• 专利附录
  • 专利说明
  • 权利要求
  • 类似技术
  • 同族专利
  • 引用文献
  • 法律状态
  • 优先权
    • 专利摘要:
    The present invention provides a sequential or cascade predictive control method, first solving a cost function and then a second cost function for two or more control objectives, comprising:a) separate the cost function into at least two or more cost functions, depending on the number of control objectives;b) define control objectives;c) first control a variable with a unitary cost function, a single term or nature of the control objectives;d) determine the possible states that minimize the cost of the first objective to be controlled; ande) control a second variable with a cost function that minimizes the cost function therof considering only the options given in point d).
    公开号:EP3683957A1
    申请号:EP17924886.9
    申请日:2017-09-15
    公开日:2020-07-22
    发明作者:Margarita NORAMBUENA VALDIVIA；José RODRIGUEZ PÉREZ
    申请人:Universidad Andres Bello；
    IPC主号:G05B13-00
    专利说明:
    [0001] The present invention relates to the control of equipment or system processes, more specifically to a predictive control method, and more specifically a sequential or cascade predictive control method, first solving a cost function and then a second cost function for Two or more control objectives. DESCRIPTION OF THE PRIOR ART
    [0002] In our today society, electricity is a fundamental pillar in the various activities carried out day by day. The use and transport of electric energy is carried out using strategic equipment based on power semiconductors. A problem of interest today is the control of these equipment, called inverters / rectifiers.
    [0003] Today there are two major control proposals, one of them is linear control, a technology that has been in the industry for more than 60 years, however, the equipment used today is not linear systems and requires the linearization of its equations to operate at a specific point of work. In addition, this control strategy requires a thorough study of control techniques to be able to implement such control..
    [0004] The other major proposal that has begun to be implemented in recent years is the (MPC) model-based predictive control strategy, the great advantage of this strategy over linear control systems is its handling of nonlinear equations in a plain and simple way , without the need to resort to the linearization of system equations. However, one of the great shortcomings of this control strategy is its implementation, because at the time of wanting to control multiple variables it requires adjusting a weight or weighting factor, whose adjustment is not standardized and requires an empirical adjustment (test / error).
    [0005] The control objectives of a system can have the following characteristics: Different physical nature (current, voltage, active / reactive power, conduction losses, torque, flow, etc., for example in electronics power applications). Show coupling between variables. Different importance between them for the desired control result.
    [0006] These characteristics of the control objectives can be managed using weight factors or ranking tables, in order to obtain the desired behavior of the system. However, the adjustment of these weight factors is not normalized and will depend on the nature of the different terms involved, therefore, we can classify a cost function in different groups:1. Cost function without weight factor: in this case only variables of a single nature are controlled, for example, system currents, or system powers, or system voltages, etc. But only one of these variables at a time. 2. Cost function with secondary terms: in this case there is a main objective of control and a secondary one clearly marked, such as system currents and switching frequency of power semiconductors. In this case, the variable of importance is the currents and the secondary term is the switching frequency. 3. Cost function with equally important terms: in this case we cannot distinguish a term of bigger importance than the other, such as, for example, the control of torque and flow in an electric machine.
    [0007] Depending on the type of cost function that is obtained, a certain process of adjusting the weight factors can be carried out, for example, for the case of group 1 it is not necessary to adjust any weight factor, but for case 2 the adjustment process is not direct and an empirical adjustment is necessary through simulation checking and observation of the effect of the same, but for case 3 the process of selection of weight factors is further complicated, because it is not possible to distinguish what is the variable that is producing the final resulting effect that can be seen in the simulations performed in order to try to adjust it.
    [0008] In the state of the art we can find documents that describe standard predictive control, for example, documents US2017260760 (A1 ) of HAMILTON SUNDSTRAND CORP., Dated 08.06.2017, entitled "Model predictive control optimization for power electronics"; US2017133849 (A1 ) of ABB SCHWEIZ AG, dated 11.05.2017, entitled "Hierachical robust model predictive voltage and var control with coordination and optimization of autonomous der voltage control"; US2017133845 (A1 ) of ABB SCHWEIZ AG, dated 11.05.2017, entitled "Model predictive control of a modular multilevel converter", describes a single cost function that may or may not control more than one control objective through a factor of weight that discriminates the importance between one and another control objective.
    [0009] US2017139427 (A1 ) of MITSUBISHI ELECTRIC RES LABORATORIES INC, dated 18.05.2017, entitled "Model predictive control of spacecraft", describes a predictive control model for the position of the satellite / aircraft, this is not rectifier / inverter control, however, it shows the versatility of the MPC that can be applied to processes not only of an electrical / electronic nature in converters / inverters, in fact the MPC control arose in chemical processes.
    [0010] The model-based predictive control (MPC for its acronym in English: Model Predictive Control), has as its most attractive feature its conceptual simplicity and its facility to incorporate multiple control objectives in the same control strategy, however, both points are opposed having to define a certain weight factor to discriminate between one and another control term.
    [0011] As shown in Figure 1 (prior art), the main steps of the basic control strategy used in the 4 patent documents previously named are: 1. Find a mathematical model that predicts the operation of the system considering known actions. "Prediction Model" block. 2. Based on the possible actions (finite number of options), the optimal performance is determined according to a single criterion, which is the evaluation of a cost function that may or may not have more than one control objective. MPC block: "Cost Function Minimization g1". 3. The state that minimizes the cost function is applied to the converter or to the system to be controlled.
    [0012] The difference of all these patent documents lies solely in the type of application, in the case of documents US2017260760 and US2017133845 is the type of converter, which changes the prediction model and the respective cost function. As for document US2017139427 , the application is totally different from the other patent documents mentioned above, however, they respect the control diagram shown in Figure 1 and appropriately changing the prediction model and the cost function in order to follow the error in the objectives of control of its interest, maintaining the same structure of the figure already explained previously.
    [0013] However, the method that is proposed to be protected in the present application has a different predictive control application structure, without losing the ability to apply to the processes already described in the documents cited, that is, it is possible to replace the structure of figure 1 (prior art) by the sequential predictive control diagram of the present application (see figure 2) and obtain the specific solutions for the same applications of the prior art documents described, but separating the cost function into two or more, depending on the number of control objectives. In the case of figure 2 it is applied to an electric machine, in this case, the control objectives are Torque and Flow, therefore, predictive control is performed for one of these objectives and all those possible states are selected that minimize the error of this first variable, then a new iteration is carried out for the second control objective, seeking to minimize the error of this second variable, having as performance options only those that minimize the first variable as action options.
    [0014] Therefore, a sequential predictive control is presented that does not require determining any weight factor without adding complexity to the system, thus maintaining both value characteristics, conceptual simplicity and simple incorporation of multiple control objectives. BRIEF DESCRIPTION OF THE FIGURES
    [0015] Figure 1 (Prior Art) shows a standard predictive control scheme. Figure 2 shows a block diagram of the sequential predictive control of the present invention for an induction machine. Figure 3 shows graphs of different parameters of the sequential predictive control of the present invention for an induction machine, where (a) Flow; (b) Torque; (c) Angular velocity; (d) Current in phase a of the machine. Figure 4 shows graphs of different parameters of the sequential predictive control of the present invention for an induction machine, where (a) Voltage of a phase a (υaN); (b) Tension between lines (υab); (c) Current in phase a of the machine. DETAILED DESCRIPTION OF THE INVENTION
    [0016] a) The present invention provides a sequential or cascade predictive control method, first solving a cost function and then a second cost function for two or more control objectives, comprising:b) a) separate the cost function into at least two or more cost functions, depending on the number of control objectives;c) b) define control objectives;d) c) first control a variable with a unit cost function, a single term or nature of the control objectives;e) d) determine the possible states that minimize the cost of the first objective to be controlled; Yf) e) control a second variable with a cost function that minimizes its cost function considering only the options given in the point d).
    [0017] On the other hand, for an application that involves power electronics, specifically for inverter / rectifier applications, the method is defined as follows: 1. Define the mathematical model of the load / network (Mcarga). 2. Define the mathematical model of the converter / inverter (Minversor) 3. Define the control objectives for the load / network (xc1, xc2) 4. Define the control objectives for the inverter converter (xi1, xi2) 5. If the objectives of load control are of a different nature (voltage, current, power, harmonics) define a cost function for each of them. If they are of the same nature (only currents, only voltages, only powers, etc.) define a single cost function for the load / network (gc1, gc2) 6. If the control objectives of the converter / inverter are of a different nature (voltage, current, switching frequency, losses) define a cost function for each of them. If they are of the same nature (only currents, only voltages, only losses, etc.) define a single cost function for the converter / inverter (gi1, gi2) 7. Using Mcarga, evaluate the possible states for the load / network and evaluate gc1, determining the two states that minimize the cost function gc1, then evaluate gc2 for the preselected states and determine the state that minimizes gc2. 8. With the resulting state, select the redundancies of the existing converter / inverter for said state and evaluate gi1 using the Minversor by selecting those that minimize the cost function gc1, subsequently, of the redundant states that minimize gi1 evaluate gi2 obtaining the state that minimizes gi2. 9. Apply resulting state from the previous step. APPLICATION EXAMPLE
    [0018] The present invention provides a method of sequential predictive control as shown in Figure 2, where the application is for an induction machine consisting of first controlling a variable with a unitary cost function (a single term or control objective) and then the following terms. In this case of the control of an electric machine, the first variable to be controlled is the Torque, whose reference value is given, in this case, by the linear control of the angular speed of the machine. The two possible driving states of the converter that minimize the Torque are determined and then the predictive control of the Flow is carried out, which minimizes its cost function considering only the two options delivered by the Torque control process. The state that minimizes the error in the Stator Electric Flow is the one applied in the system.
    [0019] Using the method of sequential predictive control of the present invention, the control of the induction machine was performed, which presents control objectives where the cost function contains equally important terms, in this case a term of greater importance cannot be distinguished than the other, the most complex control when defining the weight factor. The results of the simulation of the control performed are shown in Figure 3 and Figure 4.
    [0020] The equations that describe the system for modeling the torque and electrical flow of the machine stator are given by:ψ r k + 1 = L r L mψ s k + 1 − 1 λL mi s k + 1
    [0021] Where Lr is the rotor inductance, Lm is the mutual rotor-stator inductance and λ is a factor given by the resistances, inductances and permeability of the machine, Np is the number of pole pairs of the machine, all known as machine parameters, and the superscript k + 1 indicates that it is the estimated value at the next instant of the sampling time k.
    [0022] The cost functions g1 and g2 are determined by:g 1 =T m − T r 2
    [0023] Where the subscript m indicates that the variable is measured and the subscript r is the reference value.
    [0024] It should be noted that this sequential predictive control SMPC (Sequential Model Predictive Control), can be used to control any system, regardless of its nature, as long as there are equations that model the behavior of the system and that allow predicting it, in the case of electrical systems, all of them can be controlled without problems by the SMPC, even those with terms of equal importance in the control objectives, such as the case of the induction machine.
    权利要求:
    Claims (9)
    [0001] 1. A method of sequential or cascade predictive control, first solving a cost function and then a second cost function for two or more control objectives, CHARACTERIZED because it comprises:
    a. define the mathematical model of at least one first variable to control;
    b. define the mathematical model of at least a second variable to control;
    c. separate the cost function into at least two or more cost functions, depending on the number of control objectives;
    d. define control objectives;
    e. first control a variable with a unit cost function, a single term or nature of the control objectives:
    f. determine the possible states that minimize the cost of the first objective to be controlled; and
    g. control a second variable with a cost function that minimizes the cost function therof considering only the options given in point d).
    [0002] A sequential or cascade predictive control method for inverters / rectifiers first solving a cost function and then a second cost function for two or more control objectives, CHARACTERIZED because it comprises:
    a. define the mathematical model of the load / network (Mcarga);
    b. define the mathematical model of the converter / inverter (Minversor);
    c. define the control objectives for the load / network (xc1, xc2)
    d. define the control objectives for the inverter converter (xi1, x2)
    e. define a cost function for each load / network control objective if the control objectives are of a different nature;
    f. define a single cost / network cost function (Gc1, Gc2) if the control objectives are of the same nature;
    g. define a cost function for each control objective of the converter / inverter if the control objectives are of a different nature;
    h. define a single cost function for the converter / inverter (gi1, gi2) if the control objectives are of the same nature;
    i. evaluate the possible states for the load / network using Mcarga, and evaluate gc1, determining the two states that minimize the cost function gc1, then evaluate gc2 for the preselected states and determine the state that minimizes gc2;
    j. select the redundancies of the existing converter / inverter for said state and evaluate gi1 using the Minversor by selecting those that minimize the cost function gc1, subsequently, of the redundant states that minimize gi1 evaluate gi2 obtaining the state that minimizes gi2; Y
    k. apply state resulting from the previous step j).
    [0003] The sequential predictive control method according to claim 2 CHARACTERIZED in that the inverter / rectifier is an induction machine.
    [0004] The sequential predictive control method according to claim 3 CHARACTERIZED because a first variable to be controlled is the Torque.
    [0005] The sequential predictive control method according to claim 3 CHARACTERIZED because a second variable to be controlled is the Flow.
    [0006] The method of sequential predictive control according to claim 4 CHARACTERIZED the mathematical model of the Torque is:T e k + 1 = 3 2 N p ψ s k + 1 ⊗ i s k + 1
    [0007] The method of sequential predictive control according to claim 5 CHARACTERIZED the mathematical model of the Flow is:ψ r k + 1 = L r L mψ s k + 1 − 1 λL mi s k + 1
    [0008] The sequential predictive control method according to claims 6 and 7, CHARACTERIZED because the cost functions g1 and g2 are determined by:g 1 =T m − T r 2
    [0009] The sequential predictive control method according to claims 3 to 8, CHARACTERIZED because the first variable to be controlled is the Torque, whose reference value is given, in this case, by the linear control of the angular velocity of the machine and the two possible driving states of the converter that minimize the Torque are determined and then proceed with the predictive control of the Flow, which minimizes the cost function thereof considering only the two options delivered by the Torque control process and where the state that minimizes the error in the electric flow of the stator is the one that is applied in the system.
    类似技术:
    公开号 | 公开日 | 专利标题
    US10074978B2|2018-09-11|Model predictive control of a modular multilevel converter
    Xiang et al.2014|Stabilization of switched continuous-time systems with all modes unstable via dwell time switching
    Fei et al.2019|Adaptive fuzzy-neural-network based on RBFNN control for active power filter
    Premkumar et al.2015|Fuzzy PID supervised online ANFIS based speed controller for brushless dc motor
    Brest et al.2009|Dynamic optimization using self-adaptive differential evolution
    Larsson et al.2002|Emergency voltage control using search and predictive control
    Bouallègue et al.2012|PID-type fuzzy logic controller tuning based on particle swarm optimization
    Etesami et al.2015|Colonial competitive algorithm development toward harmonic minimization in multilevel inverters
    CN104662431B|2017-06-23|Method and apparatus for controlling multiphase converter
    Zabarankin et al.2002|Optimal risk path algorithms
    Guo et al.2003|An adaptive fuzzy sliding mode controller for robotic manipulators
    Kumar et al.2018|Review of recent advancements of direct torque control in induction motor drives–a decade of progress
    Abd Ali et al.2016|Fuzzy logic speed controller optimization approach for induction motor drive using backtracking search algorithm
    Lin et al.2007|Recurrent-fuzzy-neural-network-controlled linear induction motor servo drive using genetic algorithms
    Gupta et al.2015|Optimized switching scheme of cascaded H-bridge multilevel inverter using PSO
    Pérez‐Ventura et al.2019|When is it reasonable to implement the discontinuous sliding‐mode controllers instead of the continuous ones? Frequency domain criteria
    Kaviani et al.2009|PSO, an effective tool for harmonics elimination and optimization in multi-level inverters
    KR20150074140A|2015-07-01|Model predictive control with reference tracking
    Lu et al.2004|Switching LPV control designs using multiple parameter-dependent Lyapunov functions
    Izzo et al.2019|Machine learning and evolutionary techniques in interplanetary trajectory design
    Singh et al.2012|Optimal coordination of directional overcurrent relays: A genetic algorithm approach
    Thielemans et al.2012|Weight factor selection for model-based predictive control of a four-level flying-capacitor inverter
    Kundu et al.2017|Comparative study between different optimisation techniques for finding precise switching angle for SHE-PWM of three-phase seven-level cascaded H-bridge inverter
    Chebre et al.2011|Speed control of induction motor using genetic algorithm-based PI controller
    Sadouni et al.2012|Indirect rotor field-oriented control | of a dual star induction machine | using a fuzzy controller
    同族专利:
    公开号 | 公开日
    CN111656673A|2020-09-11|
    WO2019051621A1|2019-03-21|
    EP3683957A4|2021-04-28|
    JP2021502050A|2021-01-21|
    引用文献:
    公开号 | 申请日 | 公开日 | 申请人 | 专利标题
    法律状态:
    2019-03-22| STAA| Information on the status of an ep patent application or granted ep patent|Free format text: STATUS: THE INTERNATIONAL PUBLICATION HAS BEEN MADE |
    2020-06-19| STAA| Information on the status of an ep patent application or granted ep patent|Free format text: STATUS: REQUEST FOR EXAMINATION WAS MADE |
    2020-06-19| PUAI| Public reference made under article 153(3) epc to a published international application that has entered the european phase|Free format text: ORIGINAL CODE: 0009012 |
    2020-07-22| 17P| Request for examination filed|Effective date: 20200414 |
    2020-07-22| AK| Designated contracting states|Kind code of ref document: A1 Designated state(s): AL AT BE BG CH CY CZ DE DK EE ES FI FR GB GR HR HU IE IS IT LI LT LU LV MC MK MT NL NO PL PT RO RS SE SI SK SM TR |
    2020-07-22| AX| Request for extension of the european patent|Extension state: BA ME |
    2020-12-23| DAV| Request for validation of the european patent (deleted)|
    2020-12-23| DAX| Request for extension of the european patent (deleted)|
    2021-04-28| A4| Supplementary search report drawn up and despatched|Effective date: 20210331 |
    2021-04-28| RIC1| Information provided on ipc code assigned before grant|Ipc: G05B 13/04 20060101AFI20210325BHEP Ipc: H02P 23/14 20060101ALI20210325BHEP Ipc: H02P 27/06 20060101ALI20210325BHEP |
    优先权:
    申请号 | 申请日 | 专利标题
    [返回顶部]