Method of evaluating states of electronic power system

FIELD: physics.

SUBSTANCE: disclosed is a method of evaluating states of an electronic power system (1), having a converter (4), wherein system state vectors x(k) and x(k+1) for each of the discretisation moments k=-N+1,…,0 are varied such that the sum of the vector norm from subtracting the system state vector (k+1) and a first function f(x(k), u(k)) of the system model and the vector norm from subtracting the vector y(k) of the output value and a second function g(x(k), u(k)) of the system model for discretisation moments k=-N+1,…,0 is minimum; the system state vector x(k) at discretisation moment k=0 is then selected.

EFFECT: enabling state evaluation for a wide range of electronic power systems.

12 cl

The technical field

The invention relates to the field of evaluation methods in engineering regulation. It comes from a method of evaluation of state power electronic system in accordance with the restrictive part of the independent claim.

The level of technology

Currently, power electronic systems are used in many areas. Such a power electronic system usually includes a Converter with a number of controllable power semiconductor switches and associated control circuit for a power semiconductor circuit. With the Converter typically connected to one or more loads, which, however, can greatly vary depending on time, for example due to failure. This load can be, for example, one or more engines, and is possible in General, any electrical load. State power electronic systems, such as inductive load current capacitive load voltage, affected by such changes and labour, i.e. only with high costs, or in General not be determined, for example by measuring. Therefore, it is necessary to assess the condition of power electronic systems with estimated state can then be processed in the control unit. A common way of assessing the state power is elektronnoy system is the use of discrete-time Kalman filter, as described, for example, in "Braided extended Kalman filter for sensorless estimation in inductionmotors at high-low/zero speed", IET Control Theory Appl., 2007. For evaluation of conditions, for example using a discrete-time Kalman filter, it is necessary to carry out the first of the following steps:

(a) determining the vectors y(k) output values for the moments of sampling k=-N+1 to k=0, where N is specified by horizon sample rate, and y is the output variable, for example the output voltage of the Converter, which is determined, for example, by measuring;

b) determining the vectors u(k) regulatory impacts for the moments of sampling k=-N+1 to k=0, and the regulatory impact is, for example, the control factor of the transducer;

C) determining the first function f(x(k), u(k)) of the model system at the time of sampling k for a description of power electronic systems, the function depends on the vector u(k) regulatory impact and the vector u(k) is the system state at time sample k;

g) determining a second function g(x(k), u(k)) of the model system at the time of sampling k for a description of power electronic systems, the function depends on the vector u(k) regulatory impact and the vector u(k) is the system state at time sample k.

The problem when using discrete-time Kalman filter to estimate the States x power e is ectroni system is what are the side conditions of the States, for example, that the current of the inductive load and/or voltage of the capacitive load is limited or can be negative, they can be considered only with great cost or not at all can be taken into account. Another problem for the Kalman filter are functions of the system model f(x(k), u(k)), g(x(k), u(k)), which are piecewise affine-linear functions and describe the power electronic system. The problem is that they either can't be done or can only be extremely costly when assessed through a discrete-time Kalman filter.

Disclosure of inventions

Object of the invention is to provide a method for estimation of state power electronic systems, which would provide an assessment of conditions for a wide range of power electronic systems and would be easy to implement. This problem is solved by the characteristics of claim 1 of the formula. In dependent clauses are shown preferred embodiments of the invention.

In the method of estimating the States of a power electronic system contains a Converter circuit. The method includes the following steps:

(a) determining the vectors y(k) output values for the moments of sampling from k=-N+1 to k=0, and N is set by the horizon sampling;

b) determining the vectors u(k) re Wirayuda impacts for the moments of sampling from k=-N+1 to k=0;

C) determining the first function f(x(k), u(k)) of the model system at the time of sampling k for a description of power electronic systems, and this function depends on the vector u(k) regulatory impact and the vector x(k) is the system state at time sample k;

g) determining a second function g(x(k), u(k)) of the model system at the time of sampling k for a description of power electronic systems, and this function depends on the vector u(k) regulatory impact and the vector x(k) is the system state at time sample k.

According to the invention, the estimate of the vector x(k) is the system state at the time of sampling k=0 includes the following steps:

d) change of vectors x(k) and x(k+1) system state for each of the sampling points from k=-N+1 to k=0, so the sum from the sum of the vector norms from the subtraction of the vector x(k+1) system state and the first function f(x(k), u(k)) model of the system and vector norms from the subtraction of the vector y(k) output value and the second function g(x(k), u(k)) model systems for all moments discretization k=-N+1 to k=0 becomes minimum;

(e) the choice of the vector x(k) is the system state at the time of sampling (k=0;

the first f(x(k), u(k)) and the second g(x(k), u(k)) functions model system are affine-linear or, alternatively, piecewise affine-linear.

Required estimated state at the current time k=sodergatsja then as elements of the vector x(k) is the system state at the time of sampling k=0, i.e. in the vector x(0) of the system state. Preferably, the proposed method allows us to take into account when assessing conditions adverse conditions, expressed piecewise affine-linear relationships of States and inputs. If a power electronic system is described by a piecewise affine-linear functions g(x(k), u(k)) model, such that the function g(x(k), u(k)) of the model system can also be very easily taken into account when estimating States. In General, the proposed method provides, thereby, assess conditions for a very wide range of power electronic systems and can be implemented very simply.

These and other objectives, advantages and features of the invention become apparent from the following detailed descriptions of preferred options for its implementation in conjunction with the drawing.

Brief description of drawings

In the drawing:

- figure 1: design of power electronic systems with a control unit and a device for carrying out the assessment it States proposed method.

Used in the drawing the reference positions and their meaning are listed in the list of items. In principle, in the drawing, the same parts are denoted by the same reference position. Described are an example of the object of the invention and do not have a limiting effect.

The implementation of the invention

Figure 1 image is Agen design power electronic systems 1 unit 3 control device 2, undertaking the assessment of its state x by the proposed method. One or more United with system 1 loads for clarity, not shown. The system 1 includes a transformer 4 with a number of controllable power semiconductor switches and associated control circuit 5 to control them by means of the control signal S. This control circuit 5 generates the control signal S, for example, by pulse-width modulation on the basis of regulatory impact u, which is, for example, the control factor of the inverter. Output value from the system 1 is, for example, the output voltage of the inverter is determined for example by measuring. The estimated States x of the system 1 are, for example, the current of the inductive load and the voltage of the capacitive load. Below the proposed method is described in more detail. In step a) are determined by the vectors y(k) output values for the moments of sampling k=-N+1 to k=0, and N is set by the horizon sampling. The elements of the vectors y(k) output values are then output quantities y, defined for example by measuring the output voltage of the Converter for the moments of sampling from k=-N+1 to k=0. At the stage b) are determined by the vectors u(k) regulatory impacts for the moments of sampling from k=-N+1 to k=0, PR is than the elements of the vectors u(k) are the regulatory impacts and for the moments of sampling from k=-N+1 to k=0, for example, the coefficients of the regulation. On the stage) is determined by the first function f(x(k), u(k)) of the model system at the time of sampling k for a description of power electronic systems 1-dependent vector u(k) regulatory impact and the vector x(k) is the system state at time sample k. Next, in step g) is defined by a second function g(x(k), u(k)) of the model system at the time of sampling k for a description of power electronic systems 1-dependent vector u(k) regulatory impact and the vector x(k) is the system state at time sample k.

The first function f(x(k), u(k)) of the model system at the time of sampling k for a description of the power electronic system 1 is defined, generally, as follows:

$$f(x(k)),u(k))=\{\begin{array}{l}{A}_{1}x(k)+B_{1}u(k)+v_1,F_{1}x(k)+E_{1}u(k)\le G_1\hfill \\ A_{2}x(k)+B_{2}u(k)+v_2,F_{2}x(k)+E_{2}u(k)\le G_2\\ \mathrm{...}\hfill \end{array}{\text{A}}_{\text{M}}x(k)+B_{M}u(k)+v_M,F_{M}x(k)+E_{M}u(k)\le G_M\hfill$$'

where A₁...A_MB₁...In_MF₁...F_Mand E₁...E_Mdenote matrices, v₁...v_M- vectors, and the vectors G₁...G₂- limits, which define the first function f(x(k), u(k)) of the model system as affine-linear or piecewise affine-linear. It should be said that by a suitable choice of the vectors G₁...G_Mv₁...v_Mand matrices A₁...A_MB₁...B_MF₁...F_ME₁...E_Mcan bitopology continuous affine-linear function, if the system 1 is described thus.

The second function g(x(k), u(k)) of the model system at the time of sampling k for a description of the power electronic system 1 is defined, generally, as follows:

$$g(x(k),u(k))=\{\begin{array}{l}{C}_{1}x(k)+D_{1}u(k)+w_1,F_{1}x(k)+E_{1}u(k)\le G_1\hfill \\ C_{2}x(k)+D_{2}u(k)+w_2,F_{2}x(k)+E_{2}u(k)\le G_2\hfill \\ \mathrm{..}\hfill \\ {\text{C}}_{\text{M}}x(k)+D_{M}u(k)+w_M,F_{M}x(k)+E_{M}u(k)\le G_M\hfill \end{array}$$,

where C₁...With_MD₁...D_MF₁...F_Mand E₁...E_Mdenote matrices, v₁...v_M- vectors, and the vectors G₁...G₂- the limits of which define a second function g(x(k), u(k)) of the model system as affine-linear or piecewise affine-linear. It should be said that by a suitable choice of the vectors G₁...G_M, w₁...w_Mand matrices With₁...C_MD₁...D_MF₁...F_ME₁...E_Mcan be achieved also continuous affine-linear function, if the system 1 is described thus.

According to the invention, the estimate of the vector x(k) is the system state at the time of sampling k=0, i.e. at the moment is described by the following additional steps:

d) change of vectors x(k) and x(k+1) system state for each of the sampling points from k=-N+1 to k=0, so the sum from the sum of the vector norms from the subtraction of the vector x(k+1) system state and the first function f(x(k), u(k)) system model the vector norms from the subtraction of the vector y(k) output value and the second function g(x(k), u(k)) model system for the moments of sampling from k=-N+1 to k=0 becomes minimum;

(e) the choice of the vector x(k) is the system state at the time of sampling k=0.

Required estimated state x at the current time k=0 are then as vector elements of the vector x(k) is the system state at the time of sampling k=0, i.e. the vector x(0) of the system state. These vector elements would then be, for example, the current of the inductive load and the voltage of the capacitive load at the time of sampling k=0. Preferably, the proposed method allows us to take into account when assessing the state x also their adverse conditions. If the system 1 is described affine-linear or piecewise affine-linear functions f(x(k), u(k)), g(x(k), u(k)) model, such that the function f(x(k), u(k)), g(x(k), u(k)) of the model system can be very easily taken into account in the assessment of state X. the Proposed method provides, thus, a score of States x for a very wide range of power electronic systems 1 and may to be implemented very simply.

Estimated by the proposed method, the state x can then be processed in the unit 3 control, ie, for example, adjusted to the appropriate set of States x_ref. The control unit 3 operates primarily on the principle of predictive control model-based, as it is known, for example, from EP 1670135 A1. But prob the wives of any other principle of management or any other characteristic of a control.

The above sum can be described as the sum of J according to the following formula:

$$J=\underset{k=-N+1}{\overset{0}{\Sigma }}(\Vert x(k+1)-f(x((k),u(k))\Vert w_{xq}+\Vert y(k)-g(x(k),u(k))\Vert w_{yq})$$

where W_xand W_ycorrespondingly represent the assessment matrix in respect of the vectors x(k)x(k+1) state vector, y(k) output values. The index q denotes the chosen vector norm. Preferably as a vector norms from the subtraction of the vector x(k+1) system state and the first function f(x(k), u(k)) of the model system is chosen norm of the sum of absolute values, i.e. for expression

$$\Vert x(k+1)-f(x(k),u(k))\Vert w_{xq}$$,

where q=1.

In addition, preferably also selected the norm of the sum of absolute values as the vector norm from subtraction of the vector y(k) output value and the second function g(x(k), u(k)) model of the system, i.e. for expression

$$\Vert y(k+1)-g(x(k)\Vert w_{yq},$$

where q=1. Preferably, the norm of the sum of absolute values, i.e. q=1, is very easy.

Alternatively, it is also possible that, as a vector norms from the subtraction of the vector x(k+1) system state and the first function f(x(k), u(k)) of the model system, and as a vector norms from the subtraction of the vector y(k) output value and the second function g(x(k), u(k)) model systems, respectively, were chosen Euclidean norm, i.e. q=2.

As another alternative, it is also possible that, as a vector norms from the subtraction of the vector x(k+1) system state and the first function f(x(k), u(k)) of the model system and the image quality is as vector norms from the subtraction of the vector y(k) output value and the second function g(x(k), u(k)) model systems, respectively, were selected maximum norm, i.e. q=∞.

It should be said that there are other possible standards. In addition, one can imagine that the rules for individual deduct Commission were also selected in different ways, i.e. it would be possible, for example, so as vector norms from the subtraction of the vector x(k+1) system state and the first function f(x(k), u(k)) model system was chosen norm of the sum of absolute values, then q=1, and as vector norms from the subtraction of the vector y(k) output value and the second function g(x(k), u(k)) model system is the Euclidean norm, i.e. q=2. While any combination.

As already mentioned, the vectors x(k) and x(k+1) system state for each of the sampling points from k=-N+1 to k=0 is changed so that the sum of J on all aspects of sampling from k=-N+1 to k=0 was minimal. These changes of vectors x(k) and x(k+1) system state for each of the sampling points from k=-N+1 to k=0 can be combined in the table (reference table), then each vector y(k) output values and the vector u(k) regulatory impact for each of the sampling points from k=-N+1 to k=0 correspond to the vector x(k) and the vector x(k+1) system state. The table should then take only the required vector x(k) is the system state at the time of sampling k=0, i.e. the vector x(0), while the elements of the vector x(0) comprising the Oia system are required estimated States x in this, i.e, the point k=0. This table can be prepared in advance, i.e. the "off-line"so you don't need to make an online calculation with a large amount of calculations with the criterion that the sum of J becomes minimum. The above table can be stored in the device evaluation 2 or on a separate drive, k accesses the device evaluation 2.

If there are enough resources, computing power, for example by the processor, in particular a digital signal processor, the changes of the vectors x(k) and x(k+1) system state for each of the sampling points from k=-N+1 to k=0 can be calculated continuously, i.e. online.

The reference list of items

1 - power electronic system

2 - device evaluation

3 - control unit

4 - Converter

5 - control scheme.

1. Method of assessment of state power electronic systems (1)containing the transducer (4), comprising the steps:
(a) determining the vectors y(k) output values for the moments of sampling k=-N+1, ..., 0, where N is specified by horizon sampling;
b) determining the vectors u(k) regulatory impacts for the moments of sampling k=-N+1, ..., 0,
C) determining the first function f(x(k), u(k)) of the model system at the time of sampling k for a description of power electronic systems, and this function depends on the vector u(k) reg is stimulating effects and the vector x(k) is the system state at time sample k;
g) determining a second function g(x(k), u(k)) of the model system at the time of sampling k for a description of power electronic systems, and this function depends on the vector u(k) regulatory impact and the vector x(k) is the system state at time sample k,
characterized in that
the estimate of the vector x(k) is the system state at the time of sampling k=0 further includes the steps are:
d) change the vectors x(k) and x(k+1) system state for each of the sampling points k=-N+1, ..., 0, so that the sum from the sum of the vector norms from the subtraction of the vector x(k+1) system state and the first function f(x(k), u(k)) model of the system and vector norms from the subtraction of the vector y(k) output value and the second function g(x(k), u(k)) model systems for all moments sampling k=-N+1, ..., 0 was the lowest; and
e) choose the vector x(k) is the system state at the time of sampling (k=0;
the first f(x(k), u(k)) and the second g(x(k), u(k)) functions model system are affine-linear.

2. The method according to claim 1, characterized in that as vector norms from the subtraction of the vector x(k+1) system state and the first function f(x(k), u(k)) models the system as a vector norms from the subtraction of the vector y(k) output value and the second function g(x(k), u(k)) model systems accordingly choose the norm of the sum of absolute values.

3. The method according to claim 1, characterized in that the quality of the ve vector norms from the subtraction of the vector x(k+1) system state and the first function f(x(k), u(k)) models the system as a vector norms from the subtraction of the vector y(k) output value and the second function g(x(k), u(k)) model systems accordingly choose the Euclidean norm.

4. The method according to claim 1, characterized in that as vector norms from the subtraction of the vector x(k+1) system state and the first function f(x(k), u(k)) models the system as a vector norms from the subtraction of the vector y(k) output value and the second function g(x(k), u(k)) model of the system respectively select the maximum norm.

5. The method according to claim 1, characterized in that the change of the vectors x(k) and x(k+1) system state for each of the sampling points k=-N+1, ..., 0 unite in the lookup table, in which each vector y(k) output values and the vector u(k) regulatory impact for each of the sampling points k=-N+1, ..., 0 correspond to the vector x(k) and the vector x(k+1) system state.

6. The method according to claim 1, characterized in that the change of the vectors x(k) and x(k+1) system state for each of the sampling points k=-N+1, ..., 0 is calculated continuously.

7. Method of assessment of state power electronic systems (1)containing the transducer (4), comprising the steps:
(a) determining the vectors y(k) output values for the moments of sampling k=-N+1, ..., 0, where N is specified by horizon sampling;
b) determining the vectors u(k) regulatory impacts for momentofinertia k=-N+1, ..., 0;
C) determining the first function f(x(k), u(k)) of the model system at the time of sampling k for a description of power electronic systems (1), and this function depends on the vector u(k) regulatory impact and the vector x(k) is the system state at time sample k;
g) determining a second function g(x(k), u(k)) of the model system at the time of sampling k for a description of power electronic systems (1), and this function depends on the vector u(k) regulatory impact and the vector x(k) is the system state at time sample k,
characterized in that
the estimate of the vector x(k) is the system state at the time of sampling k=0 further includes the steps are:
d) change the vectors x(k) and x(k+1) system state for each of the sampling points k=-N+1, ..., 0, so that the sum from the sum of the vector norms from the subtraction of the vector x(k+1) system state and the first function f(x(k), u(k)) model of the system and vector norms from the subtraction of the vector y(k) output value and the second function g(x(k), u(k)) model systems for all moments sampling k=-N+1, ..., 0 was the lowest; and
e) choose the vector x(k) is the system state at the time of sampling (k=0;
the first f(x(k), u(k)) and the second g(x(k), u(k)) functions model system are piecewise affine-linear.

8. The method according to claim 7, characterized in that as vector norms from the subtraction of the vector x(k+1) comprising the Oia system and the first function f(x(k), u(k)) models the system as a vector norms from the subtraction of the vector y(k) output value and the second function g(x(k), u(k)) model systems accordingly choose the norm of the sum of absolute values.

9. The method according to claim 7, characterized in that as vector norms from the subtraction of the vector x(k+1) system state and the first function f(x(k), u(k)) models the system as a vector norms from the subtraction of the vector y(k) output value and the second function g(x(k), u(k)) model systems accordingly choose the Euclidean norm.

10. The method according to claim 7, characterized in that as vector norms from the subtraction of the vector x(k+1) system state and the first function f(x(k), u(k)) models the system as a vector norms from the subtraction of the vector y(k) output value and the second function g(x(k), u(k)) model of the system respectively select the maximum norm.
11 the Method according to claim 7, characterized in that the change of the vectors x(k) and x(k+1) system state for each of the sampling points k=-N+1, ..., 0 unite in the lookup table, in which each vector y(k) output values and the vector u(k) regulatory impact for each of the sampling points k=-N+1, ..., 0 correspond to the vector x(k) and the vector x(k+1) system state.

12. The method according to claim 7, characterized in that the change of the vectors x(k) and x(k+1) system state for each of the sampling points k=-N+1, ..., 0 compute C erevna.