Disclosure of Invention
The invention aims to solve the technical problem of providing a detection device for the ground fault of a direct current system, and the method can effectively finish the ground fault detection under the conditions that a ring network and a large ground capacitance exist in the system. Another technical problem to be solved by the present invention is to provide a method for detecting an earth fault of a dc system, which can effectively complete the earth fault detection in the presence of a ring network and a large capacitance to earth in the system.
In order to solve the technical problem, the direct current system ground fault detection device comprises a microprocessor and a plurality of current transformers, wherein the current transformers are respectively sleeved on different branches of a direct current system; the output ends of the current transformers are respectively connected with the input ends of the multi-way switches, and the output ends of the multi-way switches are connected with the input end of the sampling and holding circuit; the direct current system ground fault detection device is provided with a synchronization link, the input end of the synchronization link is connected with a positive bus and a negative bus of the direct current system, and the output end of the synchronization link is connected with the input end of the multi-way switch through an isolation link; the output end of the low-frequency signal source is connected with a positive bus and a negative bus of the direct current system and is used for injecting a low-frequency signal voltage source into the direct current system when the ground fault occurs; the detection bridge is connected with a positive bus and a negative bus of the direct current system and is used for monitoring the insulation condition of the direct current system; when the grounding fault is monitored, the grounding polarity is judged, information is sent to the microprocessor to start a low-frequency signal source, a stable sine-wave or cosine-wave low-frequency voltage signal is injected into the direct-current system, the frequency of the low-frequency signal is between 10 and 35Hz, and the amplitude of the low-frequency signal is between 20 and 30V.
The microprocessor is an ARM singlechip.
The detection bridge is a double asymmetric bridge.
The low-frequency signal source is a sine or cosine low-frequency voltage signal with the frequency of 30Hz and the amplitude of 30V.
In order to solve the technical problem, the method for detecting the ground fault of the direct current system comprises the following steps:
(1) Monitoring the insulation condition of the direct current system through the double asymmetric bridges, and monitoring the ground insulation condition of the positive bus and the negative bus in real time on line by the bridge device when the direct current system is normal;
(2) When the earth fault is monitored, judging the earth polarity, and sending information to a microprocessor to start a low-frequency signal source;
(3) Injecting a stable sine or cosine low-frequency voltage signal into the direct current system, wherein the frequency of the low-frequency signal is between 10 and 35HZ, and the amplitude of the low-frequency signal is between 20 and 30V;
(4) Detecting branch current signals through current transformers sleeved on the branches, and performing the looped network branch processing in the step (6) when the current of the current transformers is larger than 1A, or performing the general branch processing in the step (5);
(5) General branch processing:
sampling a low-frequency voltage signal loaded on a direct-current power grid bus, and calculating an amplitude value U and a phase phi according to a Fourier algorithm;
discretizing a branch current signal detected by a current transformer into an equally-spaced sampling sequence; sampling the branch current signal and the low-frequency voltage signal at the same time;
carrying out filtering pretreatment by adopting a filtering method based on 3-time B-spline wavelet transform to filter interference and higher harmonics;
extracting the amplitude and the phase of the low-frequency component from the complex branch current signal simultaneously by adopting a low-frequency component extraction method based on Morlet wavelet transform;
calculating the grounding resistance value of the branch according to the obtained amplitude and phase of the low-frequency voltage and the amplitude and phase of the low-frequency current of the branch, and comparing the grounding resistance value with a grounding judgment standard to judge whether the branch is grounded; detecting other branches in the same way;
(6) Processing a looped network branch:
selecting Morlet wavelet as wavelet base for continuous wavelet transform of current signal detected by current transformer, and selecting omega 0 Morlet wavelet of =10, and calculate corresponding yardstick, frequency window center and frequency support space of different frequency, limit different frequency components in certain bandwidth while carrying on the continuous wavelet transform;
constructing a three-dimensional figure of sampling frequency-scale-wavelet transform coefficients according to the advantages of the box dimension on the surface of the shape;
obtaining a fractal dimension value and a concave-convex parameter value on the surface of the graph, wherein the fractal dimension D and the concave-convex parameter value are functions of the side length L of the box;
after the L is selected, each L corresponds to an inverted V (L) and a D (L), and a fractal dimension-concavity and convexity point diagram is drawn on a two-dimensional plane;
and analyzing whether the fractal dimension-concave-convex point is in the fault area according to the fault area determined by the multiple test results so as to judge whether the looped network branch has the ground fault.
The low-frequency signal source is a sine or cosine low-frequency voltage signal with the frequency of 30Hz and the amplitude of 30V.
Compared with the prior art, the invention has the following beneficial effects: the invention can effectively complete the ground fault detection of the system under the conditions of looped network and large ground capacitance, is not influenced by complex and huge multi-branch power supply network, and improves the reliability of power supply of the direct current system; the invention adopts double asymmetric bridges to monitor the insulation condition of the direct current system, and solves the problem of identification of uniform reduction of the positive and negative electrode insulation.
Detailed Description
The present invention will be described in detail below with reference to examples and the accompanying drawings.
The embodiment of the ground fault detection device of the direct current system comprises the following steps:
as shown in fig. 1, the ground fault detection apparatus for a dc system of the present invention includes a microprocessor and a plurality of current transformers, which are respectively sleeved on different branches of the dc system. A plurality of current transformers in figure 1 are respectively represented by CT1, CT2, 8230, (8230) and CTn. The microprocessor is an advanced reduced instruction microprocessor, called ARM singlechip for short. The microprocessor is connected with a low-frequency signal source, a detection bridge, a display alarm and a sampling hold circuit. The output ends of the current transformers are respectively connected with the input ends of the multi-way switches, and the output ends of the multi-way switches are connected with the input end of the sampling and holding circuit. The direct current system earth fault detection device is provided with a synchronization link, the input end of the synchronization link is connected with a positive bus and a negative bus of the direct current system, and the output end of the synchronization link is connected with the input end of the multi-way switch through an isolation link. And the output end of the low-frequency signal source is connected with a positive bus and a negative bus of the direct current system and is used for injecting a low-frequency signal voltage source into the direct current system when the ground fault occurs. The detection bridge is a double-asymmetric bridge which is connected with a positive bus and a negative bus of the direct current system and used for monitoring the insulation condition of the direct current system. When the grounding fault is monitored, the grounding polarity is judged, information is sent to the microprocessor to start the low-frequency signal source, a stable sine or cosine low-frequency voltage signal is injected into the direct current system, the frequency of the low-frequency signal is between 10 and 35HZ, and the amplitude of the low-frequency signal is between 20 and 30V. In the embodiment, the low-frequency signal source adopts a sine or cosine low-frequency voltage signal with the frequency of 30Hz and the amplitude of 30V.
The direct current system ground fault detection device takes an ARM microprocessor S3C44B0X as a core, meets the requirements of processing wavelets and fractal algorithms in high performance and low power consumption, and is provided with a multi-channel 10-bit A/D converter. The current or voltage signal collected from the DC network is adjusted in phase and amplitude, isolated, sent to the multi-way selector switch, sent to the A/D converter through the sample holder, analyzed by the microprocessor, and displayed in the LED for the ground resistance value of each way and the grounding alarm prompt is found. The signal source adopts ICL8038 accurate waveform generator to produce sinusoidal signal, and the isolation link adopts ISO124 isolation amplifier to carry out the isolation with the device ground with the electric wire netting ground.
Generally, a detection device detects primary grid voltage within a specified time, the integral ground resistance of a grid is calculated through an unbalanced bridge, whether the grid is abnormal in insulation is judged, if the detected resistance value is smaller than a specified resistance value, the polarity of a grounding bus is judged, and meanwhile, low-frequency signals are injected into positive and negative buses. And then, carrying out routing inspection on the power grid branches, and judging the insulation condition of each branch so as to determine a fault point. The method comprises the steps that a branch current and low-frequency voltage are required to be detected simultaneously for one branch, the ARM microprocessor judges the magnitude of the branch current through periodic sampling and A/D conversion, if the magnitude of the branch current is larger than a set value, the branch is regarded as a looped network branch, processing is carried out through a wavelet algorithm and a fractal algorithm, and whether a fault exists is judged through analyzing the complexity of a current signal; if the current is smaller than the set value, the amplitude and the phase of the low-frequency current and voltage signals are obtained through a wavelet algorithm and a Fourier algorithm, the resistance of the branch is calculated, and the grounding condition is judged.
The embodiment of the method for detecting the ground fault of the direct current system comprises the following steps:
the invention monitors the insulation condition of the direct current system by double asymmetric bridges. When the direct current system is normal, the bridge device monitors the ground insulation condition of the positive bus and the negative bus in real time on line, once a point of ground fault is judged, the ground polarity is judged, information is sent out immediately to start a low-frequency signal source, and sine or cosine low-frequency voltage signals with the frequency being stable at 30Hz and the amplitude being constant at 30V are injected into the direct current system. And detecting branch current signals through a current transformer sleeved at the top end of each branch. The current transformer is required to have a proper transformation ratio and a proper grade to ensure the measurement accuracy. According to experience, the CT current is considered to be larger than 1A, is taken as a looped network branch (due to a large amount of harmonic circulating current components in a looped network), and enters a looped network branch processing part, otherwise, general branch processing is carried out. Fig. 2 is a detection flow diagram of the dc system ground fault detection method of the present invention.
1. General branch processing:
1. low-frequency voltage amplitude and initial phase calculation
Sampling the low-frequency voltage signal loaded on the direct-current power grid bus, and calculating the amplitude U and the phase phi according to a Fourier algorithm.
2. Branch current discretization
The branch current detected by the current transformer is a continuous signal, and is discretized into a sampling sequence with equal intervals. In order to calculate the resistance to ground of the branch, the branch current signal and the low-frequency voltage signal must be sampled at the same time. Fig. 3-1 is a non-ring network tributary original signal.
3. Branch current pre-processing
The branch current has complex components, contains various interference signals and needs to be filtered and preprocessed firstly. The filtering requirements are to filter out interference and higher harmonics as much as possible and to keep amplitude and phase information of low-frequency current undistorted. Therefore, a filtering method based on 3-order B-spline wavelet transform is adopted. Fig. 3-2 is a plot of the original signal after 3B-spline wavelet filters.
For example, the m-th order base B spline function is defined as
In the frequency domain there are:
when m =3, the corresponding filter coefficient is
Wherein h is a low-pass filter function; g is the high-pass filter coefficient.
The signal can be decomposed into components on different frequency bands using the Mallat tower algorithm as follows:
in the formula, c 0 And (n) is a sampling signal. The reconstruction algorithm is as follows:
wherein N is the number of decomposition layers.
When the branch current is sampled at the frequency of 1000Hz, the highest frequency reflected by the discrete sampling signal is 500Hz according to the sampling theorem. Wavelet decomposition is carried out twice by using a formula (1-4), reconstruction is carried out by using a formula (1-5), the reconstructed signal only contains 0-125 Hz frequency components, and high-frequency noise and high-order harmonic components in the original branch current signal are filtered.
4. Low-frequency current component amplitude and phase calculation
This is a critical step of the overall detection process. Whether the low-frequency current component is extracted accurately or not is directly related to the calculation precision of the grounding resistance. In order to simultaneously extract the amplitude and phase of the low-frequency component from the complex branch current signal, obtain higher precision and obtain higher precision, a low-frequency component extraction method based on Morlet wavelet transform is adopted here.
The Morlet wavelet is a single-frequency complex sine modulation Gaussian wave and is also the most commonly used complex-valued wavelet. The wavelet function time domain and frequency domain form is as follows:
Since the Morlet wavelet is a non-orthogonal wavelet and there is no conjugate mirror filter, it is not possible to use Mallat algorithm to perform binary wavelet decomposition and reconstruction on the signal. In practical applications, the wavelet transform of the discrete sequence s (nT) can be calculated from the riemann series:
for signal s (t) = A · cos (ω) f t + phi)), and the inverse transform formula of complex wavelet of discrete sequence is
Taking Morlet wavelet scale a =1, central frequency f 0 =30Hz. The Morlet wavelet transform algorithm and the Morlet wavelet inverse transform algorithm of the signals in a specific form can extract low-frequency current components from the branch currents, and then the amplitude and the phase of the low-frequency current are calculated through the Fourier algorithm. Fig. 3-3 is a plot after Morlet wavelet extraction.
5. Ground resistance calculation
And (4) calculating the grounding resistance value of the branch according to the amplitude and the phase of the low-frequency voltage obtained in the step (1) and the amplitude and the phase of the low-frequency current of the branch obtained in the step (4), comparing the grounding resistance value with a grounding judgment standard, judging whether the branch is grounded, and detecting other branches by the same method. The voltage of a direct current system is generally 110 or 220V, and the grounding judgment standard is 25K according to the power regulation.
2. Processing a looped network branch:
fractal science takes an irregular geometric form as a research object, and reveals the essential relation among laws, parts and the whole hidden behind the complex phenomenon through the disordered chaotic phenomenon and the irregular form of the separation of the capsule. Because the branch circuits in the direct current power grid which operate in the ring network mode are not independent, the branch circuits may change continuously with the change of the operation mode, harmonic ring current generated by the ring network exists in the branch circuits, and signal components are complex. The invention combines the natural fractal description of fractal theory, applies the fractal dimension theory, and distinguishes the signals from the aspect of signal complexity, thereby achieving the purpose of identifying the signals.
1. Ring network branch current signal preprocessing
The Morlet wavelet is selected as a wavelet base for carrying out continuous wavelet transformation on a current signal detected by a current transformer, and according to a low-frequency signal of 30Hz, the sampling current of the current transformer may be mixed with signals of 50Hz,100Hz,300Hz,600Hz and Gaussian noise. Selecting a specific wavelet scale: selecting omega 0 The Morlet wavelet of =10, and calculates the corresponding scale, frequency window center and frequency support space of different frequencies. Different frequency components can be limited in a certain bandwidth by using continuous wavelet transform, and the fractal can be favorably carried out in the next step. Fig. 4-1 shows original signals of a fault branch and a non-fault branch of a ring network.
2. Box dimension calculation for graphic surfaces
According to the advantages of the box dimension on the surface of the obtained body, a three-dimensional figure of sampling frequency-scale-wavelet transformation coefficient is constructed, so that the relation between signals is not cut, and the whole signal is always considered.
And calculating the fractal dimension of the obtained surface of the three-dimensional graph by adopting a box dimension method. Boxes with a side length of L are selected, and N (L) is the number of boxes. Estimating N (L) is a key step. Definition P (m, L) represents: probability of m points in a box with a side length of L. The box is centered at a point on the fractal surface. P (m, L) was normalized as follows:
n is the number of possible points in the box. Let S denote the points of the image (pixels of the image), if the image is covered by boxes with a side length L, then the number of boxes in which m points are:
thus, the total number of boxes desired to cover the entire image is:
and (3) respectively taking the side lengths L =3,5,7,9 and 11 of the box, amplifying the scale and the wavelet change coefficient to obtain box dimensions under different side lengths, and obtaining the fractal dimension of the graph through least square curve fitting. Fig. 4-2 is a 3D map generated by a continuous wavelet transform.
3. Obtaining a concave-convex degree parameter value of the surface of the graph;
the roughness parameter inverted V is introduced to determine the roughness of the surface of the image, and the characteristic of the roughness parameter inverted V describes the speed of the mass change of the fractal body. Similar to the method of measuring the length of the coastline, the fractal set depends on the length of the measurement standard and follows the following energy law:
M(L)=kL D (1-12)
m (L) is a function of L, the calculation ^ (L) is based on P (M, L); both the fractal dimension D and the relief a are functions of the side length L of the box.
4. Drawing dimension-relief pattern
After the L is selected, each L corresponds to an ^ (L) and a D (L), and a fractal dimension-concave-convex degree point diagram is drawn on a two-dimensional plane. And analyzing whether the fractal dimension-concave-convex point is in the fault area according to the fault area determined by the multiple test results so as to judge whether the looped network branch has the ground fault. Fig. 4-3 fractal dimension-relief plot.