄ )/3 and s _5-6 =(Δq ₅ +Δq ₆ )/2.

Erfindungsgemäß kann die Kostenfunktion in Gl. (9) als Summe von jeweiligen Kostenfunktionen der Substrangabschnitte dargestellt werden: $$S_{VB2}=s_{1-2}+s_{3-4}+s_{5-6}.$$

mit s_1-2=(Δq₁+Δq₂)/2, s_3-4=(Δ_q3+Δq₄)/2 und s_5-6=(Δ_q5+Δ_q6)/2.In 2 a second interconnection example 200 is shown schematically in a further embodiment of the method according to the invention. The interconnection example 200 includes six battery modules N=6 in a respective string of the at least one string of an AC battery. An illustrated second overall switching state 240 of the train under consideration is formed by a fourth sub-switching state 210 of a fourth sub-chain section, a fifth sub-switching state 220 of a fifth sub-chain section and a sixth sub-switching state 230 of a sixth sub-chain section. Since the three sub-string sections shown are connected in series, the same phase current 201, 202, 203, 204 flows through all three sub-string sections. Within the fourth sub-string section with the fourth sub-switching state 210, two battery modules are connected in parallel. Between the respective energy stores Both battery modules should be charge balanced. Following this, the phase current 201, 202, 203, 204 is divided equally between a first current i ₁ 211 and a second current i ₂ 212 through the respective battery module in the fourth substring section. Two further battery modules are connected in parallel within the fifth substring section with the fifth subswitching state 220 . There should be charge equalization between the respective energy stores of the two other battery modules. Following this, the phase current 201, 202, 203, 204 is divided equally between a third current i ₃ 221 and a fourth current i ₄ 222 through the respective battery module in the fifth substring section. Finally, two battery modules are connected in parallel within the sixth substring section with the sixth subswitching state 230 . There should be charge equalization between the respective energy stores of the two battery modules. Following this, the phase current 201, 202, 203, 204 is divided equally between a fifth current i ₅ 231 and a sixth current i ₆ 232 through the respective battery module in the sixth substring section. The cost function S _VB1 for the second overall switching state 240 is thus: $$S_{VB2}=\left(i_1\cdot \Delta q_1+i_2\cdot \Delta q_2\right)+\left(i_3\cdot \Delta q_3+i_4\cdot \Delta q_4\right)+\left(i_5\cdot \Delta q_5+i_6\cdot \Delta q_6\right).$$

According to the invention, the cost function in Eq. (9) can be represented as the sum of the respective cost functions of the substrand sections: $$S_{VB2}=s_{1-2}+s_{3-4}+s_{5-6}.$$

with s _1-2 =(Δq ₁ +Δq ₂ )/2, s _3-4 =(Δ _q3 +Δq ₄ )/2 and s _5-6 =(Δ _q5 +Δ _q6 )/2.

The cost function for the third substring section with the third subswitching state 130 in 1 and the cost function for the sixth substring section with the sixth subswitching state 230 are determined by the same parameters. After calculating the cost function s _5-6 and storing it in the directory, Eq. (7) and Eq. (9) can be used to save time with the same value without recalculating this contribution to the cost function of the respective overall switching state.

QUOTES INCLUDED IN DESCRIPTION

This list of documents cited by the applicant was generated automatically and is included solely for the better information of the reader. The list is not part of the German patent or utility model application. The DPMA assumes no liability for any errors or omissions.

Patent Literature Cited

• DE 102015112512 A1 [0003]
• DE 102018125728 [0007]

Claims (13)

Method for reducing runtime-relevant complexity of a control algorithm of an AC battery, in which the AC battery comprises a central controller with a scheduler and at least one string with a plurality of battery modules, in which a respective battery module of the plurality of battery modules has at least one energy store and a plurality of power semiconductor switches, which respective battery module connected in series or in parallel or in bypass to another battery module, in which the central controller specifies to the scheduler a phase voltage which leads to a desired phase current (104, 204) at an output of the AC battery, in which for realizing a respective overall switching state (140, 240) of the plurality of battery modules is specified for the phase voltage by the scheduler for each switching cycle, in which the scheduler is divided into a real-time part and an offline part which in the real-time part for each switching cycle accesses a switching state table, which contains all the overall switching states (140, 240) together with a value assigned by a cost function for a conversion of the respective overall switching state (140, 240), and from this for the at least one Line the respective overall switching status (140, 240) is formed with the smallest value of the cost function, with the cost function for all overall switching states (140, 240) of the at least one line being calculated in the offline part using a memoization algorithm in continuous sequence by • a respective string of the at least one string with a plurality of battery modules is considered to be divided into smaller sub-string sections with a respective sub-switching state (110, 120, 130, 210, 220, 230), • the cost function is calculated for all sub-switching states (110, 120, 130, 210, 220, 230), • the respective sub-switching states (110, 120, 130, 210, 220, 230) are stored in a directory together with the respective value of the cost function, and • the value of the cost function of the respective overall switching state (140, 240) is composed of the values of the cost function for sub-switching states (110, 120, 130, 210, 220, 230) of the sub-strand sections forming the respective strand of the at least one strand. procedure after claim 1 , in which the cost function is formed with a criterion from the following list: sum over respective battery module currents weighted by a respective deviation of a respective battery module state of charge compared to an average battery state of charge, temperature of the respective battery module, ripple of respective battery module currents. Method according to one of the preceding claims, in which a respective sub-string section is formed by battery modules connected in parallel with one another, the same phase current (101, 102, 103, 104, 201, 202, 203, 204) flows. procedure after claim 3 , in which a distribution of the phase current (101, 102, 103, 104, 201, 202, 203, 204) in the respective substring section, which is divided between battery modules connected in parallel with one another, is described by respective distribution coefficients. procedure after claim 4 , in which, in the case of battery modules of the respective substring section, which have complete charge equalization, the phase current (101, 102, 103, 104, 201, 202, 203, 204) is evenly divided between the battery modules that are connected in parallel with one another, as a result of which the respective distribution coefficients of the each substring section assume the same constant value and the value is inversely proportional to the number of battery modules of the respective substring section. Method according to one of the preceding claims, in which a total of all possible sub-switching states is restricted to a subset of battery modules which can be connected in parallel. Method according to one of the preceding claims, in which the memoization algorithm is used cascadingly on several levels, in that the values of the cost function for the respective sub-switching states are assembled from at least one list of values of the cost function for an individual connection of the respective battery module. A system for reducing runtime related complexity of an AC battery control algorithm, the system comprising an AC battery, the AC battery has a central controller with a scheduler and at least one string with a plurality of battery modules, in which a respective battery module of the plurality of battery modules has at least one energy store and a plurality of power semiconductor switches, which connect the respective battery module in series or in parallel or in bypass to another battery module, has, in which the central controller is configured to specify the scheduler a phase voltage, which leads to a desired phase current (104, 204) at an output of the AC battery, in which the scheduler is configured to implement the phase voltage for each switching cycle to specify the respective overall switching state (140, 240) of the plurality of battery modules, in which the scheduler is divided into a real-time part and an offline part, in which the real-time part is configured to access a switching state table for each switching cycle, which lists all overall switching t states (140, 240) together with a value assigned in each case by a cost function for a conversion of the respective overall switching state (140, 240), and from this for the at least one strand the respective overall switching state (140, 240) with the smallest value of the cost function to form, wherein the offline part is configured to calculate the cost function for all total switching states (140, 240) of the at least one line in a continuous sequence using a memoization algorithm, in which the memoization algorithm is configured to • a respective line of the at least one Strings with a plurality of battery modules to be considered divided into smaller substring sections with a respective subswitching state (110, 120, 130, 210, 220, 230), • the cost function for all subswitching states (110, 120, 130, 210, 220, 230) to calculate, • the respective sub-switching states (110, 120, 130, 210, 220, 230) together with the to store the respective value of the cost function in a directory, and • to compile the value of the cost function of the respective overall switching state (110, 120, 130, 210, 220, 230) from the values of the cost function for sub-switching states of the sub-strand sections forming the respective strand of the at least one strand . system after claim 8 , in which the cost function is formed with a criterion from the following list: sum over respective battery module currents weighted by a respective deviation of a respective battery module state of charge compared to an average battery state of charge, temperature of the respective battery module, ripple of respective battery module currents. system according to one of the Claims 8 or 9 , in which a respective sub-string section is formed by battery modules connected in parallel with one another, with the same phase current (101, 102, 103, 104, 201, 202, 203, 204) flowing through all sub-string sections of the respective string of the at least one string. system after claim 10 , in which the respective distribution coefficients describe a distribution of the phase current (101, 102, 103, 104, 201, 202, 203, 204) in the respective substring section, which is distributed among battery modules that are respectively connected in parallel. system after claim 11 , in which, in the case of battery modules of the respective substring section, which have complete charge equalization, the phase current (101, 102, 103, 104, 201, 202, 203, 204) is divided evenly among the battery modules that are connected in parallel with one another, as a result of which the respective distribution coefficients of the each substring section assume the same constant value and the value is inversely proportional to the number of battery modules of the respective substring section. system after one Claims 8 until 12 , in which the scheduler is additionally configured to use the memoization algorithm cascadingly on several levels, in that the values of the cost function for the respective sub-switching states are assembled from at least one directory of values of the cost function for a single interconnection of the respective battery module.

Images