CA2983789A1 - Multi-port converter structure for dc/dc power conversion - Google Patents

Abstract

A module for interconnecting a pair of DC sources or a pair of DC loads into a DC bus includes: a first port for each source or load; a switching cell for each first port, each cell having a pair of terminals and a switching node; a second port operatively connected to the DC bus and having a pair of terminals, one of the pair of terminals of the second port being connected to one of the terminals of one of the cells and the other of the pair of terminals of the second port being connected to one of the terminals of the other of the cells; and a filter inductor connected between the switching nodes of the cells. Systems including the module and methods utilizing the system are also disclosed.

Description

MULTI-PORT CONVERTER STRUCTURE FOR DC/DC POWER CONVERSION
Field of the Invention The invention relates to the field of power converters for dc systems.
Background A two-quadrant buck converter, also known as a synchronous buck converter, is a type of basic switch-mode dc/dc converter that is used to regulate voltage and provide efficient dc power transfer in energy systems. The traditional two-quadrant buck converter cell, shown in Figure 20 (prior art), comprises a pair of complimentary power switches and input capacitor. An output L-C low-pass filter is employed when a small high frequency ripple for the output voltage is required. For steady state operation, switch S1 is turned "on" (i.e. switch S1 is closed) and S2 is turned "off' (i.e. switch SI is opened) during time interval DTs. The converter duty cycle D, which ranges from 0 to 100%, represents the percentage time when switch S1 is on (and thus when S2 is off) during switching period T,. During each time interval D=Tõ voltage 1,, at node x becomes equal to input voltage V,õ as shown in Figure 20.
Voltage võ becomes zero when switch S1 is turned off (and thus when S2 is turned on) for the remainder of the switching period, due to the complimentary switching action of SI and S2. Based on this discussion, the voltage vx can be viewed as having an average value of D.V,õ
with a set of high frequency switching harmonics. The L-C low-pass filter is designed such that it attenuates the high frequency switching harmonics of vx and allows the output voltage Vow to be equal to the average value D.V,õ.
Assuming the output voltage is externally regulated, the inductor current //, can be made to take on either a positive or negative average value through adjustment of the converter duty cycle, thus enabling bidirectional energy transfer between input and output terminals. Therefore, voltage regulation and bidirectional energy transfer can be achieved easily by suitable control of the duty cycle D in a two-quadrant buck converter.
It should be understood that a unidirectional variant of the bidirectional buck converter in Figure 20 can be realized, by, for example, replacing switch S2 with a diode. The unidirectional buck converter can be employed for applications where only input to output power transfer capability is needed.

Multiple two-quadrant buck converter cells with associated filters can also be connected in series to form "classical cascaded buck converters". Figure 21 (i.e. prior art) shows a classical cascaded buck converter comprised of three individual dc/dc buck converter cells, each with associated output filtering, connected in series. The topology shown has three input ports and one output port. Each of the input ports and the output port consists of two terminals as shown. Observe each input port has an assigned reference terminal with its voltage defined relative to ground, i.e. vnr, vo and vo. By chosen convention the reference terminals are selected such that they correspond to the bottom connection point of each input port capacitor. To limit voltages to ground, a single reference terminal is typically connected to ground. In Figure 21, the reference terminal selected for ground connection is shown by the dotted connection from v03 to ground. However, it must be stressed this choice is entirely arbitrary, i.e. any other reference terminal in Figure 21 could have been connected to ground. The classical cascaded buck converter allows multiple input ports to exchange energy with a common output port, wherein the output voltage can be significantly higher than individual input voltages. This flexibility makes the classical cascaded buck converter suitable for a wide range of applications such as photovoltaic systems and battery management units.
Present state-of-the-art technology having similar application and functionality compared to the classical cascaded buck converter in Figure 21 is the cascaded connection of two-quadrant buck converter cells that share a single L-C low-pass filter, shown in Figure 22. However, with exception of the one ground-connected reference terminal, all other input reference terminal voltages for this topology, i.e.
vnj and võ2, are subject to undesired high frequency switching ripple voltage.
As a result, energy sources that are connected to these input ports, for example, solar panels or batteries, will suffer from significant capacitive current to ground. In contrast, the classical cascaded buck converter with multiple low-pass L-C filters as shown in Figure 21 can reduce the high frequency switching ripple voltage on vo and vn2, provided that individual L-C filter elements are sufficiently large. However, this comes at the expense of an overall increase in both the size and number of energy storage components (inductors and capacitors).
The additional components increase the loss and cost of the classical cascaded buck converter.
It is, therefore, desirable to develop a new non-isolated dc/dc converter topology that can be designed to have an arbitrarily small high frequency switching voltage ripple magnitude at all input and output reference terminals, while achieving reduced component count, cost, and loss.

2 Summary of the Invention It will be evident from the foregoing, and from a review of the detailed description that follows, that the multi-port converter topologies for dc/dc power conversion embodied within the apparatus are of significant advantage, in that, inter alia, they:
= are modular and scalable;
= can be designed to have an arbitrarily small high frequency switching voltage ripple magnitude at all input and output reference terminals;
= are capable of bidirectional energy exchange between input ports and output port;
= are capable of controlling power sharing among the input ports;
= are capable of allowing multiple inputs to exchange energy with a common output, wherein the output voltage can be significantly higher than individual input voltages;
= have relatively low rating of components; in particular, the net rating of energy storage components (capacitors and inductors) are small compared to the classical cascaded buck converter;
= are highly flexible in that they can be cascaded with modules of the same topology or cells of differing topology.
Other advantages and features associated with the multi-port converter topology will become evident upon a review of the following detailed description and the appended drawings, the latter being briefly described hereinafter.

3 Brief Description of the Drawings Figure 1 Double-input single-output converter module with generalized power switches;
Figure 2 One example of power switches realization using MOSFETs and diodes for the double-input single-output converter module in Figure 1;
Figure 3(a) One possible gating strategy with corresponding switching states for double-input single-output converter module in Figure 1;
Figure 3(b) Equivalent circuit diagram for Figure 1 corresponding to switching state #1: switch Sib and switch Sza turned on;
Figure 3(c) Equivalent circuit diagram for Figure 1 corresponding to switching state #2: switch Sia and switch S2a turned on;
Figure 3(d) Equivalent circuit diagram for Figure 1 corresponding to switching state #3: switch Sia and switch S2b turned on;
Figure 3(e) Equivalent circuit diagram for Figure 1 corresponding to switching state #4: switch Sib and switch S2b turned on;
Figure 4 Series-stacking 1-' double-input single-output converter modules in Figure 1 to form a 2k-input single-output cascaded dc/dc converter structure;
Figure 5 Series-stacking one double-input single-output converter module of Figure 1 with (k-1) two-quadrant buck converter cells to form a (k+1)-input single-output cascaded dc/dc converter structure;
Figure 6 (k+1)-input single-output cascaded converter structure where physical placements of the module and switching cell plurality are interchanged relative to Figure 5;

4 Figure 7 One possible cell sorting and selection scheme for cascaded converter structures in Figure and Figure 6 to achieve charge balancing of inputs across all possible output voltages;
Figure 8 Simulation model for Figure 1 implemented in PLECS;
Figure 9 Simulation results for double-input single-output converter module in Figure 1: power transfer from inputs to output;
Figure 10 Simulation results for double-input single-output converter module in Figure 1: power transfer from output to inputs, where power transfer is divided evenly between input ports;
Figure 11 Simulation results for double-input single-output converter module in Figure 1: power transfer from output to inputs, where power transfer is divided unevenly between input ports as assigned by the user;
Figure 12 Simulation results for prior art comprising four series-cascaded two-quadrant buck converter cells with single L-C output filter (i.e. a four input variant of Figure 22), to demonstrate inability to achieve full output voltage range when only one of the cells utilizes switch-mode operation;
Figure 13 Simulation results for a four-input single-output cascaded converter structure of Figure 5, to demonstrate ability to achieve full output voltage range when utilizing switch-mode operation for the high/low-frequency cell stack;
Figure 14 Switch gating waveforms for simulation results in Figure 13, where high-frequency cells are switched at 50 kHz;
Figure 15 Finer time scale resolution (i.e. zoomed time axis) for a chosen segment of simulated waveforms in Figure 14, to show contrast between high-frequency and low-frequency switching times;

5 Figure 16 Simulation results for a four-input single-output cascaded converter structure of Figure 5 employing battery modules, to illustrate mechanism allowing equal charge balancing between input ports such that the average battery voltages are depleted at the same rate;
Figure 17 Switch gating waveforms for simulation results in Figure 16, where high-frequency cells are switched at 50 kHz and low-frequency cells are switched at 20 Hz;
Figure 18 Finer time scale resolution (i.e. zoomed time axis) for a chosen segment of simulated waveforms in Figure 17, to show contrast between high-frequency and low-frequency switching times;
Figure 19 Double-input single-output converter module with functionally similar output capacitor configuration Figure 20 Prior art: Two-quadrant buck converter with output filtering, depicted along with corresponding operational waveforms;
Figure 21 Prior art: Classical cascaded buck converter with multiple L-C
output filters, where three buck converter cells are employed for ease of illustration; and Figure 22 Prior art : Cascaded buck converter with a single shared L-C
output filter, where three buck converter cells are employed for ease of illustration.
Detailed Description The dc/dc converter module shown in Figure 1 has two first ports, labelled with voltages "v1" and "v2", and one second port, labelled with voltage "v3". In this document, an "input"
port refers to a port that connects to a de source or load, while an "output" port refers to the port that is operationally connected to the de bus. Throughout this document, first ports and "input" ports are used interchangeably, and second ports and "output" ports are similarly interchangeable. Thus, the converter module in Figure 1 is referred to as a double-input single-output converter module.

6 The converter module in Figure 1 is comprised of two synchronous buck converter cells with a single filter inductor as shown. A key topological feature of the double-input single-output converter module is the placement of inductor L across the two inner switches Sib and S2a as shown. This configuration results in the inductor being effectively "isolated" from the output port terminals. That is, due to the imposed connection of L across non-matching switches of the two buck converter cells, the output port terminals may be connected directly to input port terminals as shown. This natural topological feature is seen as highly advantageous in achieving an arbitrarily small high frequency switching ripple magnitude at all input and output reference terminals in Figure 1, as it avoids reliance on excessively sized passive filters to achieve this goal.
The converter module in Figure 1 employs two pairs of complimentary switches:
1) Sia, Sib and 2) S2a, S2b. Similar to the convention illustrated in Figure 20, duty cycle command D1 controls the percentage time that Sia is on (and thus Sib is off) and duty cycle command D2 controls the percentage time that S2a is on (and thus S2b is off). An interleaved operation of the two pairs of switches is possible for this topology, where controlling a relative phase shift, (P, between D1 and D2 can regulate the order in which the four switches are turned on (over each switching period T5).
However, interleaved operation is optional, i.e. interleaved operation of the buck converter cells in Figure 1 is not required. The two pairs of complimentary power switches can be implemented using a number of different switching devices or technologies; one possible example of an implementation using MOSFETs and Diodes is shown in Figure 2. It should be understood that there are many possible implementations of the switches and energy storage components shown for the double-input single-output converter module in Figure 1.
Therefore, such variants are considered as being functionally similar to Figure 1.
Bidirectional energy exchange between the input ports and output port in Figure 1 is possible.
Specifically, power can be transferred either: 1) from the output port to both input ports or 2) from both input ports to the output port. A salient operational feature of the topology in Figure 1 is that power sharing among the two inputs can be achieved in a controlled manner, as will be demonstrated in the latter simulations section. It should be understood that a unidirectional variant of Figure 2 can be easily realized, by, for example, replacing two of the four MOSFET switches with diodes.

7 Double-Input Single-Output Converter Module: Theory of Operation Due to the flexible and scalable nature of the double-input single-output converter module shown in Figure 1, there are many possible methods or strategies in which to operate the converter. Therefore, the subsequent operational analysis should not be considered to be limiting. For demonstration purposes and ease of understanding, the following assumptions are imposed to illustrate the key characteristics of the topology:
= All switching devices and energy storage components (i.e. inductors and capacitors) are lossless;
= Dead time for switches is neglected to simplify mathematical analysis.
Reference is now made to Figure 1. Switches Si, and Sib are controlled by duty cycle command D1 while switches S2a and S2b are controlled by duty cycle command D2. It is assumed the two pairs of switches have the same switching period 7:,, however, this is done for ease of understanding as such an assumption is not necessary to obtain the following results. In general, four possible switching states exist for the converter module in Figure 1; these four states are illustrated in Figure 3(b), Figure 3(c), Figure 3(d) and Figure 3(e). One possible switching pattern that can generate all four switching states is given in Figure 3(a), where (I) is the (per-unitized) relative phase shift between D1 and D2 commands.
Note in Figure 3(a) the order of the switching states is indicated, however, as discussed previously, the order and duration of switching states can vary and thus the assumed diagram in Figure 3(a) is not unique. The duty cycle commands for each power switch are defined in Figure 3(a), where DI = 1 ¨ D1 and 131 = 1 ¨ D2.
The inductor voltage corresponding to the four possible switching states of Figure 1, as illustrated in Figure 3(a), are:
State #1: vz, = VI_ + V2 ¨ V3 (I) State #2: vi, = V2 ¨ V3 (2) State #3: 121, = ¨V3. (3) State #4: 19/, = V1 ¨ V3 (4)

8 The principle of inductor volt-second balance (IVSB) dictates that the average inductor voltage over one switching period in steady state is zero. Consequently, the following voltage relationship can be derived when equating the average inductor voltage to zero using IVSB:
V3 = (1¨ Di)Vi + D2V2. (5) The voltage relationship in (5) does not make any assumptions on the values of D1, D2, CP , VI., V2 or V3 within their permissible range and is independent of the switching period, and thus be considered a general design equation for this topology. However, a set of values for D1, D2, and cto can be selected to achieve a minimum inductor ripple current and a minimum capacitor ripple voltage.
Similar to the preceding analysis, a current relationship can also be found using the principle of capacitor charge balance (CCB). The principle of CCB dictates the average capacitor current over one switching period in steady state is zero. By equating the average capacitor current to zero for each port, the following current relationship can be derived ___________________________________________ ... _ (6) 1 ¨ Di D2 The polarity of currents in (6) correspond to those assumed in Figure 1.
Cascaded Topologies Formed by Series-Stacking Multiple Converter Modules The double-input single-output dc/dc converter module shown in Figure 1 can be extended to form cascaded topologies by stacking multiple modules in series. Figure 4 shows one possible example of such a cascaded topology where k double-input single-output converter modules are stacked to form a 2k-input single-output structure. Each of the k modules has the same topology as shown in Figure 1, but individual modules do not necessarily need to employ the same energy storage components, or same realization of the power switches. The cascaded converter structure shown in Figure 4 enables additional inputs (i.e. more than two) to exchange energy with a common output, wherein the output voltage can be significantly higher than individual input voltages.

9 The stack of k modules in Figure 4 can be gated (i.e. switched) on/off at a relatively high common switching frequency corresponding to fs. = 11T, , as is the case with conventional switch-mode converters. Output L-C filter components are designed to attenuate high frequency ripple associated with switch-mode operation. However, in contrast to high frequency (i.e. switch-mode) operation, substantially lower switching frequencies can also be exploited to operate the majority of modules in a "voltage stacking mode". The term "voltage stacking mode" refers to where select modules, or, select switching converter cells within individual modules, are inserted or removed from the cell stack for extended periods of time. This is inherently different from conventional switch-mode operation where cells are switched in/out at much higher switching frequencies, typically according to some form of pulse-width modulation. Taking into consideration both switch-mode operation and voltage stacking mode, values of fs for an individual cell can be anywhere from less than 1 Hz to several hundred kHz.
Higher switching frequencies (i.e. more than several hundred kHz) can also be adopted. There is no requirement the switching frequency of individual cells in Figure 4 must be equal; in general, each buck converter cell can employ a different L.
The converter structure in Figure 4 is only one example of how multiple double-input single-output converter modules of Figure 1 can be utilized to create a cascaded architecture with increased number of inputs. Reference is now made to Figure 5, where a single converter module from Figure 1 is series-stacked with (k-1) two-quadrant buck converter cells (i.e. switching cells) to form a (k+1)-input single-output cascaded structure. The resulting string of cascaded cells is partitioned into a "high/low-frequency cell stack" and a "low-frequency cell stack". The "high/low-frequency cell stack"
comprises the two switching cells in the converter module and signifies that these cells are capable of operating: 1) both as high-frequency switch-mode converters, 2) both in low-frequency voltage stacking mode, or 3) in any combination thereof.
The "low-frequency cell stack" in Figure 5 comprises the remaining switching cells (i.e. the cells that do not comprise the module) and signifies that the (k-1) series-cascaded cells operate exclusively in a low-frequency voltage stacking mode. That is, in general, the on-off states for each pair of complimentary switches in the "low-frequency cell stack" do not change during several successive high-frequency switching periods of the neighbouring "high/low-frequency cell stack".

Switching frequencies employed for the "low-frequency cell stack" can be many orders of magnitude smaller than the associated switch-mode operating frequency of the "high/low-frequency cell stack". It should be stressed that on-off state durations of complimentary switch pairs within the "low-frequency cell stack" can be made to be arbitrarily long, i.e. there is no fundamental constraint placed on the maximum 'on' state duration for each pair of complimentary switches within the "low-frequency cell stack".
Based on the above discussion, the input port reference terminal voltages vi,/, v,/,2 and vn2 to vn(k_o in Figure 5 are not subject to appreciable high-frequency switching ripple voltage. Voltages v,/,/, 17,1,2 in the high/low-frequency cell stack achieve this by exploiting the structure in Figure 1, while voltages vn2 to vn(k_i) achieve this by leveraging a very low frequency operation of the low-frequency cell stack. Note vnk is connected to ground in Figure 5 and thus does not experience any switching voltage stress.
An advantage of the topology in Figure 5 over the structures described in, for example, Z. Zheng, K
Wang, L Xu and Y Li, "A Hybrid Cascaded Multilevel Converter for Battery Energy Management Applied in Electric Vehicles," IEEE Trans. Power Electronics, vol.29, no.7, pp. 3537-3546, July 2014.
rZheng et all is that it can be designed to have arbitrarily small high-frequency switching ripple magnitude at all input and output reference terminals, while simultaneously achieving reduced component count, cost, and loss. This is evident when inspecting Figure 5 as only one interface inductor and one output filter capacitor is needed.
It should be understood the cascaded structure in Figure 5 is only one example of how the basic converter module in Figure 1 can be arranged with series-cascaded two-quadrant buck converter cells.
Any number of double-input single-output converter modules of Figure 1 can be utilized (i.e. not only limited to a single module as shown in Figure 5), as well as any number of two-quadrant series-cascaded buck converter cells. Other alternatives are possible, such as, for example, the order in which individual cells are stacked in series within the resulting string.
Reference is now made to Figure 6, where physical placement of the high/low-frequency cell stack and low-frequency cell stack are interchanged relative to Figure 5.

The topology in Figure 6 remains a (k+1)-input single-output cascaded converter structure. Apart from the physical ordering of low-frequency and high/low-frequency cell stacks within the string, this variant operates in a manner substantially similar to that of the structure of Figure 5.
Full Output Voltage Range Capability There is a practical restriction on the range of duty cycles a dc/dc converter can achieve. Namely, a converter is incapable of operating with a duty cycle very close to one, or very close to zero. Duty cycle commands within a small region above zero are in practice set to zero, while duty cycle commands within a small region below one are in practice set to unity. This practical limitation in the achievable range of duty cycles stems from the non-zero turn-on and turn-off times inherent to any semiconductor based power switching device. In addition, commercial PWM modules are typically incapable of modulating duty ratios very close to zero and one. It should be stressed the range of non-permissible duty cycles varies depending on many factors such as switch technology, switch voltage and current ratings, and switching frequency. In this work, particularly the simulations section, duty cycle commands less than 0.1 or greater than 0.9 are assumed to be unachievable for switch-mode operating cells. However, this range of values is chosen for illustrative purposes, only, and should be not considered as being typical.
The limitation in available range of duty cycles implies there is a range of average output voltages near zero, and a range of average output voltages near Vin, that a switch-mode dc/dc converter cannot achieve.
In Zheng et al., it is explicitly highlighted that the adopted operational strategy utilizes only one converter cell for high-frequency switch-mode operation. The remaining cascaded buck converter cells operate deliberately in a low-frequency voltage stacking mode. This implies only one cell is available to synthesize the deficit portion of the reference output voltage not provided by the dedicated low-frequency buck converter cells. Thus, taking into consideration the practical limitation for realizable duty cycles as described above, the topology in Zheng et al.is unable to achieve all possible average output voltages. This stems from the fact the single dedicated switch-mode buck converter cell cannot achieve all duty cycle commands. This deficiency of prior art will be demonstrated in the simulations section.

In contrast to prior techniques, the cascaded topologies in Figure 4 through Figure 6 operate in such a manner that achieves all possible average output voltages for the entire cell stack. The effect of the aforementioned duty cycle restriction is alleviated by allowing at least one double-input single-output converter module, which corresponds to at least two buck converter cells as illustrated by Figure 1, to operate in a switch-mode fashion as needed. The remaining cascaded cells need only operate in low-frequency voltage stacking mode. The key requirement is that, in order to accommodate all possible deficit portions of the reference output voltage not provided by the low-frequency cell stack, at least one module must be dedicated to switch-mode operation. Additional modules can operate as switch-mode converters, however, this is not required in order to achieve full output voltage range capability. This operational benefit will be demonstrated in the simulations section. It is important to reiterate the high/low-frequency cell stack is not limited to only one converter module, as shown by Figure 5 and Figure 6. Multiple modules can be cascaded while realizing the same benefit of full output voltage range capability.
Charge Balancing of Cells across all Possible Output Voltages The cascaded converter structures in Figure 4 through Figure 6 can be used for many applications. In particular these topologies are well suited for battery systems. Individual batteries can be connected to the input ports along the entire cell stack, thus enabling their integration with a common dc link for bidirectional energy transfer.
An advantageous operational feature of the proposed cascaded dc/dc topologies for battery systems is that charge balancing of the different cells, i.e. individual batteries, can be achieved. Moreover, recalling the preceding discussion on output voltage range capability, the cascaded structures in Figures 4-6 can achieve this cell balancing across all possible output voltages. Cell balancing is defined in this context as a means to ensure each battery along the entire cell stack is charged/discharged at the same average rate.
This is seen as a highly beneficial operating feature for battery energy management systems, as unequal charge depletion/repletion amongst the various batteries can be avoided. Cell balancing is achieved by suitable operation of the individual cells (ref. Figure 4) or suitable coordinated operation of low-frequency and high/low-frequency cell stacks (ref. Figures 5 and 6). Of course, long term (i.e.

average) charge/discharge rates of individual batteries can deliberately be made unequal, if so desired.
Reference is now made to Figure 7, which shows a high level diagram illustrating one possible operational strategy to achieve charge balancing between cells for the cascaded converter structures in Figure 5 and Figure 6, across all possible output voltages. Here it is assumed there are a total of N cells in the entire stack, with N-2 cells in the low-frequency cell stack (i.e. two cells in the module and N-2 cells in the switching cell plurality). As shown in Figure 7, there exists a cell voltage sorting block and a gating logic block. The cell voltage sorting block acts on a "slow" time scale, corresponding to the switching period of the low-frequency cell stack, and sorts/orders all input ports based on their voltage measurements. The gating logic block acts on a substantially faster time scale, corresponding to the high-frequency switching period of the high/low-frequency cell stack, to allow at least one cell within the high/low-frequency cell stack to operate in switch-mode. As a result of this combination of slow and fast control, individual cells can be removed and inserted as needed to meet all possible stack voltage demands (i.e. output port voltage references), while simultaneously ensuring charge balancing between cells. The "output port voltage reference" command in Figure 7 can be generated by various means, such as, for example, by the resulting control action of an external control logic block. An example application employing batteries is presented in the simulations section to illustrate this functionality.
It should be stressed that, as previously mentioned, the implementation in Figure 7 is not unique. There are many other alternate implementations that can similarly achieve balancing between cells of the entire stack, by suitable operation of the high/low-frequency and low-frequency cell stacks.
Deployment The double-input single-output dc/dc converter module shown in Figure 1, along with the embodiments shown in Figure 4 through Figure 6, can be used to enable bidirectional (or, if desired, unidirectional) energy exchange between multiple inputs and a common output, wherein the output voltage can be significantly higher than individual input voltages. Such operation has wide range of application in systems such as, but not limited to, photovoltaic systems and battery management units. For example, individual photovoltaic panels that use centralized or distributed based maximum power point tracking schemes can be connected to input ports of a unidirectional variant of the cascaded topology shown in Figure 4, thereby allowing maximal energy extraction from each panel (at its respective panel voltage) to a local load or external dc network.
Another example of application is to connect battery units to the input ports of the converter topology in Figure 5 and Figure 6 to enable centralized energy management of the battery system. The bidirectional energy exchange capability can be leveraged to allow individual battery units:
1) to supply energy to a load/source connected at the output or 2) to be charged by a dc source connected at the output.
Furthermore, individual battery units can be inserted or removed from the battery stack, via appropriate switching action, thus allowing balanced charging and/or discharging of select battery units for all possible output voltages.
Converter Simulations: Introduction In subsequent sections, the double-input single-output dc/dc converter module in Figure 1 and the (k+1)-input single-output cascaded dc/dc converter structure in Figure 5 are simulated using PLECS.
Specific case study scenarios are simulated to demonstrate key operational characteristics of the converters. In addition, an example case study comparing performance of the converter module in Figure 1 and prior art is carried out.
Simulation Results: Example Case Study Performance Comparison between Proposed Double-Input Single-Output Converter Module and Classical Two-Input Single-Output Cascaded Buck Converter The PLECS simulation model of the double-input single-output converter module of Figure 1 is given in Figure 8.

The nomenclature used in reference to Figure 8 is summarized in Table 1.
Table 1: Nomenclature adopted for PLECS simulations involving Figure 8 Quantity Name Vin Average voltage of input ports (i.e. average components of and 172) Vout Average voltage of output port (i.e. average component of V3) /out Average current of output port (i.e. average component of 13) Switching period D,D1,D2 Duty cycles (/) Phase shift (per-unit) between duty cycles LI Vout pp Peak-to-peak output voltage ripple of V3 louttu Peak-to-peak output current ripple of 13 dvc ,pp Equivalent single capacitor peak-to-peak voltage ripple LI iL,pp Equivalent single inductor current //, peak-to-peak ripple iiv Voltage ripple of reference terminal voltage vn1 Advantages of the double-input single-output converter module in Figure 1 over the classical two-input single-output cascaded buck converter (i.e. two cascaded buck converter cells with associated filters) can be quantified by comparing them in an application example. Such an application example is now considered, which consists of an energy management system having two equal battery voltages of Vin=60V (i.e. two inputs of 60V nominal rating) with an output voltage requirement of 170,,t =90V and iou t =10A. This corresponds to a power transfer of approx. 900 W. Equal energy storage requirements (of inductors and capacitors) and operating conditions are imposed on the converter module in Figure 1 and the classical cascaded buck converter. These imposed conditions allow key performance criteria of the topologies to be compared directly. The performance criteria of interest for this example case study are: 1) losses, 2) capacitive current to ground, and 3) harmonic distortion.
The following metric is proposed to evaluate converter performance Metric diL,pp = Ay?, (7) In order to reduce converter losses, capacitive current to ground and total harmonic distortion, it is necessary to reduce both AiL,pp and A vr, simultaneously. In other words, a "better" converter will have smaller A iL,pp and A vn values, and thus a smaller metric score constitutes a preferred converter structure based on the considered performance criteria.
Figure 8 shows the PLECS simulation circuit for the double-input single-output converter module in Figure 1. In all simulations, the switching period 7:, is set equal to 16.67us for all switches. The inductor L is 125 H, and capacitor C3 is implemented by connecting two 0.926 F
capacitors in series as shown. This is done to ensure fairness of the comparison, as the simulation model for prior art utilizes two capacitors of 0.926 F each and two inductors of 62.5p,H each. The sizing of energy storage components is chosen based on inductor current ripple and capacitor voltage ripple considerations for the prior art. Based on these parameters, the converter module in Figure 1 (and also Figure 8) and the classical cascaded buck converter have equal inductive and capacitive energy storage requirements.
Moreover, this also ensures equal operating conditions for both topologies.
A comparative case study is carried out for the two topologies as described above. Reference is now made to Figure 9, which presents simulation results for the double-input single-output converter module in Figure 1, where power transfer is approx. 900 W from input ports to output port. The depicted waveforms were obtained using the simulation model in Figure 8.
A summary of the case study simulation results comparing the two converter topologies is tabulated in Table 2. The classical series-cascading of buck converter cells as employed by prior art results in equal duty cycle commands: DI=D2=0.75. However, to achieve the same operating point, the double-input single-output converter module in Figure 1 employs unequal duty cycles:
D1=0.25 and D2=0.75. This is due to the imposed switches arrangement and chosen convention of duty cycle commands, as illustrated in Figure 3(a).
It should be noted that a set of optimal 0, DI and D2 can be found to minimize il iL,pp and .419, for Figure 1 and prior art separately.

Table 2: Summary of case study simulation results comparing performance of double-input single-output converter module in Figure 1 (and also Figure 8) and classical cascaded buck converter (prior art) V0(V) /out(A) A vout,pp(V) Ajout,pp(A) Avc,pp(V) A ii,,pp(A) lv(V) Figure 1 0 90.2 9.96 8.58 0.954 4.29 2.12 8.53 D1=0.25 0.25T 90.0 9.99 2.23 0.247 1.11 1.02 2.23 D2-0.75 0.50T 90.5 9.97 8.56 0.951 4.28 2.15 8.56 sw Prior art 0 90.1 9.99 12.6 1.41 6.32 3.17 6.31 _ ____________________________________________________________________________ D1=0.75 0.25T5õ 90.2 9.96 8.58 - 0.953 35.4 5.68 35.3 D2-0.75 0.50Tsw 90.0 10.01 2.22 0.246 26.6 2.92 26.4 Table 3 compares the computed values of performance metric (7) for Figure 1 and prior art, based on the case study results summarized in Table 2. It can be seen from Table 3 that the double-input single-output converter module in Figure 1 achieves a better metric score in comparison to the classical cascaded buck converter, for all values of 0. This implies Figure 1 is the preferred topology for the considered performance criteria.
It is important to recognize the optimal value of 0 for each converter topology, which is defined in this example as the 0 value corresponding to the lowest metric score in Table 3, is 2.27 for Figure 1 and 20.03 for prior art. For these optimal conditions, the double-input single-output dc/dc converter module has a far superior performance (as metric score is approx. 10 times lower) and therefore outperforms the classical cascaded buck converter.
Table 3: Summary of key simulation results for converter module and classical cascaded buck converter (prior art) 0 Figure 1 Prior art 0 18.2 20.03 Metric 0.257 2.27 201.07 0.50T 18.4 77.67 It should be stressed that a case study comparison between the double-input single-output converter module in Figure 1 and cascaded buck converter with single L-C output filter (i.e. Zheng et al.) is not carried out, due to the very large high frequency switching voltage ripple that naturally occurs with the latter. In this case, use of Zheng et al. would result in a Avn of 60 V, which implies computed values of metric (7) that far exceed those summarized in Table 3.
Simulation Results: Power transfer from output port to input ports of single converter module The previous simulations imposed dc power transfer from inputs to output.
Additional simulations are now performed to demonstrate the dc/dc converter topology in Figure 1 is capable of bidirectional energy exchange, i.e. power transfer from inputs to output and vice versa.
Moreover, power sharing between input ports is also demonstrated via simulation.
Two additional simulations are performed for Figure 1 to illustrate dc power transfer from output port to input ports. Specifically, these two simulated scenarios show that an energy source located at the output transfers 900 W to the inputs, and power sharing among the input ports can be controlled. Power sharing in this context implies the total power transfer can be arbitrarily split between input ports.
Figure 10 presents the simulated waveforms for power transfer from output to inputs where the power is shared equally among the two input ports. Figure 11 presents the simulated waveforms for power transfer from output to inputs where the power is shared unequally among the two input ports, as assigned by the user.
Simulation Results: Realize full output voltage range for cascaded converter structure of Figure 5 Simulations are provided to demonstrate the achievable output voltage range for Zheng et at. in relation to the proposed topology in Figure 5. The simulation models are implemented in PLECS and correspond to: 1) four series-cascaded two-quadrant dc/dc buck converter cells with single L-C output filter (Zheng et al.), and 2) a four-input single-output cascaded dc/dc converter structure of Figure 5 (i.e. k=3). Each converter structure has four batteries connected at the input terminals, where each battery has a nominal potential of 10 V. Thus, the possible range of output voltages is 0 to 40 V.
It should be reiterated the former has one buck converter cell dedicated to switch-mode operation while the latter employs one converter module (comprising two buck converter cells) for possible switch-mode operation. The remaining converter cells in each topology operate exclusively in low-frequency voltage stacking mode.
As stated previously, it is assumed that duty cycle commands less than 0.1 or greater than 0.9 are unachievable.
Figure 12 shows both the filtered and unfiltered average output voltage (i.e.
"stack voltage") corresponding to prior art, for all possible output voltage references. Here one converter cell operates as a switch mode converter at any given time, while the remaining three cells are inserted as necessary to build up to the desired stack voltage. Observe the three regions in the lower plot of Figure 12 that exhibit a "flat" voltage profile. This output response is a direct result of the duty cycle limitation of the switch-mode operating cell as described previously. This simulation result clearly shows that prior art is incapable of achieving a continuous output voltage profile. The converter cannot provide certain output voltages as shown, and therefore its operating range must be restricted to avoid these undesirable operating regions. Alternatively a rapid sorting of the cell voltages could address the problem, however, the cells pre-designated for low-frequency operation could no longer be considered as operating exclusively as such.
Figure 13 shows both the filtered and unfiltered average output voltage (i.e.
"stack voltage") corresponding to the four-input single-output structure of Figure 5, for all possible output voltage references. Individual cells within the high/low-frequency cell stack operate as switch-mode converters when needed, and cells within the low-frequency cell stack operate exclusively in voltage stacking mode. The lower plot in Figure 13 clearly demonstrates the topology of Figure 5 is capable of achieving a continuous output voltage profile. Thus, in comparison to prior art, the proposed topology can achieve better utilization of available cell voltage in achieving the full output voltage range.
Figure 14 shows gating signals for the individual converter cells of Figure 13. Figure 15 shows a finer resolution time scale for a chosen segment of simulated waveforms in Figure 14, to show contrast between high-frequency and low-frequency switching times.

Simulation Results: Charge balancing of cells for cascaded converter structure of Figure 5 Simulation results are provided to demonstrate the capability for cell balancing in Figure 5 based on the operation strategy depicted in Figure 7. These PLECS simulations utilize the same case study system as utilized in the previous simulation section. That is, a four-input single-output realization of Figure 5 is modeled in PLECS with four integrated batteries, where each battery has a nominal potential of 10 V.
Figure 16 plots the four battery voltages as a fixed amount of de power is transferred to the converter output terminals. Observe the state-of-charge of all four batteries is depleted at the same average (i.e.
long-term) rate. This charge balancing between cells is achieved by utilizing the operational strategy conceptualized in Figure 7. Although implemented as a case study example for Figure 5, this same functionality can be implemented for Figure 4 and Figure 6 (or any variants thereof). Recall cell balancing can be achieved across all possible output voltages, which is not possible using prior art.
Figure 17 shows gating signals for the individual converter cells of Figure 16. Figure 18 shows a finer resolution time scale for a chosen segment of simulated waveforms in Figure 17, to show contrast between high-frequency and low-frequency switching times.
Variations Whereas specific embodiments of the invention have been discussed, variations are possible. For example, Figure 2 shows one possible implementation of the two pairs of complimentary switches using MOSFETs and Diodes, however, the switches can be implemented using a number of different switching devices and technologies, and, similarly, energy storage components (i.e. inductors and capacitors) can also be implemented with equivalents. For example, the placement of capacitors C3a and C3b shown in Figure 19 can be employed as an alternate output capacitor configuration. All of these variations should be considered as functionally similar. It is also possible to realize unidirectional variants of the presented bidirectional topologies, by suitable implementation of the switching devices.

Furthermore, Figure 4 through Figure 6 are examples of cascaded topologies derived based on the converter module of Figure 1 and series-cascaded buck converter cells. As there are many other possible functionally similar realizations of cascaded converter structures using the basic building block in Figure 1, the illustrated embodiments should not be considered as limiting.
Furthermore, the two-quadrant buck converter cell is explicitly employed in all topologies, however, this is not essential. Other types of dc/dc converter cells may be employed, for example, buck or boost converter cells.
Whereas specific operating conditions and parameters are disclosed as part of the simulations and others, persons of ordinary skill will understand that these are included for illustration, only, and are not intended to be limiting.

Claims (4)

Claims

1. A module for interconnecting a pair of DC sources or a pair of DC loads into a DC bus, the module comprising:
a first port for each source or load;
a switching cell for each first port, each cell having a pair of terminals and a switching node;
a second port operatively connected to the DC bus and having a pair of terminals, one of the pair of terminals of the second port being connected to one of the terminals of one of the cells and the other of the pair of terminals of the second port being connected to one of the terminals of the other of the cells; and a filter inductor connected between the switching nodes of the cells.

2. A system for interconnecting a plurality of DC sources or a plurality of DC
loads into a DC bus, the system comprising:
a plurality of the modules of claim 1, provided in a number such that the number of cells equals the number of sources or loads, wherein the second ports are connected in series to one another and thereby operatively connected to the DC bus.

3. A system for interconnecting a plurality of batteries to a DC bus, the system comprising:
at least one module according to claim 1; and a plurality of switching cells, each cell having a pair of terminals and a switching node, provided such that the number of batteries equals the total of the switching cells of the module and of the plurality;
the outputs of the switching cells of the module and of the plurality being connected in series and thereby operatively connected to the DC bus.

4. A method for operating the system of claim 3 to provide a desired voltage at the DC bus, the method comprising:
cycling at least one switching cell of the module at a relatively high frequency to produce a varying first voltage; and cycling the switching cells of the plurality at a relatively low frequency to produce a varying second voltage, characterized in that:
at all times, the sum of the first voltage and the second voltage is the desired voltage; and the first voltage and the second voltage vary to balance the state of charge of the batteries.