A neural network-based power distribution system harmonic detection method and device

Abstract

The application discloses a power distribution system harmonic detection method and device based on a neural network, and belongs to the technical field of power quality disturbance detection. The method comprises the following steps: in step S1, voltage time domain signals are respectively acquired at R monitoring points of the power distribution system; in step S2, a neural network model outputs power quality disturbance event information according to the voltage time domain signals; in step S3, if the power quality disturbance event type is the power harmonic event, a current time domain signal is acquired by a power signal detection instrument; if it is other events, the power quality disturbance event information is displayed; in step S4, a probabilistic neural network outputs a harmonic component recognition result according to the current time domain signal and displays the result. The application can effectively reduce the complexity of power quality disturbance event judgment, and improve the accuracy of power harmonic detection of the power distribution system.

Description

Technical Field

[0001] This invention belongs to the field of power quality disturbance detection technology, and particularly relates to a method and device for harmonic detection in power distribution systems based on neural networks. Background Technology

[0002] Power quality disturbances refer to deviations or fluctuations in voltage, current, or frequency within a power system. These deviations or fluctuations can adversely affect the normal operation of power equipment, the efficiency of power utilization, and the stability of the power system. They typically include voltage spikes, voltage sags, and power harmonics. To ensure the stable operation of the power system and the normal use of power equipment, these power quality disturbances need to be monitored, analyzed, and addressed. Current technologies utilize power quality monitoring devices to monitor changes in voltage, current, and frequency in real time, enabling the timely detection and handling of potential power quality problems. Simultaneously, measures can be taken to prevent and reduce the occurrence of power quality disturbances, such as optimizing the power grid structure, improving equipment performance, and strengthening equipment maintenance.

[0003] Currently, methods for determining the type of power quality disturbance events can be divided into two categories based on signal type. The first category is time-domain waveform discrimination methods, which utilize artificial intelligence to analyze the time-domain waveform of acquired power signals to determine the load type and harmonic sources. However, when using artificial intelligence for analysis, a large amount of power quality disturbance event data is often required to train the AI ​​model to accurately obtain the characteristics of power quality disturbance events and identify the event type, resulting in high complexity. The second category is harmonic spectrum discrimination methods, which typically rely on Fast Fourier Transform (FFT) to convert the distorted waveform from the time domain to the frequency domain, and then analyze the spectrum to identify the disturbance event type. However, this method suffers from problems such as spectral leakage and picket fence effect when processing non-stationary signals, potentially leading to inaccurate analysis results. Furthermore, because the entire judgment process cannot be simplified, it is difficult to meet the real-time requirements for power quality disturbance event judgment. In addition, power equipment with nonlinear characteristics is increasingly being used in distribution systems, and the power harmonics they cause have become one of the main manifestations of power quality problems. Therefore, how to promptly identify power harmonic components after determining a power quality disturbance event, so as to accurately eliminate the impact of power harmonics on power quality, is also an urgent problem to be solved in this field.

[0004] Therefore, there is an urgent need for a new method and device for detecting harmonics in power distribution systems. This method and device can quickly detect power quality disturbances in power distribution systems and accurately identify power harmonic components based on the detection results, thereby enabling rapid response to power disturbances and improving the stability of power distribution system operation and user experience. Summary of the Invention

[0005] To address the shortcomings of the existing technology, this invention provides a method for harmonic detection in a power distribution system based on a neural network, the method comprising:

[0006] Step S1: Set up R monitoring points in the power distribution system, and set up power signal detection instruments at the R monitoring points to acquire voltage time domain signals respectively; where R ≥ 2;

[0007] Step S2: Preprocess the voltage time-domain signal, the preprocessed voltage time-domain signal carries the location information of the corresponding monitoring point; sequentially input the preprocessed voltage time-domain signal into the neural network model, the neural network model outputs power quality disturbance event information based on the voltage time-domain signal; the power quality disturbance event information includes the location information of the monitoring point corresponding to the voltage time-domain signal and the type of power quality disturbance event, the power quality disturbance event types include five types: normal, power harmonic event, voltage surge event, voltage sag event, and power outage event;

[0008] Step S3: If the power quality disturbance event type is the power harmonic event, determine the corresponding power signal detection instrument based on the monitoring point location information, and obtain the current time domain signal by the power signal detection instrument; if the power quality disturbance event type is other events, display the power quality disturbance event information.

[0009] Step S4: Preprocess the current time-domain signal, input the preprocessed current time-domain signal into the probabilistic neural network, the probabilistic neural network outputs the harmonic component identification result based on the current time-domain signal, and displays the harmonic component identification result.

[0010] Furthermore, the neural network model includes a wavelet layer, a hidden layer, and an output layer, wherein the wavelet layer is used to extract signal features from the voltage time-domain signal, and the hidden layer and the output layer are used to calculate the probability that the signal belongs to the power quality disturbance event type based on the signal features.

[0011] Furthermore, the weight matrix W from the wavelet layer to the hidden layer WH Specifically:

[0012]

[0013]

[0014]

[0015] x i =x(nT)

[0016] t i =xc (nT)

[0017] Where, d i Let x be the scale parameter of the wavelet transform, and let x be the voltage time-domain signal. c For the fundamental frequency time domain signal, x i Let t be the sampling sequence of the voltage time-domain signal. i Let T be the sampling sequence of the fundamental frequency time-domain signal, T be the sampling period, n be the number of wavelet nodes in the wavelet layer, and K be the number of nodes in the hidden layer.

[0018] Furthermore, the weight matrix W from the hidden layer to the output layer HO Specifically:

[0019]

[0020]

[0021] Where m is the number of output nodes in the output layer, m = 5, corresponding to the five types of power quality disturbance events mentioned above, and Category1 and Category2 represent category labels.

[0022] Furthermore, the calculation of the probability that the signal belongs to the power quality disturbance event type specifically involves:

[0023]

[0024]

[0025]

[0026] Where σ is the smoothing parameter.

[0027] Furthermore, the probabilistic neural network includes an input layer, a hidden layer, and an output layer, wherein the output layer includes 6 output nodes, corresponding to the 3rd, 5th, 7th, 9th, 11th, and 13th harmonics, respectively.

[0028] Furthermore, after step S4, the method further includes: providing an active filter at the corresponding monitoring point, and injecting a harmonic current with the same magnitude but opposite direction as the harmonic component into the active filter according to the harmonic component identification result output by the probabilistic neural network, so as to eliminate the harmonic components in the current time domain signal.

[0029] The present invention also provides a harmonic detection device for a power distribution system based on a neural network, the device comprising:

[0030] The acquisition module is used to set up R monitoring points in the power distribution system, and to set up power signal detection instruments at the R monitoring points to acquire voltage time-domain signals respectively; wherein R ≥ 2;

[0031] The detection module is used to preprocess the voltage time-domain signal, and the preprocessed voltage time-domain signal carries the location information of the corresponding monitoring point. The preprocessed voltage time-domain signal is then sequentially input into a neural network model, and the neural network model outputs power quality disturbance event information based on the voltage time-domain signal. The power quality disturbance event information includes the location information of the monitoring point corresponding to the voltage time-domain signal and the type of power quality disturbance event, which includes five types: normal, power harmonic event, voltage surge event, voltage sag event, and power outage event.

[0032] The judgment module is used to: if the power quality disturbance event type is the power harmonic event, determine the corresponding power signal detection instrument based on the monitoring point location information, and have the power signal detection instrument acquire the current time domain signal; if the power quality disturbance event type is other events, display the power quality disturbance event information.

[0033] The identification module is used to preprocess the current time-domain signal, input the preprocessed current time-domain signal into a probabilistic neural network, and output the harmonic component identification result based on the current time-domain signal and display the harmonic component identification result.

[0034] The present invention also provides a computer-readable storage medium storing a computer program that, when executed by a processor, implements the aforementioned neural network-based harmonic detection method for power distribution systems.

[0035] The present invention also provides an electronic device, the electronic device including a memory and a processor, the memory storing a computer program, and the processor executing the computer program to implement the aforementioned neural network-based harmonic detection method for power distribution systems.

[0036] This invention first utilizes a neural network to determine the type of power quality disturbance events at multiple monitoring points. Based on the determination that the power quality disturbance event is a power harmonic event, it then identifies the power harmonic components, enabling a comprehensive and accurate identification of power quality disturbance events and improving the effectiveness of power harmonic component detection. When using the neural network to determine the type of power quality disturbance event, wavelet transform and probabilistic neural networks are combined. The wavelet transform is used as the input layer of the probabilistic neural network, enabling this input layer to function like a filter. It can directly extract signal features from the input time-domain signal, avoiding problems such as spectral leakage and picket fence effects that may occur when using fast Fourier transform for time-frequency conversion beforehand. Furthermore, the aforementioned neural network has a fast learning speed and does not require iterative correction of the network weight matrix; only the smoothing parameters need to be considered, effectively reducing the complexity of power quality disturbance event judgment. When identifying power harmonic components, this invention sets the probabilistic neural network to detect odd harmonics of nonlinear loads, effectively improving the accuracy of power harmonic detection in the power distribution system and facilitating subsequent harmonic component elimination. Attached Figure Description

[0037] The above and other objects, features, and advantages of exemplary embodiments of the present disclosure will become readily apparent upon reading the following detailed description with reference to the accompanying drawings. In the drawings, several embodiments of the present disclosure are illustrated by way of example and not limitation, and like or corresponding reference numerals denote like or corresponding parts, wherein:

[0038] Figure 1 This is a flowchart illustrating a harmonic detection method for a power distribution system based on a neural network according to an embodiment of the present invention.

[0039] Figure 2 This illustrates a neural network according to an embodiment of the present invention.

[0040] Figure 3 This is a schematic diagram illustrating the structure of a neural network-based harmonic detection device for a power distribution system according to an embodiment of the present invention. Detailed Implementation

[0041] To make the objectives, technical solutions, and advantages of this invention clearer, the invention will be further described in detail below with reference to the accompanying drawings. Obviously, the described embodiments are merely some embodiments of this invention, and not all embodiments. All other embodiments obtained by those skilled in the art based on the embodiments of this invention without creative effort are within the scope of protection of this invention.

[0042] The terminology used in the embodiments of this invention is for the purpose of describing particular embodiments only and is not intended to limit the invention. The singular forms “a,” “the,” and “the” as used in the embodiments of this invention and the appended claims are also intended to include the plural forms, and “multiple” generally includes at least two unless the context clearly indicates otherwise.

[0043] It should be understood that although the terms first, second, third, etc., may be used to describe... in the embodiments of the present invention, these... should not be limited to these terms. These terms are only used to distinguish... For example, first... may also be referred to as second... without departing from the scope of the embodiments of the present invention, and similarly, second... may also be referred to as first...

[0044] It should be understood that the term "and / or" used in this article is merely a description of the relationship between related objects, indicating that three relationships can exist. For example, A and / or B can represent: A existing alone, A and B existing simultaneously, and B existing alone. Additionally, the character " / " in this article generally indicates that the preceding and following related objects have an "or" relationship.

[0045] Depending on the context, the words “if” or “suppose” as used here can be interpreted as “when” or “in response to determination” or “in response to detection.” Similarly, depending on the context, the phrases “if determination” or “if detection (of the stated condition or event)” can be interpreted as “when determination” or “in response to determination” or “when detection (of the stated condition or event)” or “in response to detection (of the stated condition or event).”

[0046] It should also be noted that the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that an article or device that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such an article or device. Without further limitation, an element defined by the phrase "comprising one..." does not exclude the presence of other identical elements in the article or device that includes said element.

[0047] Power quality disturbances mainly include events such as voltage surges, voltage dips, power harmonics, and power outages.

[0048] Specifically, a voltage surge refers to a sudden increase in the effective value of the voltage to more than 110% of the rated value within a short period of time. This phenomenon is usually caused by factors such as a faulty voltage regulator or a sudden decrease in the load on a damaged circuit.

[0049] Voltage sag: also known as voltage transient, voltage fluctuation, or voltage drop, refers to a brief decrease in voltage in a power system. It is usually caused by grid faults, sudden load changes, equipment failures, or other disturbances. The duration of a voltage sag can range from milliseconds to seconds, and its magnitude may also vary, but it generally manifests as a drop in the effective voltage value to a certain percentage of the rated value (e.g., between 10% and 90%).

[0050] Power harmonics refer to waveform distortions in a power system where the frequency is an integer multiple of the fundamental frequency (typically 50Hz or 60Hz). The generation of power harmonics mainly stems from various factors present in the power system, including nonlinear loads, power electronic devices, centralized isolators, and resonant circuits. Common nonlinear loads include UPS, switching power supplies, rectifiers, frequency converters, and inverters.

[0051] To achieve the detection of power quality disturbances and harmonics in the power distribution system, such as... Figure 1 As shown, this invention discloses a harmonic detection method for power distribution systems based on neural networks, the method comprising:

[0052] Step S1: Set up R monitoring points in the power distribution system, and set up power signal detection instruments at the R monitoring points to acquire voltage time domain signals respectively; where R ≥ 2.

[0053] Taking the IEEE 14-BUS node system as an example, multiple monitoring points are set up in the system based on the potential power quality disturbances. Each monitoring point represents a critical area or important node in the system, and the distribution of these points should cover the entire system as much as possible to comprehensively reflect the power quality status. In this embodiment, there are eight monitoring points. Power signal detection instruments are installed at these eight monitoring points to acquire voltage time-domain signals over a specific time period. Although existing technologies use power signal detection instruments to directly determine the type of power quality disturbance, their detection accuracy and comprehensiveness are unsatisfactory. Therefore, this invention uses power signal detection instruments to collect voltage time-domain signals and analyzes these signals using the neural network model proposed in this invention to improve the accuracy of disturbance type determination.

[0054] Step S2: Preprocess the voltage time-domain signal, the preprocessed voltage time-domain signal carries the location information of the corresponding monitoring point; sequentially input the preprocessed voltage time-domain signal into the neural network model, the neural network model outputs power quality disturbance event information based on the voltage time-domain signal.

[0055] The preprocessing of the voltage time-domain signal refers to filtering the voltage time-domain signal using a passive filter (such as an LC filter composed of inductors and capacitors) or an active filter (such as an operational amplifier-based filter) to remove additional noise. Since active filters can introduce additional amplification capabilities, they are more advantageous when processing complex signals; therefore, this embodiment uses an active filter to preprocess the voltage time-domain signal.

[0056] In step S2, the preprocessing of the voltage time-domain signal further includes combining the monitoring point location information corresponding to the voltage time-domain signal with the voltage time-domain signal to add the monitoring point location information to the voltage time-domain signal. This is because, in this embodiment, monitoring is performed on eight locations where power quality disturbances may occur. It is necessary to distinguish the monitoring results from different locations to accurately and quickly pinpoint the location of the power quality disturbance. Simultaneously, if the power quality disturbance is determined to be a power harmonic event, it facilitates the rapid activation of the corresponding power signal detection instrument to collect the voltage time-domain signal and identify the power harmonic components.

[0057] In this embodiment, either the monitoring point number or the monitoring point location coordinates can be used as the monitoring point location information. In one possible implementation, when using the monitoring point number as the monitoring point location information, adding the monitoring point location information to the voltage time-domain signal may take the following form:

[0058] Monitoring point number 1: Voltage time domain signal 1;

[0059] Monitoring point number 2: Voltage time domain signal 2;

[0060] ...

[0061] Monitoring point number 8: Voltage time domain signal 8.

[0062] In this embodiment, the power quality disturbance event information output by the neural network model includes the monitoring point location information corresponding to the voltage time-domain signal and the power quality disturbance event type. This monitoring point location information allows for rapid location of the power quality disturbance event and identification of its specific type.

[0063] In this embodiment, the power quality disturbance event types include five types: normal, power harmonic events, voltage surge events, voltage sag events, and power outage events. Power quality disturbance types in the art include steady-state disturbances and transient disturbances, each of which includes various different scenarios. Therefore, in one possible implementation, the power quality disturbance event types can be expanded based on common scenarios of power quality disturbance events in the power distribution system.

[0064] In step S2, as Figure 2 As shown, the neural network model is implemented by combining wavelet transform and probabilistic neural network. Figure 2 In the neural network model, the wavelet layer, hidden layer, and output layer are used to extract signal features from the voltage time-domain signal, and the hidden layer and the output layer are used to calculate the probability that the signal belongs to the power quality disturbance event type based on the signal features.

[0065] In the aforementioned neural network model, the wavelet layer functions as a filter, removing the fundamental frequency component from the input voltage time-domain signal. The signal features are then obtained through wavelet transform, effectively preserving the characteristics of power quality interference events. Furthermore, since wavelet transform has the ability to characterize local signal features in both the time and frequency domains, it is particularly beneficial for detecting transients and singularities in signals. Therefore, the wavelet layer can capture minute changes in the input voltage time-domain signal, improving the accuracy of identifying power quality interference events.

[0066] The hidden and output layers in the aforementioned neural network model are the same as those in probabilistic neural networks. Probabilistic neural networks learn quickly during training, requiring only a one-time setup and eliminating the need for iterative adjustments to the network weight matrix. They focus solely on adjusting the smoothing parameters, exhibiting strong adaptive and self-learning capabilities. They continuously optimize their parameters during training, improving classification accuracy. This allows them to maintain a high recognition rate even when faced with new types of perturbations.

[0067] In the neural network model described in this embodiment, the weight matrix W from the wavelet layer to the hidden layer... WH Specifically:

[0068]

[0069]

[0070]

[0071] x i =x(nT)

[0072] t i =x c (nT)

[0073] Where, d i Let x be the scale parameter of the wavelet transform, and let x be the voltage time-domain signal. c For the fundamental frequency time domain signal, x i Let t be the sampling sequence of the voltage time-domain signal. iLet T be the sampling sequence of the fundamental frequency time-domain signal, T be the sampling period, n be the number of wavelet nodes in the wavelet layer, and K be the number of nodes in the hidden layer.

[0074] The weight matrix W from the hidden layer to the output layer HO Specifically:

[0075]

[0076]

[0077] Where m is the number of output nodes in the output layer, m = 5, corresponding to the five types of power quality disturbance events mentioned above, and Category1 and Category2 represent category labels.

[0078] In the neural network model described in this embodiment, the probability of the input voltage time-domain signal belonging to the power quality disturbance event type is calculated as follows:

[0079]

[0080]

[0081]

[0082] Where σ is the smoothing parameter.

[0083] In this embodiment, when training the aforementioned neural network model, voltage time-domain signals of known power quality disturbance event types are used as training data, and the training set is divided into a training set and a test set. The known power quality disturbance event types include the five disturbance event types described in this embodiment. During training, the smoothing parameters are fine-tuned to improve model performance. The neural network model is then evaluated based on accuracy, recall, and F1 score to determine if its performance meets the requirements.

[0084] Step S3: If the power quality disturbance event type is the power harmonic event, determine the corresponding power signal detection instrument based on the monitoring point location information, and obtain the current time domain signal by the power signal detection instrument; if the power quality disturbance event type is other events, display the power quality disturbance event information.

[0085] In this embodiment, after obtaining the power quality disturbance event types output by the neural network model, processing is performed according to different power quality disturbance event types. Among the five power quality disturbance event types in this embodiment, power harmonic events belong to steady-state disturbances, voltage sags, voltage surges, and power outages belong to transient disturbances, and normal power quality does not require further processing. Therefore, in the case of a transient disturbance, i.e., a voltage sag, voltage surge, or power outage, the judgment result of the above power quality disturbance event type is output and the corresponding monitoring point location information is displayed, allowing the user to promptly handle the power quality disturbance event based on the power quality disturbance event type and monitoring point location information.

[0086] In the case of a steady-state disturbance, i.e. a power harmonic event, the power signal detection instrument corresponding to the monitoring point location information associated with the power harmonic event acquires the current time-domain signal for identification of the power harmonic components.

[0087] Step S4: Preprocess the current time-domain signal, input the preprocessed current time-domain signal into the probabilistic neural network, the probabilistic neural network outputs the harmonic component identification result based on the current time-domain signal, and displays the harmonic component identification result.

[0088] In this embodiment, the current time-domain signal is preprocessed, specifically by using a filter to reduce noise. Subsequently, the preprocessed current time-domain signal is input into a probabilistic neural network for harmonic component identification.

[0089] The inventors discovered through research that power conversion equipment, such as nonlinear loads like rectifiers, frequency converters, and inverters, is the primary source of harmonics in power grids. Power conversion equipment typically exhibits odd-order harmonic characteristics, with the most severe harmonic frequencies located at the 3rd, 5th, 7th, 9th, 11th, and 13th harmonics. Therefore, to accurately identify power harmonic components, the output layer of the probabilistic neural network described above was adjusted to have six output nodes, corresponding to the identification of the 3rd, 5th, 7th, 9th, 11th, and 13th harmonics, respectively. The input and hidden layers of the probabilistic neural network are conventional forms known in the art. By improving this probabilistic neural network, power harmonic components can be identified more accurately, enabling a rapid response to power harmonic events.

[0090] In this embodiment, the method of acquiring training data for the probabilistic neural network varies depending on the application scenario of the power distribution system. In one possible implementation, for systems containing arc lamps, or in scenarios where the load fluctuates drastically over time, such as high-speed rail power consumption, the current time-domain signal is acquired at a predetermined interval of 1 minute for 2 hours. For conventional power systems, the predetermined interval is 15 minutes for acquiring the current time-domain signal for 24 hours. By acquiring training data in a targeted manner, the accuracy of the probabilistic neural network in identifying harmonic components is further improved.

[0091] In this embodiment, the training process and application method of the probabilistic neural network are the same as those in the conventional art, and will not be described again here.

[0092] After identifying specific power harmonic components, the identification results are displayed to the user. In one possible implementation, the user can provide an active filter at the corresponding monitoring point. Based on the harmonic component identification results output by the probabilistic neural network, the active filter injects a harmonic current with the same magnitude but opposite direction as the harmonic component to eliminate the harmonic components in the current time-domain signal.

[0093] In another embodiment of the present invention, a harmonic detection device for a power distribution system based on a neural network is also provided, such as... Figure 3 As shown, the device 300 includes:

[0094] Acquisition module 301, wherein the acquisition module is used to set R monitoring points in the power distribution system, and to set power signal detection instruments at the R monitoring points to acquire voltage time domain signals respectively; wherein R≥2;

[0095] The detection module 302 is used to preprocess the voltage time-domain signal, the preprocessed voltage time-domain signal carrying the location information of the corresponding monitoring point; the preprocessed voltage time-domain signal is sequentially input into a neural network model, the neural network model outputs power quality disturbance event information based on the voltage time-domain signal; the power quality disturbance event information includes the location information of the monitoring point corresponding to the voltage time-domain signal and the type of power quality disturbance event, the power quality disturbance event types include five types: normal, power harmonic event, voltage surge event, voltage sag event, and power outage event;

[0096] The judgment module 303 is used to: if the power quality disturbance event type is the power harmonic event, determine the corresponding power signal detection instrument based on the monitoring point location information, and have the power signal detection instrument acquire the current time domain signal; if the power quality disturbance event type is other events, display the power quality disturbance event information.

[0097] The identification module 304 is used to preprocess the current time-domain signal, input the preprocessed current time-domain signal into a probabilistic neural network, and output the harmonic component identification result based on the current time-domain signal and display the harmonic component identification result.

[0098] It should be noted that the computer-readable medium described in this disclosure can be a computer-readable signal medium or a computer-readable storage medium, or any combination thereof. A computer-readable storage medium can be, for example,—but not limited to—an electrical, magnetic, optical, electromagnetic, infrared, or semiconductor system, apparatus, or device, or any combination thereof. More specific examples of a computer-readable storage medium may include, but are not limited to: an electrical connection having one or more wires, a portable computer disk, a hard disk, random access memory (RAM), read-only memory (ROM), erasable programmable read-only memory (EPROM or flash memory), optical fiber, portable compact disk read-only memory (CD-ROM), optical storage device, magnetic storage device, or any suitable combination thereof. In this disclosure, a computer-readable storage medium can be any tangible medium containing or storing a program that can be used by or in conjunction with an instruction execution system, apparatus, or device. In this disclosure, a computer-readable signal medium can include a data signal propagated in baseband or as part of a carrier wave, carrying computer-readable program code. Such propagated data signals can take various forms, including but not limited to electromagnetic signals, optical signals, or any suitable combination thereof. A computer-readable signal medium can be any computer-readable medium other than a computer-readable storage medium, which can send, propagate, or transmit a program for use by or in connection with an instruction execution system, apparatus, or device. The program code contained on the computer-readable medium can be transmitted using any suitable medium, including but not limited to: wires, optical fibers, RF (radio frequency), etc., or any suitable combination thereof.

[0099] The aforementioned computer-readable medium may be included in the aforementioned electronic device; or it may exist independently and not assembled into the electronic device.

[0100] Computer program code for performing the operations of this disclosure can be written in one or more programming languages ​​or a combination thereof, including object-oriented programming languages ​​such as Java, Smalltalk, and C++, and conventional procedural programming languages ​​such as the "C" language or similar programming languages. The program code can be executed entirely on the user's computer, partially on the user's computer, as a standalone software package, partially on the user's computer and partially on a remote computer, or entirely on a remote computer or server. In cases involving remote computers, the remote computer can be connected to the user's computer via any type of network—including a local area network (LAN) or a wide area network (WAN)—or can be connected to an external computer (e.g., via the Internet using an Internet service provider).

[0101] The flowcharts and block diagrams in the accompanying drawings illustrate the architecture, functionality, and operation of possible implementations of systems, methods, and computer program products according to various embodiments of this disclosure. In this regard, each block in a flowchart or block diagram may represent a module, segment, or portion of code containing one or more executable instructions for implementing a specified logical function. It should also be noted that in some alternative implementations, the functions indicated in the blocks may occur in a different order than those indicated in the drawings. For example, two consecutively indicated blocks may actually be executed substantially in parallel, and they may sometimes be executed in reverse order, depending on the functions involved. It should also be noted that each block in the block diagrams and / or flowcharts, and combinations of blocks in the block diagrams and / or flowcharts, may be implemented using a dedicated hardware-based system that performs the specified function or operation, or using a combination of dedicated hardware and computer instructions.

[0102] The units described in the embodiments of this disclosure can be implemented in software or in hardware. The names of the units are not, in some cases, intended to limit the specific unit.

[0103] The preferred embodiments of the present invention have been described above to make the spirit of the present invention clearer and easier to understand, and are not intended to limit the present invention. All modifications, substitutions, and improvements made within the spirit and principles of the present invention should be included within the protection scope summarized by the appended claims.

Claims

1. A method for harmonic detection in a power distribution system based on a neural network, characterized in that, The method includes: Step S1: Set up R monitoring points in the power distribution system, and set up power signal detection instruments at the R monitoring points to acquire voltage time domain signals respectively; where R ≥ 2; Step S2: Preprocess the voltage time-domain signal, the preprocessed voltage time-domain signal carrying the location information of the corresponding monitoring point; sequentially input the preprocessed voltage time-domain signal into the neural network model, the neural network model outputs power quality disturbance event information based on the voltage time-domain signal; the power quality disturbance event information includes the location information of the monitoring point corresponding to the voltage time-domain signal and the power quality disturbance event type, the power quality disturbance event type includes five types: normal, power harmonic event, voltage surge event, voltage sag event, and power outage event; the neural network model includes a wavelet layer, a hidden layer and an output layer, wherein the wavelet layer is used to extract signal features from the voltage time-domain signal, and the hidden layer and the output layer are used to calculate the probability that the signal belongs to the power quality disturbance event type based on the signal features; Step S3: If the power quality disturbance event type is the power harmonic event, determine the corresponding power signal detection instrument based on the monitoring point location information, and obtain the current time domain signal by the power signal detection instrument; if the power quality disturbance event type is other events, display the power quality disturbance event information. Step S4: Preprocess the current time-domain signal, input the preprocessed current time-domain signal into the probabilistic neural network, the probabilistic neural network outputs the harmonic component identification result based on the current time-domain signal, and displays the harmonic component identification result; The probabilistic neural network is used to detect odd harmonics of nonlinear loads. Wherein, the weight matrix W from the wavelet layer to the hidden layer WH Specifically: x i =x(nT) t i =x c (nT) Where, d i Let x be the scale parameter of the wavelet transform, and let x be the voltage time-domain signal. c For the fundamental frequency time domain signal, x i Let t be the sampling sequence of the voltage time-domain signal. i Let T be the sampling sequence of the fundamental frequency time-domain signal, T be the sampling period, n be the number of wavelet nodes in the wavelet layer, and K be the number of nodes in the hidden layer.

2. The method for harmonic detection in a power distribution system based on a neural network according to claim 1, characterized in that, The weight matrix W from the hidden layer to the output layer HO Specifically: Where m is the number of output nodes in the output layer, m = 5, corresponding to the five types of power quality disturbance events mentioned above, and Category1 and Category2 represent category labels.

3. The method for harmonic detection in a power distribution system based on a neural network according to claim 2, characterized in that, The calculation of the probability that the signal belongs to the power quality disturbance event type is specifically as follows: Where σ is the smoothing parameter.

4. The method for harmonic detection in a power distribution system based on a neural network according to claim 1, characterized in that, The probabilistic neural network includes an input layer, a hidden layer, and an output layer. The output layer includes six output nodes, which correspond to the 3rd, 5th, 7th, 9th, 11th, and 13th harmonics, respectively.

5. The method for harmonic detection in a power distribution system based on a neural network according to claim 1, characterized in that, After step S4, the method further includes: providing an active filter at the corresponding monitoring point, and injecting a harmonic current with the same magnitude but opposite direction as the harmonic component into the active filter according to the harmonic component identification result output by the probabilistic neural network, so as to eliminate the harmonic components in the current time domain signal.

6. A harmonic detection device for a power distribution system based on a neural network, characterized in that, The device includes: The acquisition module is used to set up R monitoring points in the power distribution system, and to set up power signal detection instruments at the R monitoring points to acquire voltage time-domain signals respectively; wherein R ≥ 2; The detection module preprocesses the voltage time-domain signal, which carries the location information of the corresponding monitoring point. The preprocessed voltage time-domain signal is then sequentially input into a neural network model, which outputs power quality disturbance event information based on the voltage time-domain signal. This power quality disturbance event information includes the location information of the monitoring point corresponding to the voltage time-domain signal and the type of power quality disturbance event, which includes five types: normal, power harmonic event, voltage surge event, voltage sag event, and power outage event. The neural network model includes a wavelet layer, a hidden layer, and an output layer. The wavelet layer extracts signal features from the voltage time-domain signal, and the hidden layer and the output layer calculate the probability that the signal belongs to the power quality disturbance event type based on the signal features. The judgment module is used to: if the power quality disturbance event type is the power harmonic event, determine the corresponding power signal detection instrument based on the monitoring point location information, and have the power signal detection instrument acquire the current time domain signal; if the power quality disturbance event type is other events, display the power quality disturbance event information. The identification module is used to preprocess the current time-domain signal, input the preprocessed current time-domain signal into a probabilistic neural network, and output harmonic component identification results based on the current time-domain signal and display the harmonic component identification results. The probabilistic neural network is used to detect odd harmonics of nonlinear loads. Wherein, the weight matrix W from the wavelet layer to the hidden layer WH Specifically: x i =x(nT) t i =x c (nT) Where, d i Let x be the scale parameter of the wavelet transform, and let x be the voltage time-domain signal. c For the fundamental frequency time domain signal, x i Let t be the sampling sequence of the voltage time-domain signal. i Let T be the sampling sequence of the fundamental frequency time-domain signal, T be the sampling period, n be the number of wavelet nodes in the wavelet layer, and K be the number of nodes in the hidden layer.

7. A computer-readable storage medium storing a computer program, characterized in that, When the computer program is executed by the processor, it implements the neural network-based harmonic detection method for power distribution systems as described in any one of claims 1 to 5.

8. An electronic device, the electronic device comprising a memory and a processor, characterized in that, The memory stores a computer program, and the processor executes the computer program to implement the neural network-based harmonic detection method for power distribution systems as described in any one of claims 1 to 5.

Images