CN112083299B - DC system insulation fault prediction method based on Kalman filtering - Google Patents

Abstract

A DC system insulation fault prediction method based on Kalman filtering includes the steps of firstly, obtaining insulation resistance value of a DC bus to ground, and establishing a database of insulation resistance of the bus to ground; calculating the average insulation resistance of the current database

Average insulation resistance

Average insulation resistance from the next moment

Resistance difference DeltaR of (2) _i I.e.1, N); calculating the resistance difference DeltaR _i With corresponding detection interval time DeltaT _i Ratio X of (2) _i The method comprises the steps of carrying out a first treatment on the surface of the Analyzing the variation trend of the ratio coefficient by using Kalman filtering, and predicting the ratio coefficient X at the next moment; calculating an insulation resistance increase value delta r=x×t at the next moment; calculating to obtain a predicted insulation resistance value R=R at the next moment _n +Δr; wherein: r is R _n In order to predict the actual insulation resistance value of the insulation resistance value R at the previous time, T is the interval time between the predicted insulation resistance value R and the previous time. According to the invention, the insulation resistance at the next moment can be predicted according to the historical data of the insulation resistance, so that the system fault point can be pre-positioned, and the positioning precision and efficiency can be improved.

Description

A DC system insulation fault prediction method based on Kalman filtering

Technical Field

The invention relates to the technical field of power system detection, and in particular to a DC system insulation fault prediction method based on Kalman filtering.

Background Art

The DC system is an important part of the power system. It is widely used in power plants, substations and other places, and has a huge branch network. It is used to supply DC loads such as relay protection, control, signal, computer monitoring, emergency lighting, AC uninterruptible power supply, etc. If the insulation state of the DC system changes drastically, it may cause the relay protection device to refuse to operate or malfunction, which will have a great negative impact on the safe and stable operation of the power grid. Therefore, it is necessary to test the insulation resistance of the DC system.

In recent years, the following schemes are mainly used for the detection of insulation resistance of DC systems: "Design of online insulation monitoring device based on STM32" published in the journal "Computer Measurement and Control" in 2019, uses the method of unbalanced bridge and balanced bridge to measure insulation resistance and locate fault points, but the error of this method is large and the operation is cumbersome. "Study on DC system fault detection based on improved unbalanced bridge" published in the journal "Electrical Automation" in 2018 improves the balanced bridge method so that the fault can be effectively and accurately detected, but the fault point cannot be accurately located, and it cannot be effectively detected under strong interference conditions such as substations. "220kV substation DC grounding fault analysis and search measures" published in the journal "Automation Application" in 2015, and "DC system insulation monitoring technology based on dynamic difference method" published in the journal "Proceedings of the Chinese Society of Electrical Engineering" in 2015, on the basis of the leakage current detection method, use the dynamic difference method to detect insulation resistance. This method requires the detection of leakage current and obtains the leakage current change, but this value is small and difficult to measure accurately.

Summary of the invention

In view of the above-mentioned deficiencies in the prior art, the technical problem to be solved by the present invention is: how to provide a DC system insulation fault prediction method that can predict the insulation resistance at the next moment based on the historical data of the insulation resistance, thereby enabling the pre-location of the system fault point, which is beneficial to improving the positioning accuracy and efficiency.

In order to solve the above technical problems, the present invention adopts the following technical solutions:

A DC system insulation fault prediction method based on Kalman filtering, characterized by comprising the following steps:

S1. Obtain the insulation resistance value of the DC busbar to the ground and establish a database of the insulation resistance of the busbar to the ground;

S2. After obtaining the insulation resistance value of the DC bus to the ground each time, calculate the average insulation resistance of the current database

与后一时刻的平均绝缘电阻

的电阻差值ΔR_i,i∈[1,N)；S3. Calculate the average insulation resistance at the previous moment

The average insulation resistance at the next moment

The resistance difference ΔR _i ,i∈[1,N);

S4, calculating the ratio of the resistance difference ΔR _i and the corresponding detection interval time ΔT _i as the ratio coefficient _Xi ;

S5. Analyze the changing trend of the ratio coefficient by using Kalman filtering, and predict the ratio coefficient X at the next moment;

S6. Calculate the insulation resistance increase ΔR at the next moment:

ΔR＝X×T

S7. Calculate the predicted insulation resistance value R at the next moment:

R＝ _Rn +ΔR

Where: _Rn is the actual insulation resistance value at the moment before the predicted insulation resistance value R, and T is the interval time between the predicted insulation resistance value R and the previous moment.

Furthermore, in step S1, a branch insulation resistance detection unit with the following structure is used to obtain the insulation resistance value of the DC bus to ground, including an instrument amplifier, detection resistors R3 and R4 arranged in series between the positive and negative poles of the DC bus, and a grounding resistor R5, one end of the grounding resistor R5 is grounded, and the other end is connected between the detection resistors R3 and R4; the positive power supply terminal of the instrument amplifier is connected to the positive pole of the DC bus through a first isolated power supply, and the negative power supply terminal of the instrument amplifier is connected to the negative pole of the DC bus through a second isolated power supply; the in-phase input terminal of the instrument amplifier is grounded, and the inverting input terminal is connected between the detection resistors R3 and R4; the instrument amplifier also has an external gain control resistor R6.

As an optimization, the instrument amplifier is an AD620 instrument amplifier.

As an optimization, the first isolated power supply and the second isolated power supply are both flyback power supplies.

Furthermore, the output end of the instrument amplifier is connected to a single-chip microcomputer, and a display screen is provided on the single-chip microcomputer.

Furthermore, the single chip microcomputer is also provided with a wireless communication module.

As an optimization, the wireless communication module is a Bluetooth module.

Furthermore, the single chip computer is also provided with a storage module for storing the resistance value of the detection resistor.

Furthermore, the state equation and observation equation predicted by the Kalman filter analysis are as follows:

_Xk+1 ＝A* _Xk +B* _Uk+1 + _Wk+1

Z _k+1 =H*X _k+1 +V _k+1

Where: _Xk is the insulation resistance growth rate of the kth detection; _Xk+1 is the predicted insulation resistance growth rate of the next detection; Uk ₊₁ is the control input; Zk ₊₁ is the observation of the state matrix _Xk+1 ; Wk ₊₁ is the system noise; _Vk+1 is the observation noise; A, B, H are parameter matrices;

Among them, the system noise and observation noise satisfy the following formula:

E[W _k (W _k ) ^T ]＝Q,E[V _k (V _k ) ^T ]＝Y

Where: _Wk is the system noise; _Vk is the observation noise; Q is the covariance matrix of the system noise; Y is the covariance matrix of the observation noise;

The state prediction equation of the Kalman filter prediction model is:

The state update equation of the Kalman filter prediction model is:

为已知Z_k+1前的状态预测值；

为已知Z_k+1后的最优估计值；K为卡尔曼增益矩阵，

表示真实值与预测值之间的协方差；P_k+1表示真实值与最优估计值之间的协方差；并满足下式：Where:

is the predicted value of the state before Z _k+1 is known;

is the optimal estimate after Z _k+1 is known; K is the Kalman gain matrix,

represents the covariance between the true value and the predicted value; P _k+1 represents the covariance between the true value and the optimal estimated value; and satisfies the following formula:

表示先验状态误差；e_k+1表示后验状态误差。Where:

represents the prior state error; e _k+1 represents the posterior state error.

In summary, the present invention can predict the insulation resistance at the next moment based on the historical data of the insulation resistance, thereby enabling pre-location of the system fault point, which is beneficial to improving the positioning accuracy and efficiency, etc.

BRIEF DESCRIPTION OF THE DRAWINGS

FIG1 is a system block diagram of a branch insulation resistance detection unit.

FIG. 2 and FIG. 3 are equivalent circuit diagrams of the bridge circuit in FIG. 1 .

FIG. 4 is a flowchart of system prediction analysis in an embodiment.

FIG5 is a simulation diagram of insulation resistance measurement.

Figure 6 is a diagram of the output voltage when the positive and negative insulation resistances are the same.

FIG7 is a graph showing the output voltage when the positive insulation resistance decreases and the negative insulation resistance remains unchanged.

FIG8 is a graph showing the output voltage when the negative insulation resistance decreases and the positive insulation resistance remains unchanged.

FIG9 is a graph showing the variation trend of the insulation resistance value over time.

Figure 10 is a measured picture.

DETAILED DESCRIPTION

This embodiment adopts a dynamic differential response method to detect the insulation resistance of the DC system online. By picking up the slightly changing voltage signal in the detection circuit, the insulation resistance value of the DC system is calculated and updated in real time, the resistance change curve is drawn, and its change trend function is fitted. The change trend function is used to estimate the predicted insulation resistance alarm threshold generation time, and the function is corrected in combination with the periodic feeder system leakage current detection. The method uses a high common-mode input voltage meter amplifier as an analog front end to monitor the insulation resistance, realizes the isolation of the busbar and the measurement unit, and the weak signal pickup generated by the slight change of the insulation resistance under high temperature and electromagnetic interference environment. And based on the Kalman filter analysis, the time of occurrence of the insulation fault and the change trend of the insulation resistance value are predicted, solving the disadvantages of low measurement accuracy, slow response, and inability to monitor online in real time in the feeder insulation resistance measurement circuit in the existing DC feeder system. Finally, the correctness and feasibility of the method were verified by using MATLAB simulation analysis and on-site physical simulation test.

The online detection of the dynamic differential response of the insulation resistance of the DC system in this embodiment specifically includes two parts, among which the first part is to use an instrument amplifier as an analog front-end detection unit, and to use the high-gain, high common-mode rejection ratio, and high common-rail AD620 to isolate the DC system and amplify the dynamic differential mode signal to measure the insulation resistance of the positive and negative ends of the load branch to the ground, thereby eliminating the disadvantage that the traditional method of measuring the insulation resistance needs to be switched, and realizing real-time monitoring of the insulation resistance. The second part is the insulation fault prediction, which uses Kalman filtering to model the insulation resistance data measured by the differential dynamic response, predict its change trend, and draw a change curve, and then before the predicted alarm threshold arrives at the time node, the branch leakage current is manually detected and the curve is corrected to reduce the impact of the insulation resistance change at the same time on the measurement.

Part I:

As shown in Figure 1, the branch insulation resistance detection unit includes an instrument amplifier, detection resistors R3 and R4 arranged in series between the positive and negative poles of the DC bus, and a grounding resistor R5, one end of the grounding resistor R5 is grounded, and the other end is connected between the detection resistors R3 and R4; the positive power supply end of the instrument amplifier is connected to the positive pole of the DC bus through a first isolated power supply, and the negative power supply end of the instrument amplifier is connected to the negative pole of the DC bus through a second isolated power supply; the in-phase input end of the instrument amplifier is grounded, and the inverting input end is connected between the detection resistors R3 and R4; the instrument amplifier also has an external gain control resistor R6. In this embodiment, the instrument amplifier is an AD620 instrument amplifier; the first isolated power supply and the second isolated power supply are both flyback power supplies; the output end of the instrument amplifier is connected to a single-chip microcomputer, and the single-chip microcomputer is provided with a display screen, a Bluetooth communication module and a storage module for storing the resistance value of the detection resistor.

The detection unit measures the insulation resistance of the two DC buses in the branch at the same time. In the figure, R represents the branch load, _R1 and _R2 represent the insulation resistance of the positive and negative DC buses to the ground, respectively, U is the voltage between the positive and negative buses, and _R3 and _R4 are detection resistors. The detection device takes power from the bus isolation through a flyback power supply, and amplifies the voltage across the resistor _R5 through the instrument amplifier AD620 to calculate the insulation resistance of the load branch to the ground.

Under natural conditions, the insulation resistance of the load branch to the ground is irreversibly reduced gradually. Even under the same conditions, the insulation resistance of the positive and negative busbars to the ground will not change in the same way in a very short time. In this embodiment, the output voltage of the balanced bridge in an unbalanced state is used to detect the insulation resistance.

When the DC system is working normally, the detection device (i.e., detection unit) is hung on the load branch, and the resistance of the load branch to ground and the voltage _U1 of the negative feedback terminal of the instrument amplifier to the negative bus are measured. The measuring resistors _R3 and _R4 are adjusted by the balanced bridge principle to make the voltage across _R5 zero, and the whole measuring device is dynamically balanced.

When the insulation resistance of the positive bus to the ground decreases, the voltage at the positive feedback terminal of the instrument amplifier AD620 becomes U ₂ to the negative bus voltage. The input voltage of the instrument amplifier is ΔU, which is amplified G times by AD620 and outputs U ₀ . According to the Thevenin theorem, the bridge circuit shown in Figure 1 can be equivalent to the two-port network shown in Figure 2. Short-circuit the power supply to obtain the circuit of Figure 3, where U _o is the equivalent power supply and R _i is the equivalent resistance.

Based on this, the equivalent internal resistance of the bridge can be calculated:

According to the circuit in Figure 2, the output voltage when the bridge is connected to a load is obtained:

The instrument amplifier calculates the output voltage U ₀ to obtain R ₁ ′, which can be transmitted to the host computer through the Bluetooth communication module and stored in the SD card of the host computer to correct the alarm threshold time in real time.

When the insulation resistance of the negative bus to the ground decreases, U ₂ may be equal to U ₁ , that is, ΔU is zero. At this time, it is considered that the insulation resistance of the negative bus to the ground is reduced, which simplifies the process of calculating the negative bus to the ground. At the same time, compared with the existing method of detecting insulation resistance, this solution does not need to repeatedly switch on and off the resistance, and can measure the insulation resistance value of the DC system in real time, eliminating the measurement error of the traditional balanced bridge for the same change in the insulation resistance of the positive and negative bus to the ground within a period of time. At the same time, the historical insulation resistance value is stored in the SD card, and the data is fitted to the equation and curve of the change of insulation resistance over time. Combined with the method of artificial feeder leakage current detection during the inspection period, the equation is corrected, and the time when the insulation resistance reaches the threshold can be predicted more accurately.

Part II:

Kalman filter prediction and analysis of insulation faults

In order to predict the time of the alarm threshold so as to carry out manual detection and correction in advance for the branch, the concepts of state variables and state space are introduced on the basis of Kalman filtering theory, and the observation equation and state equation are established. In the measurement that completely contains noise, the estimated value at the previous moment and the observed value at this moment are used to update the state variable. According to the linear unbiased minimum mean square error estimation criterion, a recursive algorithm is used to make the best estimate of the state variable of the filter, thereby obtaining the best estimate of the useful signal after filtering out the noise.

The steps to use the Kalman filter algorithm to predict the insulation resistance value change trend are as follows:

1) Obtain the insulation resistance value of the DC busbar to the ground, and establish a database of the insulation resistance of the busbar to the ground. In the database, the system collects the insulation resistance value N times, and collects M data each time;

2) Calculate the average insulation resistance of each acquisition

Where: _Rj is the actual value of the insulation resistance corresponding to each data set collected, and i and j are both positive integers;

与下一次绝缘电阻平均值

的差值ΔR_i,i∈[1,N)，得到电阻差值序列ΔR₁、ΔR₂……ΔR_N；3) Calculate the current average insulation resistance

The next insulation resistance average

The difference ΔR _i , i∈[1,N) is obtained to obtain the resistance difference sequence ΔR ₁ , ΔR ₂ …… ΔR _N ;

4) Calculate the ratio of the difference ΔR _i and the corresponding detection interval time ΔT _i to obtain the ratio coefficient _Xi , and form a ratio coefficient sequence: _Xi , _X2 ... _XN ;

5) Use Kalman filtering to analyze the changing trend of the ratio coefficient and predict the ratio coefficient X at the next moment;

6) Calculate the insulation resistance increase ΔR at the next moment:

ΔR＝X×T (7)

7) Calculate the predicted insulation resistance value R at the next moment:

R＝ _Rn +ΔR (8)

Where: _Rn is the actual insulation resistance value at the moment before the predicted insulation resistance value R, and T is the interval time between the predicted insulation resistance value R and the previous moment.

According to the insulation resistance value information library detected by the dynamic differential response, a branch insulation detection model is established. The data detected by the dynamic differential response is processed by Kalman filtering to correct the error caused by the measurement. By picking up the slightly changing voltage signal in the detection circuit, based on the parameters of the established model, the insulation resistance value of the DC system is calculated and updated in real time, the resistance change curve is drawn, and its change trend function is fitted. Finally, the predicted insulation resistance alarm threshold generation time is estimated, and the function is corrected in combination with the periodic feeder system leakage current detection. When the predicted data generates abnormal data in a certain period of time, the dynamic differential response detection frequency is accelerated, and the current data is corrected. After the error correction is completed, if the insulation coefficient is abnormal, the fault signal is immediately sent to the processor to disconnect the fault branch, so as to achieve the pre-positioning of the system fault point and estimate its impact, so as to make the DC system grounding line positioning more efficient and accurate. The system prediction analysis process is shown in Figure 4.

The premise of Kalman filtering is that both system noise and measurement noise obey zero-mean Gaussian distribution, and each noise component is independent of each other, so the noise covariance matrix is a diagonal matrix. Therefore, the following assumptions are made for system noise and observation noise:

E[W _k (W _k ) ^T ]＝Q,E[V _k (V _k ) ^T ]＝Y (10)

Where: _Wk is the system noise; _Vk is the observation noise; Q is the covariance matrix of the system noise; Y is the covariance matrix of the observation noise. By processing the noisy observation signal, the estimated value of the real signal when the error is minimized can be obtained in an average sense.

If the insulation resistance growth rates of several consecutive times are used as a series to predict the next insulation resistance growth rate, the following state equation and observation equation can be established:

X _k+1 ＝A*X _k +B*U _k+1 +W _k+1 (11)

Z _k+1 =H*X _k+1 +V _k+1 (12)

Where: _Xk is the insulation resistance growth rate of the kth detection; Xk ₊₁ is the predicted insulation resistance growth rate of the next detection; Uk ₊₁ is the control input (generally zero); Zk ₊₁ is the observation value of the state matrix _Xk+1 ; A, B, and H are parameter matrices.

From the state prediction equation and state update equation in the Kalman filter prediction model, we can get:

为已知Z_k+1前的状态预测值；

为已知Z_k+1后的最优估计值；K为卡尔曼增益矩阵，卡尔曼增益矩阵K为最优估计值与真实值之间的协方差偏导等于0时的系数，用以融合测量值与最优估计值。Where:

is the predicted value of the state before Z _k+1 is known;

is the optimal estimate after Z _k+1 is known; K is the Kalman gain matrix. The Kalman gain matrix K is the coefficient when the partial derivative of the covariance between the optimal estimate and the true value is equal to 0, which is used to fuse the measured value and the optimal estimate.

make:

其中：

表示先验状态误差；e_k+1表示后验状态误差；

表示真实值与预测值之间的协方差；P_k+1表示真实值与最优估计值之间的协方差；P矩阵用于度量最优估计值的精确程度。

in:

represents the prior state error; e _k+1 represents the posterior state error;

represents the covariance between the true value and the predicted value; P _k+1 represents the covariance between the true value and the optimal estimate; the P matrix is used to measure the accuracy of the optimal estimate.

From formula (11) to (15), we can get:

Calculate the Kalman gain matrix K under the optimal estimation condition:

The above equations (13) and (16) are state prediction equations, and equations (14) (17) and (18) are state update equations. Equations (13) to (18) constitute the Kalman filter recursive equation, which is used to predict the optimal estimated value of the insulation resistance growth rate at each moment.

The model proposed in this embodiment introduces external comprehensive influencing factors that affect the insulation condition, selects system insulation related information, expands the sample data volume, and then uses the Kalman filter method to predict the system insulation resistance value, and then obtains the insulation level trend. It fully considers the dynamic adaptability of external factors that affect the changes in the insulation condition of the DC system, provides accurate prediction data, and improves the accuracy of insulation detection. At the same time, it achieves more efficient DC system grounding line positioning based on judging the trend of changes in the insulation level of the DC system.

Simulation and Experimental Results Analysis

Using MATLAB/Simulink software, a simulation model as shown in Figure 5 is established.

By artificially changing the positive and negative insulation resistances to simulate the change of insulation resistance in the DC system, and detecting and recording the corresponding voltage values, the instantaneous relationship between insulation resistance and voltage is obtained. The voltage between the positive and negative busbars in the DC system is set to 220V. When _R1 , _R2 , _R3 , _R4 , are 10MΩ, _R5 is 1KΩ, and the external gain resistor _R6 of the instrument amplifier is 499Ω, the output voltage U0 is ₀ , as shown in Figure 6.

_R1 gradually changes from 10MΩ to 2MΩ, that is, the simulated positive insulation resistance decreases, _R2 , _R3 , and _R4 remain unchanged, _R5 is 1KΩ, and the external gain resistor _R6 of the instrument amplifier is 499Ω. The output voltage trend diagram is shown in Figure 7.

_R1 is 2MΩ, _R2 gradually changes from 10MΩ to 2MΩ, that is, the simulated negative insulation resistance decreases, _R3 and _R4 are 10MΩ, _R5 is 1KΩ, and the external gain resistor _R6 of the instrument amplifier is 499Ω, the output voltage trend diagram is shown in Figure 8.

By picking up the slightly changing voltage signal in the detection circuit, a graph showing the insulation resistance changing over time as shown in FIG9 is drawn after processing, and a horizontal comparison is made with the data detected by the traditional existing detection method. The data measured by this method is consistent with the data measured by the traditional method.

The physical device was measured and compared with the traditional measurement method at a local substation. When the insulation level was normal, the results measured by the traditional measurement method were shown in Figure 10 (the time in the figure was displayed incorrectly because the debugging equipment did not correct the system time). The voltage of a section of the positive bus was 253.2V, the insulation to the ground was abnormal, and the insulation resistance to the ground was 100Ω. The insulation resistance measured by this device was 100Ω, which verified its feasibility and accuracy.

This method is optimized on the basis of the DC leakage current detection method. It overcomes the disadvantage of low measurement sensitivity caused by the small leakage current value. It uses the Kalman filter algorithm to predict the change trend of the insulation resistance, can judge the insulation fault in real time, and more accurately estimate the time when the insulation resistance reaches the alarm threshold. It solves the disadvantages of low measurement accuracy, slow response, and inability to conduct online real-time monitoring of the feeder insulation resistance measurement circuit in the existing DC feeder system.

The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions and improvements made within the spirit and principles of the present invention should be included in the protection scope of the present invention.

Claims (8)

1. A DC system insulation fault prediction method based on Kalman filtering, characterized in that it comprises the following steps: S1. Obtain the insulation resistance value of the DC busbar to the ground and establish a database of the insulation resistance of the busbar to the ground; S2. After obtaining the insulation resistance value of the DC bus to the ground each time, calculate the average insulation resistance of the current database

S3. Calculate the average insulation resistance at the previous moment

The average insulation resistance at the next moment

The resistance difference ΔR _i ,i∈[1,N); S4, calculating the ratio of the resistance difference ΔR _i and the corresponding detection interval time ΔT _i as the ratio coefficient _Xi ; S5. Analyze the changing trend of the ratio coefficient by using Kalman filtering, and predict the ratio coefficient X at the next moment; S6. Calculate the insulation resistance increase ΔR at the next moment: ΔR＝X×T S7. Calculate the predicted insulation resistance value R at the next moment: R＝ _Rn +ΔR Where: _Rn is the actual insulation resistance value at the moment before the predicted insulation resistance value R, T is the interval time between the predicted insulation resistance value R and the previous moment; In the step S1, a branch insulation resistance detection unit with the following structure is used to obtain the insulation resistance value of the DC bus to the ground, including an instrument amplifier, detection resistors R3 and R4 arranged in series between the positive and negative electrodes of the DC bus, and a grounding resistor R5, one end of the grounding resistor R5 is grounded, and the other end is connected between the detection resistors R3 and R4; the positive power supply terminal of the instrument amplifier is connected to the positive electrode of the DC bus through a first isolated power supply, and the negative power supply terminal of the instrument amplifier is connected to the negative electrode of the DC bus through a second isolated power supply; the in-phase input terminal of the instrument amplifier is grounded, and the inverting input terminal is connected between the detection resistors R3 and R4; the instrument amplifier also has an external gain control resistor R6. 2. The DC system insulation fault prediction method based on Kalman filtering as described in claim 1 is characterized in that the instrument amplifier is an AD620 instrument amplifier. 3. The DC system insulation fault prediction method based on Kalman filtering as described in claim 1 is characterized in that both the first isolated power supply and the second isolated power supply are flyback power supplies. 4. The DC system insulation fault prediction method based on Kalman filtering as described in claim 1 is characterized in that the output end of the instrument amplifier is connected to a single-chip microcomputer, and a display screen is provided on the single-chip microcomputer. 5. The DC system insulation fault prediction method based on Kalman filtering as claimed in claim 4, characterized in that a wireless communication module is also provided on the single chip microcomputer. 6 . The DC system insulation fault prediction method based on Kalman filtering according to claim 5 , wherein the wireless communication module is a Bluetooth module. 7. The DC system insulation fault prediction method based on Kalman filtering according to claim 4, characterized in that the single chip microcomputer is also provided with a storage module for storing the resistance value of the detection resistor. 8. The DC system insulation fault prediction method based on Kalman filtering according to claim 1, characterized in that the state equation and observation equation analyzed and predicted by the Kalman filter are as follows: _Xk+1 ＝A* _Xk +B* _Uk+1 + _Wk+1 Z _k+1 =H*X _k+1 +V _k+1 Where: _Xk is the insulation resistance growth rate of the kth detection; _Xk+1 is the predicted insulation resistance growth rate of the next detection; Uk ₊₁ is the control input; Zk ₊₁ is the observation of the state matrix _Xk+1 ; Wk ₊₁ is the system noise; _Vk+1 is the observation noise; A, B, H are parameter matrices; Among them, the system noise and observation noise satisfy the following formula:

E[W _k (W _k ) ^T ]＝Q,E[V _k (V _k ) ^T ]＝Y Where: _Wk is the system noise; _Vk is the observation noise; Q is the covariance matrix of the system noise; Y is the covariance matrix of the observation noise; The state prediction equation of the Kalman filter prediction model is:

The state update equation of the Kalman filter prediction model is:

Where:

is the predicted value of the state before Z _k+1 is known;

is the optimal estimate after Z _k+1 is known; K is the Kalman gain matrix,

represents the covariance between the true value and the predicted value; P _k+1 represents the covariance between the true value and the optimal estimated value; and satisfies the following formula:

Where:

represents the prior state error; e _k+1 represents the posterior state error.

Images