Modular Control Valves. Typically used with Proportional valves with Zero-Lap type spool configurations that could allow oil movement when the system is shut down. This valve will block flow when de-energized and will allow bi-directional flow when energized. Both the A and B work ports incorporate a 2-way, 2-position bi-directional, normally closed poppet type solenoid valve. Energizing the solenoid will allow full flow to the system. Had these forces been modules combined at a notional column for a 4.
For light steel walls, an analogy with masonry is used in which an approximate length of wall of 2. The robust- ness of the test house described earlier was demonstrated during its controlled demolition, in which walls were progres- sively removed, as illustrated in Fig Guidance on structural integrity is given in SCI P Full-scale test on a two-storey house Fig Load test on A full-scale test house, illustrated in Fig 19, was built inside a a pair of modules factory in Newport, UK The magnitude of the unfactored design wind load corner is 0.
The design imposed load on the floors is 1. The rooms were sealed using a plastic membrane and progressively filled with water to a measured depth and the total load applied to the floors and roof was in excess of kN to achieve an imposed load factor of 1.
Lateral loads were applied by jacking between steel restraint frames and spreader beams located at first floor and eaves level adjacent to the stairwell gable, and the rear face of the house. A series of tests was conducted using combinations of vertical and horizontal loading. The light steel wall frames consist of 75mm deep C sections of 1.
The series of tests was repeated on perforated walls the brick-clad structure. Table 4 presents the measured deflec- tions for shear loads resisted by the perforated front and rear Measuring position Bare frame FE result Brick-clad building FE result walls, which represents the more flexible case.
Table 5 presents 1. In Mid-side 7. Lateral deflections were predicted by a Finite Eaves Corner The brickwork cladding was Corner These 1st floor Corner 4. BS cross-refers to these requirements, but does Corner 6. He has authored over 20 SCI publications in light steel framing, composite construction and fire engineering.
He was iour under unusual actions. Technology in the School of the Built Environment. The position is partly sponsored The module dimensions were 3.
The thick- by SCI. He is a former employee of Arup Associates and has a PhD in 1. Uniform load was applied using water containers. Production Engineering. He has authored a wide range of technical publications on construction, building physics and building technology Vibration tests gave a natural frequency of over 12Hz for an imposed floor load of 0. He is a charted engineer and was awarded his doctorate for research 5.
The serviceability load tests on the floor gave a deflection into post-tensioned masonry. He joined the University after a period in the of 8. He has published a numerous of papers on various aspects modules; firstly, when support to one longitudinal side removed, of structural engineering research.
He is a member of various European and national technical committees. John Grubb is a structural engineer and consultant to SCI.
He has co-authored removed. The first test showed that the modules were able to several SCI publications in the area of light steel framing. He obtained his degree at span as a deep beam with one longitudinal support removed Nottingham University. He was formerly Technical Director for Conder Structures and the maximum deflection was 23mm. The deflections for the in Winchester. His professional career started with Ove Arup and partners in Removal of one corner support led to a where he took his PhD.
He also holds a degree in Civil Engineering with deflection of 19mm, which demonstrates the torsional action of Architecture from Leeds University. He is a London Technology Network Business Fellow fostering both horizontally and vertically at their corners, although the academic-industry links. Removal of support to an edge Robustness of light steel framing may be achieved by design- or corner module causes the modules above to act as cantilevers.
The self weight of the modules is typically 5t when fitted out. For an acci- Acknowledgments dental load of 2. Photographic material and addi- This paper reviews modern methods of light steel construction tional information was provided by Feilden Clegg Bradley, that are used in the residential sector, and identifies mixed Michael Barclay Partnership, The Forge Company, Unite forms of skeletal, planar and volumetric construction that are Modular Solutions, HTA Architects, The Design Buro, economic in the medium-rise sector.
The structural behaviour Kingspan Off-site, Corus Living Solutions, Ayrshire Metal, of light frames demonstrates considerable reserve in stability Terrapin, Yorkon, the Swedish Institute of Steel Construction, and structural integrity.
Background tests have shown that a and Bouwen met Staal. Re-thinking Construction. The Report of the Gann, D. The Policy Press, Lessing, J. Barlow, J. Japanese lessons on with steel frames. A study of the OpenHouse Minister, customer-focussed house building, system.
The Swedish Institute of Steel 2. Non-traditional housing: A collection of BRE Venables, T. Construction Report , publications. The Building Research et al. BS Structural use of steelwork in building Establishment, offsite manufacture.
The Housing Forum, Part 5: Code of practice for design of cold formed 3. Delivering stability: Securing future housing From all the above, it is extracted that the agreement between the differential equation solutions of the continuous model and the simulations using the circuit model which includes power switches is rather satisfactory.
The three-phase direct modulation control scheme has been also simulated. The results are shown in the following figures. The output line-line voltage is a 2N-1 -level voltage as expected, shown in Figure 2. To achieve this both the modulation indices and the triangular references were phase shifted by o between the two arms of the same phase leg.
To get the different switching combinations in the converter, the modulation reference signals are phase- shifted between the two arms of the same phase leg, but the triangular carrier signals are kept the same.
The two modulation approaches are summarized in Figure 2. Of course some of the module switching combinations might never be used. The second case brings along some problems though, due to the voltage changes across the arm impedances in order to keep the DC-link voltage constant.
The output phase voltage, which is shown in Figure 2. The first thought is to increase the arm inductances, so that they operate as filters for the arm currents. Simulations with larger values of arm inductors have shown that the second harmonic of the arm currents is suppressed satisfactorily.
This however presents two main problems: a it increases the cost of the system unnecessarily and b it does not ensure stable performance under transient conditions, including changes in the load conditions or the request for increase or decrease in the voltage in which the capacitors are operated.
Thus, the need for a more efficient control structure is derived. The goal of the control scheme is to bring the circulating current to be a plain DC current. In this way, each arm current becomes a DC plus a fundamental frequency component. This current, which passes through the modules in the detailed model or through the variable capacitor in the continuous model, creates a second harmonic component in the AC output voltage.
In a control system that includes a reference signal for the AC voltage, this component may be extinguished through its proper modification. The next chapter describes such control systems for a Modular Multilevel Converter. A continuous model was also developed assuming infinite number of modules and infinite switching frequency, and its comparison with the more realistic detailed model showed that it can be an accurate way to study the stabilization of the converter.
For the convenience of the reader some of the equations are repeated. The study is made on the equivalent circuit of a single phase M2C, as depicted in Figure 3. This approach is helpful in solving problems related to circuit stability and harmonic elimination. The main goal is to design a control system that would ensure the two mentioned conditions [6]. Finally, equation 3. The AC voltage equation contains some voltage drop components caused by the inner inductance and resistance in the arms.
In other words the arm inductance L and the line resistance R form a fix, passive inner impedance for the AC current. At this point, a first approximate step for the derivation of the arm voltages will be made. Let the inner EMF be defined as in 3.
This means that the difference current idiff becomes a DC current. Accordingly, it will not cause an inductive voltage drop. Further the arm resistance should be low and therefore it could be neglected, as an initial approach shown in 3. Therefore, by subtracting the same voltage contributions udiff from both arms of the same phase leg of the converter, the idiff can be controlled independently of the AC- side quantities. This very last statement makes it more reasonable to finally define the arm voltages as follows in 3.
For this reason the following quantities are defined in 3. As far as the total energy deviation is concerned, it can be clearly observed from 3. According to 3. Therefore, this DC component of the circulating current can be used to control the total energy stored in all capacitors of one converter leg. On the other hand, and according to 3. The fundamental frequency component of idiff though, which has the same frequency as the output voltage does affect the even distribution of the capacitor voltages between the two arms of the same converter leg.
A similar effect is created in the second term of equation 3. As mentioned in the beginning of the second chapter the arm currents should be able to carry a DC component such that the product of the circulating current and the voltage of the DC-link provides the power that is actually transferred to the AC side.
These currents should also contain a fundamental frequency component, the sum of which should equal the AC output current.
It was shown earlier though, that the arm currents in the direct modulation scheme contain a strong second harmonic component. If it is assumed that the total capacitor voltages are available at any time, a proper voltage feedback control and modulation concept would be able to eliminate the harmonics in the arm currents and stabilize the converter through the energy stored in each leg. This requires of course the measurement of all submodule capacitor voltages and the calculation of voltage references for each arm.
This means that the term udiff can be neglected considering that the arm impedances R and L are small enough. However, this approximation is very likely to lead to instabilities. Thus, the need of controllers design is derived. These controllers will contribute to the term udiff of the reference equations of each arm, such that the ideal conditions, described by 3. Then the total energy deviation of the leg is determined by 3. If the converter is treated as a three-phase unit though, such as in a motor drive application, then this reference is given by an external AC controller.
In steady-state, the average of the last term is given by 3. Thus, 3. In a first approach, a PI controller is used. The integral part is necessary to eliminate the static errors in the average energy level for various active output power from the AC circuit.
Without the integral part, the capacitor average voltage drops when active power is transferred from the DC to the AC side and increases when power is delivered from the AC to the DC side. A derivative part might also be needed to avoid the undesirable oscillation caused by the phase-lag that is being introduced by the PI controller. The controller described above determines finally a DC component, which is added to udiff. Assuming very small value of udiff, compared to the term uD 2 , finally 3.
The only quantity that can be freely varied without causing deviations from the eV reference is idiff. In order to balance the energies between the two arms in the long run, the capability of creating a continuous derivative of WCdiff is necessary. Therefore, the only way to influence the energy balance between the two arms of the same leg is by creating a circulating current idiff, which produces an average value when multiplied with the inner EMF eV in the converter.
This can only be achieved if the circulating current idiff is given an AC component with the same frequency and in phase with eV. Let the inner EMF reference in the converter be given as a phasor in a rotating coordinate system, as in 3. Inserting 3. This means that the measured energy difference WCdiff is rather oscillatory. However, only the average of WCdiff is of interest in the feedback control scheme.
Accordingly a filter must be provided in order to extract the average components WCdiff and to ensure that the interaction of the balance control system with the total energy loop is avoided. This filter also attenuates the inevitable high frequency noise that occurs in the system measurements as well. So far, this balance control structure is expressed in terms of an AC contribution to the circulating current idiff. The difference current however can not be directly controlled, but it must be manipulated through the applied difference voltage udiff.
Assuming that the correct phase is used, the differential equation for the dynamics of the unbalanced energy is given by 3. In a normal case, the use of only a proportional gain should be adequate for the controller, since for a typical application the reference for the unbalanced energy is set to zero.
However, it might be necessary to add an integral part for possible static errors during transient operations as well. However, for the simulation and implementation on the real system, an adaptation for the time domain is needed. Assuming that the total capacitor voltage for each arm is measured and thus available at each time instant, the M2C stabilization in runtime is achieved as shown in 3.
The global closed-loop control structure for the M2C is illustrated in Figure 3. Both the continuous and the detailed model were investigated. Table 3. As far as the energy references of the controllers are concerned, they were chosen to fulfill the typical steady state operation conditions of an M2C.
In the detailed model, this voltage is evenly distributed among the submodules, meaning V per capacitor. Finally, WCdiff ,ref is set to zero. According to the classic modulation concept, the output phase voltage in the detailed model consists of 6 levels. In Figure 3. The strong second order harmonic, which was observed in the direct modulation control structure, is significantly suppressed.
This proves the effectiveness of the controller in steady state operation. The arm currents contain a DC and half of the AC fundamental frequency components. The circulating current is very close to being a plain DC current, corresponding to the active load that is delivered to the AC side. The control structure also contributes to the second harmonic voltage component in the arm capacitor voltages. However, a second harmonic voltage still exists as a result of the multiplication of the insertion index and the fundamental frequency component in the arm current.
This is not always a disadvantage though, since for load currents at certain power factors, it may expand the available voltage range, which otherwise may be limited by the arm capacitor voltage ripple.
Compared to the direct modulation, the voltage variation does not exceed the value of 35 V peak-peak anymore. The output phase voltage third order harmonic is also suppressed. Output current THD factor reaches 0. Finally, Figure 3. Figure 3. The continuous model shows an idealized output voltage condition, since an infinite number of modules and infinite switching frequency are assumed, leading to almost pure sinusoidal waveform.
In order to get the full 11 levels per phase, the converter was operated with a full inner EMF reference eVpk of V, which leads to higher converter currents as well. The results are given for steady state operation in Figures 3. The voltage across Zarm appears to have a larger frequency and sometimes larger steps as well, which have no impact on the circulating current peak values.
The output current appears to be smoother and the voltage THD is also reduced due to the extra levels that are gained through this modulation concept. It is interesting to depict the transient response of the physically measured signals, which are the capacitor voltages. Figures 3. It is observed that the total energy controller ensures that the whole phase leg stored capacitor voltage reaches the desired reference value quite fast.
In addition, the balancing mechanism keeps both arm capacitor voltages of the phase leg within the same limits, so that no arm capacitor voltage gets too high. The even distribution between the individual submodule capacitors though, is still a matter of the balancing algorithm of the modulator. The agreement between the two models, as illustrated in Figure 3. As a remark, the comparison shows that during steady state operation, more oscillation is observed in the detailed model than in the continuous one.
During the transient though, the continuous model seems to have some more overshoot and damping than the detailed. An adjustment of the controller gains can change the system behavior accordingly.
In addition, a derivative term can be added in the total energy controller or some lead-lag filtering. This, however, slows down the transient response. This sharp increase of the idiff is expected, since the controllers demand from the capacitors to be charged in a very short time.
These quantities are to remain the same before and after the transient. In the detailed model there is a slight increase of the peak-peak output voltage though. This is caused by the higher charging of the individual submodule capacitors. However, the fundamental frequency output voltage amplitude is kept unchanged. As mentioned before, the reference for this controller is usually set to zero. For demonstration purposes though, it is made non zero. This asymmetrical change is executed by a temporary fundamental frequency component, which is shown clearly in Figure 3.
At second 2 of the simulation time, the amplitude of the voltage reference eVpk is changed from V to V. The output AC voltage and current respond very fast, as depicted in Figures 3. Since the eVpk reference value becomes quite small, this affects the peak-peak value of the modulation index. The smaller modulation index is finally translated by the modulator into the use of less output voltage levels. This is shown in Figure 3. On the other hand, the control mechanism that affects the total capacitor voltages, takes about 0.
It is observed that initially, the total voltage increases. This is explained by Equation 3. The principle is that the calculation of the modulation indices nU and nL is performed in runtime. This is based on the steady state analytical solution of the AC quantities equations, which satisfy the conditions given by 3. As a first step for the control system design, it is assumed that the circulating current idiff ideally consists of only one DC component, as shown in 3.
Then the following condition stated in 3. The integration constant is freely selected to be the average steady state leg total energy. When idiff and udiff are constants, 3. Integration of this equation gives the steady state function of the energy difference described by 3.
Again, the integration constant WCdiff0 is freely selected to be the average steady state energy difference. The results are described by 3. The global structure of the open-loop control for a M2C is depicted on Figure 3. Again, both the continuous and the detailed model were investigated. The parameters used for simulation of the system are shown in Table 3. The switching frequency is still set to 2. The result of the classic modulation principle is a 6-level voltage, as expected. As illustrated in Figure 3.
As far as the circulating current in Figure 3. However, as the inner EMF reference eVpk is decreased, the circulating current peak-peak value decreases as well. The THD factor reduction proves the effectiveness of the controller in steady state, since the whole design concept was based on the mathematical assumption that the circulating current consists only of a plain DC component.
Tables 3. The voltage reference value eVpk is set to V. It is interesting to show the results of how the mechanism impacts the stored energy of each arm. The energy ripple is related to the capacitor voltages variation, whose mean value follows the reference with great accuracy. The stability of the system is therefore ensured. The role of the modulator once again is to keep each individual capacitor voltage within the same limits.
It is to be observed that there is almost no oscillation in the transient and steady state behavior of the converter phase leg. In addition, steady state is reached faster than in the closed-loop controller. There is more oscillation in the continuous model than in the detailed one. The reason for the sharp increase of the current again has to do with the fact that the capacitors are asked to be charged to a higher level.
Since this condition is achieved within shorter time than in the closed-loop control structure, the peak value of the idiff is also higher than in the feedback control. However, the peak value of the output voltage in the detailed model is slightly increased due to the fact that the total capacitors voltage in each arm becomes higher after the step energy change.
There is a fast response in the output voltage and current, as illustrated in Figures 3. Again Figure 3. Moreover, compared to the closed-loop controller, the total capacitors voltage reaches steady state faster, as shown in Figure 3. Again the system stability is ensured by the control mechanism that is imposed. The voltage reference eVpk was set once again to V, in order to acquire all the levels in the output phase voltage, as shown in Figure 3.
The output current depicted in Figure 3. Compared to the 6-levels concept the pulsed voltage across the arm impedance is thicker and shows larger steps Figure 3.
The THD factor is decreased, compared to the classic modulation concept, proving a good alternative operation condition for the M2C. The voltage balancing is divided into two types of controllers: a an averaging and b a balancing one. The results of the two controllers are then combined to extract references for each individual submodule. Finally, each submodule sinusoidal reference is compared to a different triangular carrier according to Carrier Shifted SPWM and is fed with pulses independently.
This means that there is no need for a balancing algorithm in the modulator, in contrast with the control structures described in previous paragraphs and chapters.
For this control method, it is assumed that the capacitor voltages are measured, thus their values are known at any time instant. The result of the voltage major control loop, which is given by 3. Finally, the voltage command obtained by the current minor loop u Aref is given by 3. The individual capacitor voltages are forced to follow their command uCref through the use of the balancing control, which is illustrated in Figure 3.
This balancing control scheme is based on the polarity of the currents and contributes to the arm voltage references with the value of uBref. When uCref is higher than uCj , a positive active power is to be taken from the DC side into the submodules.
When iU is positive, the product of uBjref and iU forms the positive active power. When iU is negative, the polarity of u Bjref is inversed to take the positive active power.
The same applies for the lower arm but with the opposite polarities for the current, according to Figure 3. This control law is finally represented by the equations 3.
This ensures harmonic elimination and current controllability. The equivalent switching frequency of the whole system is finally 2Nfc. The simulation converter parameters are the same with the closed loop control described in previous paragraphs.
The controller gains were set according to values proposed in [7]. The current waveform appearing in Figure 3. The effectiveness of the circulating current controller is proved by Figures 3. Due to higher switching frequency, the voltage across the line impedance Zarm appears to be thicker, as shown in Figure 3.
The balancing mechanism ensures that all the intermediate capacitor voltages are kept within the same limits, without the need of any balancing algorithm in the modulator. The discussion was made mainly in terms of system stability and power quality. The problems that were encountered in Chapter 2, where an ideal direct modulation control scheme was presented, were overcome through the use of effective control structures.
It has been shown that the circulating current idiff is the main way to directly influence the dynamics of the system. In all cases, the system was stabilized and gained very good transient behavior by trying to approach the condition, in which idiff consists of only a DC component in steady state.
In this way, the strong second order harmonic of the arm currents was sufficiently suppressed. In the closed-loop systems the stability was ensured based on the assumption that each submodule capacitor voltage is measured and available at each time instant. In the open-loop controller though, only the output current is measured and used for the estimation of the capacitor voltages.
This seems very promising, since it decreases the complexity of a real system significantly, where all the procedures are to function with great accuracy and precision. In the next chapter the implementation of control systems on a 10 kVA Modular Multilevel Converter is discussed for experimental validation of the above. The prototype consists of three phases and a number of 10 modules per phase leg.
The prototype consists of three phases and 5 modules per arm. It has been dimensioned with V per capacitor for a total rated power of 10 kVA.
The control of the converter is implemented in the processing unit, which manages with analog measurements, calculations and user interface. Finally, the FPGA processes the modulation data, performs the sorting algorithms and produces control signals for the converter. The reader is encouraged to refer to the Annexes at the end of the thesis for the detailed diagrams of the processor and FPGA programs, which were created to run the converter prototype.
The processor part was implemented as part of this thesis work [10]. However, for reasons of completeness and for a more comprehensive understanding of the experimental setup and the global control structure, description and discussion is made for the FPGA part, according to [9] and the prototype hardware as well, according to [8]. It is equipped with two switches on a heat sink, as depicted in Figure 4.
Each submodule needs the blocking ability of about V. In addition, they are easy to drive and their cost is quite low compared to the use of IGBTs. Since the switching frequency of the converter is quite low, the switching losses are not significant compared to the conducting losses. Therefore, it is desirable to choose switches with a small resistance RDSon.
The connectors for these signals are shown in Figure 4. The dead time of the switchings between the upper and lower MOSFETs is directly implemented in the driving unit and it is adjusted through an external resistor.
In Figure 4. As explained in the modulation concept description in previous paragraphs, the submodule capacitor voltages are to be available at each time instant, such that the modulator executes the sorting and selection algorithms. Therefore, the need for measuring all the submodule capacitor voltages is derived. The VCO converts the analog measurement signal of the capacitor voltage into a variable frequency digital signal, which is finally sent to the FPGA through optic fibers. The center frequency of the VCO is set to be adequately high, since this results in more accurate measurement.
For the submodules, electrolytic capacitors were chosen, since film capacitors in the milli-Farad range with a rated voltage of more than V are very large. Since a 3. Due to their design, these capacitors are relatively slow.
Moreover, their placement away from the submodules implies the use of cables, which present inevitable inductances. For this reason, film capacitors are placed in parallel to the electrolytic capacitors on the PCB board.
These capacitors are to supply fast transient currents, when the submodules are enabled, and also operate as filters to the oscillations produced by the creation of the LC circuit between the cable inductances and the submodule capacitor. A film capacitor of 4. The capacitors are to withstand high voltages of about V for the case of the failure of one or two submodules and a possible voltage oscillation between the electrolytic capacitor and the film capacitor.
In the presented prototype, electrolytic capacitors with a value of 3. Figure 4. For the analog measurements, an interface card between the control unit and the sensors is needed. This analog interface board initially supplies power to the sensors. It also receives, shapes the measurement signals and finally transmits them to the data acquisition unit. This card is also in charge of sending the arm current polarities to the FPGA, which are required by the modulator for the selection process, as mentioned already in previous chapters.
The polarity of the currents is also possible to be determined by implementing the differentiation of the capacitor voltages. However, this implicates a time delay. Since the arm currents need to be measured for the implementation of specific control systems as well, it was decided to directly specify their polarities from their value at any time instant. Through the analog interface card, accurate analog data flow from the converter to the processor and the FPGA is ensured.
The final form of the PCB for the three-phase system is illustrated in Figure 4. In addition, six voltage sensor connections are soldered, as shown in Figure 4. Again, two channels are provided as a backup. The card provides also a first order active noise filter for each measurement.
However, it is desirable to eliminate the noise as close to its origin as possible. The connection to data acquisition card is achieved through the connection ports, depicted in Figure 4.
Since this card is connected through a long cable, the digital current polarity signals need to be driven out using current drivers. The auxiliary power supply of the card is depicted in Figure 4.
Again, TracoPower converters are used. The reason is that they provide galvanic insulation, they are relatively easy to handle and their length does not cause problems such as being inductive or producing large time delays, even when it becomes considerably high.
0コメント