Method for determining insulation resistance and discharge capacitance of an ungrounded power supply system
The method addresses the challenge of low-frequency line voltage variations in ungrounded power supply systems by employing unipolar or bipolar coupling and recursive QR decomposition, facilitating efficient and continuous measurement of insulation resistance and leakage capacitance.
Patent Information
- Authority / Receiving Office
- KR · KR
- Patent Type
- Patents
- Current Assignee / Owner
- BENDER GMBH & CO KAGE
- Filing Date
- 2024-03-19
- Publication Date
- 2026-07-15
AI Technical Summary
Existing methods for determining insulation resistance and leakage capacitance in ungrounded power supply systems are hindered by low-frequency line voltage variations, leading to unstable measurements and increased computational complexity, particularly in PV installations where solar radiation causes fluctuations.
A method utilizing unipolar or bipolar coupling of a measuring voltage with active conductors, combined with recursive QR decomposition of measurement-value matrices, allows for rapid and robust determination of insulation resistance and leakage capacitance by integrating line voltage into current-voltage relationships without the need for filtering, enabling continuous measurements even under low-frequency line voltage fluctuations.
Enables fast and reliable identification of insulation resistance and leakage capacitance, reducing computational effort and storage requirements, while adapting to changing network parameters, thus overcoming the limitations of existing methods.
Smart Images
Figure 112025108541695-PCT00033_ABST
Abstract
Description
Technology Field
[0001] The present invention relates to a method for determining the insulation resistance and leakage capacitance of an ungrounded power supply system according to the preamble of claim 1. Background Technology
[0002] To meet high demands for operational safety, fire safety, and touch safety, network configurations of ungrounded power supply systems are used, which are also referred to as isolated networks (IT networks) or IT power supply systems (French: islands and lands It is referred to as an IT. In this type of power supply system, the active parts are isolated from the ground potential—with respect to "ground." The advantage of such a network is that, because a closed circuit cannot be formed between the network's active conductor and ground due to an ideally infinitely large impedance value, the function of the connected electrical consumer is not disrupted during an insulation fault (first fault), such as the accidental grounding of the active conductor in an ungrounded power supply system. As the actual components (real parts) in the parallel circuit possess leakage capacitance as virtual components, the electrical resistance associated with the ground potential (with respect to ground) in this context forms the complex-valued insulation impedance of the ungrounded power supply system.
[0003] Accordingly, since another possible fault (second fault) in another active conductor can cause a fault loop and the resulting fault current flow can cause the shutdown of the facility, including the shutdown of operation, in conjunction with the overcurrent protection device, the electrical resistance of the ungrounded power supply system relative to the ground potential, referred to as insulation resistance, must be monitored using a standard insulation monitoring device (IMD).
[0004] In addition to passive insulation monitoring devices that use the line voltage of an ungrounded power supply system as a driving source for a measuring current to detect insulation faults, actively operating insulation monitoring devices are known at the level of the art. These devices have a measuring path extending between one or more active conductors of an ungrounded power supply system and the ground potential, and include an internal measuring-voltage generator. The measuring voltage generated by the measuring-voltage generator actively drives a measuring current, and this measuring current flows back into the measuring path through the active conductor(s) and through the insulation resistance and leakage impedance, causing a voltage drop across a measuring resistance switched in series with the measuring-voltage generator. The insulation resistance and leakage impedance are determined using the voltage drop recorded across the measuring resistance.
[0005] An active method is known for superimposing a square-shaped measuring voltage consisting of continuous measuring pulses over an ungrounded power supply system to be monitored. However, a reliable calculation of insulation resistance is possible only when the measuring voltage is stable, which can take several minutes at larger leakage capacitances.
[0006] In addition, undesirable but inevitable line voltage variations can interfere with the measurement. High-frequency line voltage variations (greater than a few Hertz) can be removed by a filter. However, low-frequency line voltage variations of only a few Hertz are problematic because they hinder the verification of stability and distort the calculated insulation resistance. Consequently, large voltage variations within the low-frequency domain make measurement impossible because the measured voltage is unstable.
[0007] Patent EP 2 433 147 B1 discloses a method for identifying insulation resistance before the measured voltage at the measured resistance reaches a stable state. In this case, the stabilization process is predicted by a mathematical model, and the parameters of this mathematical model are iteratively adapted until the theoretical and the measured voltage curve match as closely as possible. Subsequently, the insulation resistance can be calculated based on the model parameters. Interfering line voltage variations are compensated by a filter and by the formation of a difference between two consecutive measured pulses. However, the removal of low-frequency line voltage variations also appears to be problematic. These drawbacks persist in the time-consuming calculation of the matrix inverse transformation to determine the model parameters. means of solving the problem
[0008] The object of the present invention is to provide the ability to perform measurements of insulation resistance and leakage capacitance as quickly, reliably, and robustly as possible in an ungrounded power supply system; in particular, to minimize interference factors of low-frequency line voltage variations with efficient implementation from a computational perspective.
[0009] This objective is achieved by a method having the features of claim 1.
[0010] Based on the technical level of the industry, a measuring voltage is coupled in series to a measured resistance in a unipolar or bipolar manner with one of the active conductors in each case, and a voltage drop—induced by a measuring current driven by the measuring voltage—is measured through the corresponding measured resistance.
[0011] In DC power supply systems, positive coupling has proven useful because it particularly efficiently resolves the problem of low-frequency line voltage variations. In alternating current (AC) and three-phase alternating current (3AC) power supply systems, unipolar coupling is sufficient as line frequency can be removed by filtering.
[0012] From the voltage curves of the measurement voltage, line voltage, and voltage drop across the corresponding measurement resistance, a time and value-discrete sample sequence of the measurement voltage, line voltage, and measurement voltage is generated.
[0013] The continuous signal curves of the line voltage (nominal voltage of an ungrounded power supply system), the applied measuring voltage generated by the measuring-signal generator, and the voltage drop captured by the measuring resistor are converted into time and value-discrete signals by a sampling device (analog-to-digital converter (A / D converter)), and thus made accessible as a sample sequence for digital signal processing.
[0014] The sample sequences generated accordingly form the input and output variables of the functional equivalent circuit of an observed ungrounded power supply system with insulation monitoring, and its mathematical description is provided as a linear network (current-voltage relationship) by physical laws (Ohm's law) and Kirchhoff's laws.
[0015] Based on these laws, a linear difference equation is implemented, wherein the measured voltage (corresponding to the sum of the measured voltage drops for unipolar coupling or bipolar coupling) can be expressed as a function of the measured voltage and the line voltage, and the network parameters of the ungrounded power supply system to be identified represent the coefficients of the linear difference equation. In this context, the network parameters are formed from the measured resistance, the insulation resistance to be determined, and the leakage capacitance to be determined. For the simplification of calculations, the resistance value is expressed as a conductance value.
[0016] By integrating the line voltage into the current-voltage relationship of the linear difference equation, it is not necessary to remove the line voltage from the measurement voltage, for example, by filtering or other signal-processing means. This advantageously reduces the effort associated with the circuitry and allows the measurement method to prevent interference.
[0017] In addition, faster and continuous measurements are possible for slow, low-frequency changes in line voltage. This is particularly advantageous in PV installations because the DC line voltage fluctuates according to the intensity of solar radiation.
[0018] Compared to the general method of the prior art, in which a stable line voltage and a stabilized measurement current are assumed to identify insulation resistance, uninterrupted measurement can be continued using the method according to the present invention.
[0019] In the following steps, a measured-value equation system of N linear difference equations is implemented from a sample sequence of a measured voltage, line voltage, measured voltage, and network parameter for k = 1, 2 to N measurement times with a sample period T.
[0020] An overdetermined measured value equation system is derived for time points where N > 4. As a general rule, a solution vector, and consequently a set of network parameters (coefficient vector) that correctly solve all current N linear difference equations, does not exist in such an overdetermined-value equation system.
[0021] For this reason, the estimated network parameters are calculated as an approximate solution to the measure-value equation system, and the sum of the squared residuals between the (actual) network parameters and the estimated network parameters is minimized.
[0022] Next, the estimated network parameters are considered optimal in the sense of an approximate solution when the sum of the squared residuals resulting from the remaining (residual) error between the actual network parameters and the estimated network parameters is minimized.
[0023] These minimization or compensation problems are solved by minimizing the sum of squared residuals by QR decomposition of the measurement-value matrix, which features a system of measurement-value equations, and the QR decomposition is calculated recursively.
[0024] A system of measurement-value equations containing N linear difference equations can be written as a measurement-value matrix equation, where a sample sequence of measurement voltages (measurement-value vectors) is derived as a factor by multiplying the measurement-value matrix by a coefficient vector (network-parameter vector). The elements of the measurement-value matrix correspond to line voltages, samples of measurement voltages, and preceding samples of measurement voltages (due to the iterative nature of different equations).
[0025] The inverse matrix transformation, which requires significant computation and is necessary for solving minimization problems by the gradient method, can be avoided by QR decomposition of the measurement-value matrix. Under undesirable conditions, such as when the measurement voltage or line voltage is zero, this will result in zero columns and consequently a matrix that cannot be inversely transformed. In this case, the inverse matrix transformation will not be possible, in contrast to QR decomposition.
[0026] In contrast, QR decomposition is numerically significantly more robust and enables the computation of estimated network parameters even from poorly conditioned measurement-value matrices. The computation of estimated network parameters as intended by the present invention is sensitive to incorrect measurement data and rounding errors, and accordingly, it becomes more accurate.
[0027] Starting from the geometric interpretation of the minimization problem, error squares can be viewed as the square of the Euclidean norm. Through this, the minimization problem can be attributed to a QR matrix equation by applying the QR decomposition of the measure-value matrix. In contrast to the measure-value matrix equation representing a system of measure-value equations composed of different equations, the (matrix) equation that solves the minimization problem and represents the result vector as a matrix product of the upper triangle matrix and the estimated network-parameter vector is described here as the QR matrix equation.
[0028] The solution for the estimated network-parameter vector is identified by inverse substitution using the upper triangle matrix (R) of the previously (recursively) calculated QR decomposition and the resulting vector of the (recursively) determined QR matrix equation.
[0029] According to the present invention, the QR decomposition is calculated recursively. In this context, the results calculated in each method cycle from the QR decomposition—the upper triangle matrix (R) and the resulting vector of the QR equation—are updated in the following cycle by considering the currently available measurements, which means that a stepwise approximation of the actual network parameters is made.
[0030] As only the currently available set of measured values is processed, the need for storage and computational effort are significantly reduced by recursive calculation. This means that it can be efficiently implemented in a microcontroller.
[0031] In addition to the advantage of requiring minimal storage space and computation time, recursive QR decomposition enables rapid adaptation to changing (actual) network parameters. If the currently effective insulation resistance or leakage capacitance changes, measurement and calculation can be easily performed through continuous adaptation.
[0032] The corresponding conductor-related insulation resistance and corresponding leakage capacitance are calculated from the estimated network parameters.
[0033] As is common practice in the industry, insulation resistance and leakage capacitance are not calculated directly from the voltage drop measured through a measuring resistance, but rather from estimated network parameters. Therefore, to perform reliable calculations, it is not necessary for the measured voltage to reach a strictly settled state.
[0034] On the one hand, the method according to the present invention can rapidly identify insulation resistance and leakage capacitance, and on the other hand, can utilize greater freedom when selecting the signal shape of the voltage to be measured. Thus, for example, a sinusoidal voltage to be measured or a mixture of sinusoidal voltage curves from different frequencies can be selected as the measurement signal.
[0035] The method step is repeated with the estimated network parameters calculated in correspondence, while considering the samples available at the time of current measurement.
[0036] Therefore, practical adaptation to the currently available insulation state of the ungrounded power supply system is achieved, and resource-efficient and rapid identification of insulation resistance and leakage capacitance is ensured by recursive calculation.
[0037] In another embodiment, the linear difference equation is derived by converting the linear algebraic equation in the frequency domain into a time-continuous difference equation and its time-discrete implementation, and the linear algebraic equation describes the current-voltage relationship.
[0038] The starting point for the implementation of the linear difference equation is to describe the current-voltage relationship in the frequency domain (image domain) through linear algebraic equations, and the current-voltage relationship is derived from the equivalent circuit diagram of an observed ungrounded power supply system that is isolated and monitored. Preferably, a Laplace transform is used to describe the relationship within the frequency domain. When transferred to the time domain, this becomes a time-discrete differential equation, and the linear difference equation follows from the subsequent time-discrete. Its application to successive measurement points (samples) yields a system of measurement-value equations that can be expressed as measurement-value matrix equations and has measurement-value matrices for determined network parameters.
[0039] Preferably, recursive QR decomposition is performed based on a recursive matrix.
[0040] An initial matrix having appropriate initial parameters for the upper triangle matrix of the QR decomposition and the result vector of the QR matrix equation is enhanced by a set of current measurement values for the recursive matrix during the initialization phase of the method. In each subsequent method cycle, this set of measurement values is replaced with a corresponding set of current measurement values. The upper triangle matrix (R) and the result vector are recalculated iteratively in each method cycle, taking into account the corresponding set of current measurements.
[0041] In addition, QR decomposition is performed by Givens rotation.
[0042] By the method of Gibbons rotation to compute the upper triangle matrix, the rotation matrix is computed in each method cycle, and zero entries are generated in the recursion matrix in the target method. Consequently, the updated upper triangle matrix (R) and the updated result vector can be used at the end of each method cycle, and the QR matrix equation can be solved through inverse substitution using the updated upper triangle matrix (R) and the updated result vector, thereby determining the estimated network-parameter vector.
[0043] Advantageously, the upper triangle matrix and the resulting vector are weighted using the Forgetting Factor in the recursive QR decomposition.
[0044] The weights of the measured values are introduced by multiplying the upper triangle matrix (R) of the QR matrix equation and the result vector by a factor (pocketing factor).
[0045] The pocketing factor has the effect of weighting the currently measured value more heavily than the previously measured value. It typically has a value ranging from 0.95 to 1. The larger the factor, the more older the measured value is considered in the calculation. Consequently, parameter changes are identified more slowly; however, short-term interference is filtered out more effectively. Conversely, a smaller factor enables faster identification of parameter changes, with the disadvantage that measurements are more susceptible to interference. Recursive calculations using the pocketing factor allow for tracking such changes.
[0046] Additional advantageous embodiments are derived from the following description and drawings, which describe preferred embodiments of the present invention using examples. Brief explanation of the drawing
[0047] Figure 1 illustrates a functional equivalent circuit diagram of an insulated, monitored, ungrounded power supply system. FIG. 2 illustrates a flowchart of a method according to the present invention. Figure 3 shows the measured voltage in a square-pulse shape. Figure 4 illustrates the voltage drop measured through the measured resistance when the line voltage changes. Figure 5 illustrates the temporal progression of the calculated insulation resistance. Figure 6 illustrates the temporal progression of the calculated leakage capacitance. Specific details for implementing the invention
[0048] Figure 1 shows the monitored line voltage (U N A functional equivalent circuit diagram of an ungrounded power supply system (2) having ) is shown.
[0049] The method according to the present invention can be applied to both DC power supply systems and single-phase or multi-phase AC power supply systems, and the measurement voltage (U G In this example, the measuring voltage) is coupled to two active conductors (L1, L2) in a DC power supply system in a bipolar manner.
[0050] In each case, the effective insulation impedance between the active conductors (L1, L2) and the ground (PE) is represented by the insulation resistance (R1, R2) (the real part of the insulation impedance) in a parallel circuit having leakage capacitance (C1, C2) (the imaginary part of the insulation impedance).
[0051] For insulation monitoring, the measurement-voltage generator uses a measurement voltage (U G It generates ), and this measurement voltage flows through the active conductors (L1, L2), through the insulation resistors (R1, R2), and through the leakage capacitance (C1, C2), and the measurement resistor (R M Voltage drop (U) at , measured resistance) M1 , U M2By driving a measuring current that induces ), the voltage drop (U M1 , U M2 ) is measured and evaluated to determine insulation resistance (R1, R2) and leakage capacitance (C1, C2).
[0052] Actually, the measurement voltage (U G ) is coupled through a series connection having a high-impedance coupling resistor and a low-impedance measured resistance. For the simplicity of calculation, both resistors are used as the corresponding measured resistance (R M It was summarized as ).
[0053] Figure 2 illustrates a flowchart of the method according to the present invention.
[0054] After the initialization step of the method in which initial values of parameters for the upper triangle matrix of QR decomposition and the result vector of the QR matrix equation are set, the measurement voltage (U G ) is switched in steps (S1 and S2) and the corresponding voltage drop (U M1 , U M2 The measurement of ) is due to the measurement current, and the measurement resistance (R M It is stabilized through ).
[0055] Through sampling and quantification, the measurement voltage (U G ), line voltage (U N ) and voltage drop (U M1 , U M2 Time and value-discrete sample sequences U G (k), U N (k), U M (k) is generated in step (S3).
[0056] A linear difference equation is implemented in step (S4) based on the current-voltage relationship derived from the functional equivalent circuit diagram.
[0057] Measured voltage (U M )(This is a voltage drop (U) due to the positive coupling illustrated exemplarily M1 , U M2 Composed of the sum of ) and the voltage drop (U) for unipolar coupling only M1 , U M2 It starts with an explanation within the frequency domain through algebraic equations for (which can only correspond to ).
[0058]
[0059] (Equation 1)
[0060] The time-continuous differential equation is the measured voltage (U M This leads to transferring ) to the time domain, and the linear difference equation is derived from subsequent time discretization with the index (k) of the sample sequence and the sample period T; in the linear difference equation, the measurement voltage (U G (k)) and line voltage (U N The measured voltage (U) which is a function of (k)) M (k)) can be expressed as follows.
[0061]
[0062] (Equation 2)
[0063] In matrix notation, the following is derived.
[0064]
[0065] (Equation 3)
[0066] At measurement points k = 1, 2 to N, an implementation of an overdetermined measured-value equation system is derived therefrom:
[0067]
[0068] (Equation 4)
[0069] This is a measured-value vector y ), measurement-value matrix (Ψ), and network-parameter vector (coefficient vector)( i It can be written as a measurement-value matrix equation having ):
[0070] y=Ψθ
[0071] (Equation 5)
[0072] Network-parameter vector( i ) is an element( i i It consists of )(network parameters), and these elements are derived from the network configuration variables to be determined (G1 = 1 / R1, G2 = 1 / R2, C1 and C2) and the measured resistance (R M It can be formed from ).
[0073] Since the overdetermined equation system generally does not have a solution, an approximation result is calculated in step (S6), and ideally, the (residual) error (which represents only measurement noise) e ) remains in the approximation result. Network parameter vector of equation (4) and equation (5) i From ), estimated network parameters( Estimated network-parameter vector having ) ) is produced accordingly:
[0074]
[0075] (Equation 6)
[0076] error( e If the sum of the squares of ) is minimized, the estimated network parameters ( )(network-parameter vector( The element of ) is considered optimal:
[0077]
[0078] (Equation 7)
[0079] The result of this minimization problem by the gradient method requires an inverse matrix transformation, and the calculation can be numerically unstable. To bypass this inverse matrix transformation, equation (7) is interpreted geometrically and considered as the square of the Euclidean standard:
[0080]
[0081] (Equation 8)
[0082] In step (S7), the (complete / whole) QR decomposition is an orthonormal matrix ( Q ) and upper triangle matrix( R Applied to the measurement-value matrix (Ψ) having ), and 0 is a zero matrix.
[0083]
[0084] (Equation 9)
[0085] Due to orthogonality, on the left side Q T If you multiply by Q T Q =1 is produced:
[0086]
[0087] (Equation 10)
[0088] Measurement-value vector( y The product of ) can likewise be divided as follows:
[0089]
[0090] (Equation 11)
[0091] Here c is described as a result vector, and the vector( ) represents the remaining error.
[0092] Q Since it is orthonormal, the Euclidean standard does not change, and the reason is that on the left sideQ T This is because multiplying—such multiplication preserves length—and the formula of the equation (8) to be minimized is transformed into the following.
[0093]
[0094] (Equation 12)
[0095] Formula (12) within the equation ) is the QR matrix equation
[0096]
[0097] (Equation 13)
[0098] It has a minimum value when this is satisfied. Subsequently, the estimated network-parameter vector ( ) is the upper triangle matrix( R ) and result vector( c It can be easily determined by inverse substitution considering ).
[0099] R and c A recursive method is used for the calculation of. As an initial value, the identity matrix is R Appropriate initial network parameters selected for ( )go c It is selected for. In one example of four network parameter values to be estimated, the following matrix is produced as the identity matrix:
[0100]
[0101] (Equation 14)
[0102] This identity matrix is now the first set of measured values (ψ T It is enhanced by y') and leads to a recursive matrix.
[0103]
[0104] (Equation 15)
[0105] By Gibbons rotation, the rotation matrix (G) on the left side i,j By repeatedly multiplying ), a 0 (zero) element can be generated in the target manner. Element ( i , j If ) must become 0, the rotation is the line( i and j It only affects ). As a result, the recursive matrix from equation (15) can be transformed into the following form.
[0106]
[0107] (Equation 16)
[0108] Therefore, the following is calculated:
[0109]
[0110] (Equation 17)
[0111] The initial value is the newly calculated upper triangle matrix ( R' ) and the newly calculated result vector( c' It is repeatedly replaced by ). The newly calculated upper triangle matrix ( R' ) and the newly calculated result vector( c' ) with respect to the estimated network-parameter vector according to the QR matrix equation (13) ) can be determined by inverse substitution.
[0112] At the end of each method cycle, the corresponding insulation resistance (R1, R2) and corresponding leakage capacitance (C1, C2) are network parameters estimated in step (S8). It is calculated from ).
[0113] By considering the line voltage in the description of the current-voltage relationship (obtained from the equivalent circuit diagram), insulation resistance and leakage capacitance can be determined in a phase-segregated manner, that is, separately for each conductor. Distribution to individual active conductors simplifies fault locations in ungrounded DC power supply systems.
[0114] Based on the recursive matrices described in equations (15) and (16), the method cycle is a newly calculated upper triangle matrix ( R' ) and the newly calculated result vector( c' ) as the set of measured values (ψ) available at the time of current measurement T It is continuously repeated as y').
[0115] In this context, prior to the subsequent method cycle, the newly calculated upper triangle matrix ( R' ) and newly calculated c' pocketing factor ( By multiplying by ), the temporal weight of the measured value can be obtained.
[0116] FIG. 3 shows a square-pulse-shaped measurement voltage (U G ) is illustrated, which is generated by a measurement-voltage generator and superimposed on an ungrounded DC power supply system (2). As the maximum modulation amplitude is 1.2 V, the measurement voltage (U G ) has an amplitude of approximately 1 V. The pulse width is 8 s.
[0117] FIG. 4 shows the line voltage (U N When ) has a low-frequency line-voltage change of 0.1 Hz, the measured resistance (R M Voltage drop (U) measured through ) M1 ) city.
[0118] Voltage drop (U M1 ) is the changing line voltage (U N) and square-pulse-shaped measurement voltage (U G It consists of a superposition of ). It is evident that low-frequency line-voltage variations are dominant and thus exert interference on the measurement. Therefore, depending on the level of technology in the art, interference suppression, for example through filtering, is required. Measurements of this kind for interference suppression, particularly those requiring significant effort from a circuit perspective in low-frequency line-voltage variations, may not be necessary when applying the method according to the present invention. Advantageously, the present invention aims to detect usable measured values even under these difficult conditions, namely spoofed measuring signals.
[0119] FIG. 5 shows a line voltage (U) characterized by the line-voltage change of FIG. 4. N Insulation resistance (R) calculated from the stepwise (test-based) change in actual insulation resistance after 40 seconds under the condition (in this case, as the total resistance of the parallel circuit of insulation resistances (R1 and R2)). f Illuminates the temporal progression of ).
[0120] Calculated insulation resistance (R f Initially, about 20 seconds elapse until the actual value is nearly identical. After a stepwise change in the actual insulation resistance at 40 seconds, about 30 more seconds elapse until the measurement method provides the actual value of the insulation resistance again.
[0121] FIG. 6 shows the leakage capacitance (C) calculated from the stepwise change of the actual leakage capacitance after 40 seconds, similar to FIG. 5. e )(In this case, the time progression of the sum of the leakage capacitances (C1 and C2) is illustrated.
Claims
Claim 1 Line voltage (U N A method for determining the insulation resistance (R1, R2) and leakage capacitance (C1, C2) with respect to ground (PE) of an ungrounded DC power supply system having ) and active conductors (L1, L2): a voltage for measurement (U G ) in each case, between one of the active conductors (L1, L2) and ground (PE), the resistance (R) is measured in a unipolar or bipolar manner. M A step (S1) of coupling in series with the ) and the corresponding measuring resistor (R M Voltage drop (U) through ) M1 , U M2 It includes a step (S2) of measuring ) and the measurement voltage (U G ), the above line voltage (U N ) and the corresponding voltage drop (U M1 , U M2 Time and value-discrete sample sequence (U G (k), U N (k), U M (k)) generating step (S3), the measurement voltage (U G (k)) and the line voltage (U N Measurement voltage (U) as a function of (k) M (k)) having the above-mentioned measuring resistance (R M A network parameter (θ) formed from the insulation resistance (R1, R2) to be determined and the leakage capacitance (C1, C2) to be determined i Step (S4) of implementing a linear difference equation having ), a sample sequence ((U) for measurement time points k = 1, 2 to N with a sampling period T. G (k), U N (k), U M (k))) and network parameters (θ i Step (S5) of implementing a measurement-value equation system from N linear difference equations using ), estimated network parameters A step (S6) of calculating as an approximate solution of the measurement-value equation system, wherein the network parameter (θ i ) and the above estimated network parameters (θ i Step (S6), in which the sum of squared residuals between ) is minimized; Step (S7), in which the sum of squared residuals is minimized through QR decomposition of the measurement-value matrix (Ψ), wherein the QR decomposition is recursively calculated; and the corresponding insulation resistances (R1, R2) and the corresponding leakage capacitances (C1, C2) are the estimated network parameters Step of calculating from (S8), considering samples available at the current measurement point, correspondingly calculated and estimated network parameters A method characterized by including a step of continuously repeating the above method steps. Claim 2 A method according to claim 1, wherein the linear difference equation is derived by converting a linear algebraic equation in the frequency domain into a time-continuous difference equation and its time-discrete implementation, and wherein the linear algebraic equation describes a current-voltage relationship. Claim 3 A method according to claim 1 or 2, characterized in that the recursive QR decomposition is performed based on a recursive matrix. Claim 4 A method according to claim 1, characterized in that the QR decomposition is performed by Gibbons rotation. Claim 5 In claim 1, the upper triangle matrix (R) and the result vector (c) are pocketing factors in the recursive QR decomposition. A method characterized by being weighted using