CN114330461B - V-I track generation method and device for non-invasive load identification and neural network - Google Patents
V-I track generation method and device for non-invasive load identification and neural network Download PDFInfo
- Publication number
- CN114330461B CN114330461B CN202210255087.1A CN202210255087A CN114330461B CN 114330461 B CN114330461 B CN 114330461B CN 202210255087 A CN202210255087 A CN 202210255087A CN 114330461 B CN114330461 B CN 114330461B
- Authority
- CN
- China
- Prior art keywords
- voltage
- reactive current
- track
- time derivative
- value
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Active
Links
- 238000000034 method Methods 0.000 title claims abstract description 48
- 238000013528 artificial neural network Methods 0.000 title claims abstract description 9
- 238000012545 processing Methods 0.000 claims description 87
- 238000011176 pooling Methods 0.000 claims description 78
- 238000010586 diagram Methods 0.000 claims description 21
- 238000010606 normalization Methods 0.000 claims description 20
- 238000013507 mapping Methods 0.000 claims description 19
- 238000013527 convolutional neural network Methods 0.000 claims description 17
- 238000000605 extraction Methods 0.000 claims description 15
- 230000004913 activation Effects 0.000 claims description 13
- 238000012958 reprocessing Methods 0.000 claims description 12
- 230000011218 segmentation Effects 0.000 claims description 9
- 238000009499 grossing Methods 0.000 claims description 8
- 230000008707 rearrangement Effects 0.000 claims description 8
- 230000008859 change Effects 0.000 claims description 5
- 230000001131 transforming effect Effects 0.000 claims description 3
- 238000003062 neural network model Methods 0.000 abstract description 12
- 230000005611 electricity Effects 0.000 description 10
- 230000006870 function Effects 0.000 description 3
- 238000005070 sampling Methods 0.000 description 3
- 230000002159 abnormal effect Effects 0.000 description 2
- 230000006399 behavior Effects 0.000 description 2
- 230000009286 beneficial effect Effects 0.000 description 2
- 238000004364 calculation method Methods 0.000 description 2
- 238000013145 classification model Methods 0.000 description 2
- 238000012937 correction Methods 0.000 description 2
- 238000001514 detection method Methods 0.000 description 2
- 239000000463 material Substances 0.000 description 2
- 230000001052 transient effect Effects 0.000 description 2
- OKTJSMMVPCPJKN-UHFFFAOYSA-N Carbon Chemical compound [C] OKTJSMMVPCPJKN-UHFFFAOYSA-N 0.000 description 1
- 238000012951 Remeasurement Methods 0.000 description 1
- 230000004075 alteration Effects 0.000 description 1
- 229910052799 carbon Inorganic materials 0.000 description 1
- 238000007405 data analysis Methods 0.000 description 1
- 238000000354 decomposition reaction Methods 0.000 description 1
- 238000005516 engineering process Methods 0.000 description 1
- 230000004927 fusion Effects 0.000 description 1
- 238000012986 modification Methods 0.000 description 1
- 230000004048 modification Effects 0.000 description 1
- 230000003287 optical effect Effects 0.000 description 1
- 239000013307 optical fiber Substances 0.000 description 1
- 238000005192 partition Methods 0.000 description 1
- 230000008569 process Effects 0.000 description 1
- 230000009467 reduction Effects 0.000 description 1
- 238000006467 substitution reaction Methods 0.000 description 1
- 238000012549 training Methods 0.000 description 1
Images
Landscapes
- Image Analysis (AREA)
Abstract
The invention discloses a V-I track generation method, a device and a neural network for non-invasive load identification, wherein the method comprises the following steps: when a switching event is determined to occur, acquiring an initial V-I track, wherein the initial V-I track is a track of voltage and reactive current changing along with time; acquiring a time derivative of the voltage and a time derivative of the reactive current based on the initial V-I track; the initial V-I trajectory is mapped to the color space based on the time derivative of the voltage and the time derivative of the reactive current to obtain a color V-I trajectory. Therefore, the initial V-I track is mapped to the color space based on the time derivative of the voltage and the reactive current to obtain the color V-I track, so that the information carrying capacity of the V-I track is improved, the V-I track can be suitable for various load identification scenes of electric appliances, and meanwhile, the formed color V-I track can be trained by utilizing a neural network model, and the accuracy of the V-I track on non-invasive load identification is improved.
Description
Technical Field
The invention relates to the technical field of non-invasive load identification, in particular to a V-I track generation method and device for non-invasive load identification and a neural network.
Background
Non-intrusive load identification means that detection equipment is installed at a load group port (for example, an incoming electricity meter of household electricity), so that switching of each electric appliance in a load group and detection of electricity utilization behaviors can be achieved, at present, the total electricity utilization amount of residents and the types and the number of the resident household appliances are gradually increased, effective electricity utilization data statistics can be provided for a smart grid by analyzing electricity consumption of various household appliances and electricity utilization behaviors of the residents through a non-intrusive load identification technology, through big electricity utilization data analysis, resident electricity utilization strategies in different regions and different seasons are formulated in an auxiliary mode, electricity utilization efficiency of the residents can be improved, and energy saving and carbon reduction are achieved.
Non-intrusive load identification methods can be generally classified into two broad categories: the load identification method based on the event comprises the steps of identifying the switching event of the electric appliance by extracting various load characteristics, further calculating the switching state and the power load of the electric appliance to realize power consumption decomposition, and extracting characteristics by the load identification method based on the event, wherein the characteristics comprise transient characteristics and steady characteristics, and the transient characteristics comprise power change, starting current waveform, voltage noise, edge size or kurtosis of current waveform peak and the like; the steady state characteristics include power (including active, reactive, apparent power), steady state voltage and current waveforms, V-I trajectories, steady state current and voltage harmonics, and the like.
The V-I track is usually used as a load identification method based on events to extract features, the track is a closed curve drawn according to the relation between voltage and reactive current, but the traditional V-I track is a gray image, only has one channel and contains less information, and cannot be applied to various load identification scenes of electrical appliances; in addition, the traditional gray level V-I track is difficult to train by using a neural network model, and the accuracy of the V-I track on non-invasive load recognition is reduced.
Disclosure of Invention
The present invention is directed to solving, at least to some extent, one of the technical problems in the related art. Therefore, a first objective of the present invention is to provide a V-I trajectory generation method for non-intrusive load identification, in which an initial V-I trajectory is mapped to a color space based on time derivatives of voltage and reactive current to obtain a color V-I trajectory, so as to improve information carrying capacity of the V-I trajectory, and make it suitable for various load identification scenarios of electrical appliances, and meanwhile, the formed color V-I trajectory can be trained by using a neural network model, which is beneficial to improving accuracy of the V-I trajectory for non-intrusive load identification.
A second object of the invention is to propose a V-I trajectory generation device for non-intrusive load recognition.
A third object of the invention is to propose a computer-readable storage medium.
A fourth object of the invention is to propose an electronic device.
A fifth object of the present invention is to provide a convolutional neural network.
In order to achieve the above object, a first embodiment of the present invention provides a V-I trajectory generation method for non-intrusive load identification, where the method includes: when a switching event is determined to occur, obtaining an initial V-I track, wherein the initial V-I track is a track of voltage and reactive current changing along with time; acquiring a time derivative of the voltage and a time derivative of the reactive current based on the initial V-I track; the initial V-I trajectory is mapped to the color space based on the time derivative of the voltage and the time derivative of the reactive current to obtain a color V-I trajectory.
According to the V-I track generation method for non-intrusive load identification, when a switching event is determined to occur, an initial V-I track of voltage and reactive current changing along with time is obtained, a time derivative of the voltage and a time derivative of the reactive current are obtained based on the initial V-I track, and the initial V-I track is mapped to a color space according to the time derivatives of the voltage and the reactive current to obtain a color V-I track. Therefore, the initial V-I track is mapped to the color space based on the time derivative of the voltage and the reactive current to obtain the color V-I track, so that the information carrying capacity of the V-I track is improved, the V-I track can be suitable for various load identification scenes of electric appliances, and meanwhile, the formed color V-I track can be trained by utilizing a neural network model, and the accuracy of the V-I track on non-intrusive load identification is improved.
According to an embodiment of the invention, the method further comprises: collecting a voltage signal and a current signal; carrying out Fourier transform on the voltage signal and the current signal to obtain voltage and current; obtaining reactive current according to the voltage and the current; and when the change conditions of the voltage and the reactive current meet preset conditions, determining that a switching event occurs.
According to an embodiment of the invention, before fourier transforming the voltage signal and the current signal to obtain the voltage and the current, the method further comprises: and smoothing the voltage signal and the current signal.
According to one embodiment of the present invention, smoothing a voltage signal and a current signal includes: and performing one-dimensional convolution processing on the voltage signal and the current signal.
According to one embodiment of the invention, obtaining an initial V-I trajectory comprises: acquiring voltage and reactive current of at least one power frequency period before a switching event occurs, and acquiring voltage and reactive current of at least one power frequency period after the switching event occurs; carrying out average processing on the voltage and the reactive current of at least one power frequency period before the switching event occurs to obtain the voltage and the reactive current of a first power frequency period, and carrying out average processing on the voltage and the reactive current of at least one power frequency period after the switching event occurs to obtain the voltage and the reactive current of a second power frequency period; and generating an initial V-I track according to the voltage and the reactive current of the first power frequency period and the voltage and the reactive current of the second power frequency period.
According to one embodiment of the invention, mapping the initial V-I trajectory to the color space according to the time derivative of the voltage and the time derivative of the reactive current to obtain a color V-I trajectory comprises: aiming at the trace points on the initial V-I trace, determining the R value, the G value and the B value of the trace points according to the time derivative of the voltage and the time derivative of the reactive current; and generating a color V-I track according to the R value, the G value and the B value of the track point.
According to one embodiment of the invention, determining the R, G and B values of the trace points from the time derivative of the voltage and the time derivative of the reactive current comprises: when the time derivative of the voltage is greater than or equal to 0 and the time derivative of the reactive current is greater than or equal to 0, determining the R value of the trace point as a first value; when the time derivative of the voltage is greater than or equal to 0 and the time derivative of the reactive current is less than 0, determining the G value of the trace point as a first value; when the time derivative of the voltage is less than 0 and the time derivative of the reactive current is greater than or equal to 0, determining the B value of the trace point as a first value; and when the time derivative of the voltage is less than 0 and the time derivative of the reactive current is less than 0, determining that the R value and the B value of the trace point are both first values.
In order to achieve the above object, a second embodiment of the present invention provides a V-I trajectory generation apparatus for non-intrusive load identification, the apparatus including: the acquisition module is used for acquiring an initial V-I track when the switching event is determined to occur, wherein the initial V-I track is a track of voltage and reactive current changing along with time; and the mapping module is used for acquiring the time derivative of the voltage and the time derivative of the reactive current based on the initial V-I track and mapping the initial V-I track to the color space according to the time derivative of the voltage and the time derivative of the reactive current to obtain a color V-I track.
According to the V-I track generation device for non-intrusive load identification, an initial V-I track of voltage and reactive current changing along with time is obtained through an obtaining module when a switching event is determined to occur; and acquiring a time derivative of the voltage and a time derivative of the reactive current based on the initial V-I track through a mapping module, and mapping the initial V-I track to a color space according to the time derivatives of the voltage and the reactive current to obtain a color V-I track. Therefore, the initial V-I track is mapped to the color space based on the time derivative of the voltage and the reactive current to obtain the color V-I track, so that the information carrying capacity of the V-I track is improved, the V-I track can be suitable for various load identification scenes of electric appliances, and meanwhile, the formed color V-I track can be trained by utilizing a neural network model, and the accuracy of the V-I track on non-intrusive load identification is improved.
According to an embodiment of the present invention, the obtaining module is specifically configured to: acquiring voltage and reactive current of at least one power frequency period before a switching event occurs, and acquiring voltage and reactive current of at least one power frequency period after the switching event occurs; carrying out average processing on the voltage and the reactive current of at least one power frequency period before the switching event occurs to obtain the voltage and the reactive current of a first power frequency period, and carrying out average processing on the voltage and the reactive current of at least one power frequency period after the switching event occurs to obtain the voltage and the reactive current of a second power frequency period; and generating an initial V-I track according to the voltage and the reactive current of the first power frequency period and the voltage and the reactive current of the second power frequency period.
According to an embodiment of the present invention, the mapping module is specifically configured to: aiming at the trace points on the initial V-I trace, determining the R value, the G value and the B value of the trace points according to the time derivative of the voltage and the time derivative of the reactive current; and generating a color V-I track according to the R value, the G value and the B value of the track point.
According to an embodiment of the present invention, the mapping module is specifically configured to: when the time derivative of the voltage is greater than or equal to 0 and the time derivative of the reactive current is greater than or equal to 0, determining the R value of the trace point as a first value; when the time derivative of the voltage is greater than or equal to 0 and the time derivative of the reactive current is less than 0, determining the G value of the trace point as a first value; when the time derivative of the voltage is less than 0 and the time derivative of the reactive current is greater than or equal to 0, determining the B value of the trace point as a first value; and when the time derivative of the voltage is less than 0 and the time derivative of the reactive current is less than 0, determining that the R value and the B value of the trace point are both first values.
To achieve the above object, a third embodiment of the present invention proposes a computer-readable storage medium on which a V-I trajectory generation program is stored, which when executed by a processor implements the V-I trajectory generation method as in the first embodiment.
According to the computer-readable storage medium for non-intrusive load identification, the initial V-I track is mapped to the color space based on the time derivative of the voltage and the reactive current through the V-I track generation method for non-intrusive load identification to obtain the color V-I track, so that the information carrying capacity of the V-I track is improved, the V-I track can be suitable for various load identification scenes of electric appliances, and meanwhile, the formed color V-I track can be trained by using a neural network model, and the accuracy of the V-I track on non-intrusive load identification is improved.
To achieve the above object, a fourth aspect of the present invention provides an electronic device, including: the device comprises a memory, a processor and a V-I track generation program which is stored on the memory and can run on the processor, and when the processor executes the program, the V-I track generation method in the embodiment of the first aspect is realized.
According to the electronic equipment provided by the embodiment of the invention, through the V-I track generation method for non-intrusive load identification, the initial V-I track is mapped to the color space based on the time derivative of the voltage and the reactive current to obtain the color V-I track, so that the information carrying capacity of the V-I track is improved, the V-I track can be suitable for various load identification scenes of electric appliances, and meanwhile, the formed color V-I track can be trained by using a neural network model, and the accuracy of the V-I track on non-intrusive load identification is favorably improved.
In order to achieve the above object, a fifth embodiment of the present invention provides a convolutional neural network for performing feature extraction on a color V-I trajectory generated by the V-I trajectory generation method in the first embodiment, where the convolutional neural network includes: the multi-level residual error network is used for performing multi-level feature extraction on the input image to obtain feature maps of different levels, and splicing the feature maps of the different levels to obtain M1 feature maps; the reprocessing unit is used for carrying out random uniform rearrangement, pooling and splicing treatment on the M1 characteristic graphs to obtain M2 characteristic graphs; and the depth separable convolution unit is used for performing depth separable convolution processing on the M2 feature maps to obtain M3 feature maps, wherein M1, M2 and M3 are all positive integers.
According to the convolutional neural network disclosed by the embodiment of the invention, multi-level feature extraction and splicing are carried out on input images through a multi-level residual network so as to obtain M1 feature maps, random uniform rearrangement, pooling and splicing processing are carried out on M1 feature maps at the processing unit so as to obtain M2 feature maps, and depth separable convolution processing is carried out on M2 feature maps through a depth separable convolution unit so as to obtain M3 feature maps, so that more abundant feature maps can be extracted from an input color V-I track map, and the accuracy of model prediction is further improved.
According to one embodiment of the invention, a multi-level residual network comprises: the input end of the 1 st convolution unit is used as the input end of a convolution neural network, the input end of the (i + 1) th convolution unit is connected with the first output end of the ith convolution unit, and the N convolution units are used for performing multi-level feature extraction on an input image to obtain feature maps of different levels; the input end of the ith first maximum pooling layer is connected with the second output end of the ith convolution unit and is used for performing maximum pooling on the feature map of the corresponding convolution unit, N is an integer greater than or equal to 2, and i is greater than or equal to 1 and is less than N; and the input end of the first splicing layer is respectively connected with the output end of the Nth convolution unit and the output end of each of the N-1 first maximum pooling layers, and is used for splicing the feature map output by the Nth convolution unit and the feature map subjected to maximum pooling to obtain M1 feature maps.
According to one embodiment of the invention, each convolution unit of the N convolution units comprises: the device comprises a convolution layer, a first normalization processing layer and a ReLU activation layer.
According to one embodiment of the invention, the pooling cores of each of the N-1 first maximum pooling layers are different in size.
According to one embodiment of the invention, the reprocessing unit includes: the input end of the shuffling layer is connected with the multi-level residual error network and is used for shuffling the M1 characteristic graphs; the input end of the splitting layer is connected with the output end of the shuffling layer, the splitting layer comprises K output ends and is used for carrying out uniform random splitting processing on the M1 characteristic graphs after shuffling processing to obtain K groups of characteristic graphs and outputting the K groups of characteristic graphs through the K output ends, and K is an integer greater than or equal to 2; the spatial pyramid pooling layer comprises K-1 second maximum pooling layers, and the input ends of the K-1 second maximum pooling layers are correspondingly connected with the 1 st to the K-1 st output ends in the segmentation layers and are used for performing maximum pooling on the feature maps output by the 1 st to the K-1 st output ends; and the input end of the second splicing layer is respectively connected with the Kth output end of the segmentation layer and the output end of each second maximum pooling layer in the K-1 second maximum pooling layers, and the second splicing layer is used for splicing the characteristic diagram output by the Kth output end and the characteristic diagram after maximum pooling to obtain M2 characteristic diagrams.
According to one embodiment of the invention, the pooling core size of each of the K-1 second largest pooling layers is different.
According to one embodiment of the present invention, a depth separable convolution unit includes: the input end of the channel-by-channel convolution layer is connected with the reprocessing unit and is used for carrying out channel-by-channel convolution processing on the M2 feature maps; the input end of the point-by-point convolution layer is connected with the output end of the channel-by-channel convolution layer and is used for performing point-by-point convolution processing on the feature map subjected to the channel-by-channel convolution processing; the input end of the second normalization processing layer is connected with the output end of the point-by-point convolution layer and is used for performing normalization processing on the feature map subjected to the point-by-point convolution processing; and the input end of the ReLU activation layer is connected with the output end of the second normalization processing layer and is used for performing activation processing on the feature maps after normalization processing to obtain M3 feature maps.
Additional aspects and advantages of the invention will be set forth in part in the description which follows and, in part, will be obvious from the description, or may be learned by practice of the invention.
Drawings
FIG. 1 is a flow diagram of a V-I trajectory generation method for non-intrusive load identification, in accordance with one embodiment of the present invention;
FIG. 2 is a flow chart for obtaining voltage and reactive current according to one embodiment of the present invention;
FIG. 3 is a schematic structural diagram of a V-I trajectory generation apparatus for non-intrusive load identification, in accordance with one embodiment of the present invention;
FIG. 4 is a schematic diagram of a convolutional neural network according to one embodiment of the present invention;
FIG. 5 is a schematic diagram of a multi-level residual network according to an embodiment of the present invention;
FIG. 6 is a schematic diagram of the structure of a convolution unit according to one embodiment of the present invention;
FIG. 7 is a schematic diagram of a reprocessing unit according to one embodiment of the present invention;
FIG. 8 is a schematic diagram of the structure of a depth separable convolution element in accordance with one embodiment of the present invention.
Detailed Description
Reference will now be made in detail to embodiments of the present invention, examples of which are illustrated in the accompanying drawings, wherein like or similar reference numerals refer to the same or similar elements or elements having the same or similar function throughout. The embodiments described below with reference to the drawings are illustrative and intended to be illustrative of the invention and are not to be construed as limiting the invention.
A V-I trajectory generation method, an apparatus, a computer-readable storage medium, an electronic device, and a convolutional neural network for non-intrusive load recognition according to embodiments of the present invention are described below with reference to the accompanying drawings.
FIG. 1 is a flow diagram of a V-I trajectory generation method for non-intrusive load identification, in accordance with one embodiment of the present invention. As shown in fig. 1, the V-I trajectory generation method for non-intrusive load recognition includes the steps of:
and S101, when the switching event is determined to occur, acquiring an initial V-I track, wherein the initial V-I track is a track of voltage and reactive current changing along with time.
Specifically, the waveforms of active currents of different electrical appliances are difficult to distinguish from the waveforms except for different amplitudes, the waveforms of reactive currents of different electrical appliances are large in difference, different electrical appliances are easy to distinguish by adopting the reactive current waveforms, and therefore when a switching event is determined to occur, a V-I track drawn by voltage and reactive current is selected to carry out non-invasive load identification.
It should be noted that, as shown in fig. 2, when determining whether a switching event occurs, the method further includes the following steps:
step S201, collecting a voltage signal and a current signal.
Specifically, when it is determined whether a switching event occurs to the electrical appliance, the preset voltage acquisition device and the preset current acquisition device are adopted to respectively acquire voltage signals and current signals of different electrical appliances so as to analyze and identify whether the switching event occurs to the corresponding electrical appliance.
Step S202, Fourier transform is carried out on the voltage signal and the current signal to obtain voltage and current.
Specifically, since the voltage signal and the current signal acquired by the voltage acquisition device and the current acquisition device are converted from an ADC (Analog-to-Digital Converter), and are not true voltage and current values, fourier transform processing needs to be performed on the acquired voltage signal and current signal, and then the acquired voltage signal and current signal are multiplied by corresponding coefficients, respectively, to obtain true voltage and current values.
In some embodiments, before fourier transforming the voltage signal and the current signal to obtain the voltage and the current, the method further comprises: and smoothing the voltage signal and the current signal.
Further, smoothing the voltage signal and the current signal includes: and performing one-dimensional convolution processing on the voltage signal and the current signal.
That is, before performing fourier transform on the voltage signal and the current signal to obtain a voltage and a current, smoothing processing is performed on the collected voltage and current signal to remove abnormal values and noise in the voltage and current signal, and meanwhile, in order to reduce the amount of calculation in the smoothing processing, an average filter with a filter kernel size of 5 is selected, and further, since the collected voltage and current signal is a time sequence signal, one-dimensional convolution processing can be performed on the voltage signal and the current signal so that a plurality of such signals can be processed in parallel, thereby improving the signal processing efficiency, that is, finally, one-dimensional convolution with a convolution kernel length of 5 can be used to smooth the collected voltage and current signal to remove abnormal values and noise in the voltage and current signal, and optionally, the convolution kernel is [0.2,0.2,0.2,0.2,0.2 ]. The voltage signal and the current signal in the fourier transform of the voltage signal and the current signal are one-dimensional convolution processed voltage signal and current signal.
Step S203, reactive current is obtained according to the voltage and the current.
Specifically, the real voltage and current can be obtained after one-dimensional convolution processing and Fourier transform are carried out on the collected voltage signal and current signal, and the obtained current can be obtained according to the Fryze power theorySplit into active currentAnd reactive currentAs shown in the following formula (1):
active powerThe average value of the instantaneous power in one period T is shown in the following formula (2):
Assume a sampling frequency ofWith an alternating current frequency ofThe number of sampling points in one cycle of the alternating current isThe time interval between two sampling points isThen the active power is:
According to the relation between active power and active current, the active currentCan be calculated as follows:
substituting the formulas (3) and (4) into the formula (1) and finishing to obtain reactive currentComprises the following steps:
and step S204, when the change conditions of the voltage and the reactive current meet preset conditions, determining that a switching event occurs.
Specifically, whether a switching event occurs is judged according to the obtained voltage and reactive current, and when the change conditions of the voltage and the reactive current meet preset conditions, the switching event is determined to occur; otherwise, no further measures are taken until a switching event occurs.
In some embodiments, obtaining an initial V-I trajectory includes: acquiring voltage and reactive current of at least one power frequency period before a switching event occurs, and acquiring voltage and reactive current of at least one power frequency period after the switching event occurs; carrying out average processing on the voltage and the reactive current of at least one power frequency period before the switching event occurs to obtain the voltage and the reactive current of a first power frequency period, and carrying out average processing on the voltage and the reactive current of at least one power frequency period after the switching event occurs to obtain the voltage and the reactive current of a second power frequency period; and generating an initial V-I track according to the voltage and the reactive current of the first power frequency period and the voltage and the reactive current of the second power frequency period.
That is to say, when it is determined that a switching event occurs, the voltage and the reactive current of at least one power frequency period before the switching event occurs and the voltage and the reactive current of at least one power frequency period after the switching event occurs are respectively obtained, for example, the voltage and the reactive current of 100 power frequency periods before the switching event occurs and the voltage and the reactive current of 100 power frequency periods after the switching event occurs can be obtained, for the voltage and the reactive current of 100 power frequency periods before the switching event occurs, 128 points can be evenly divided into each power frequency period, the voltage and the reactive current corresponding to each point are obtained, the voltages and the reactive currents of the same points of 100 power frequency periods are added to calculate an average value, and the voltage and the reactive current of the first power frequency period can be obtained; for the voltage and the reactive current of 100 power frequency periods after the switching event occurs, 128 points can be equally divided in each power frequency period, the voltage and the reactive current corresponding to each point are obtained, the voltage and the reactive current of the same point of the 100 power frequency periods are added to calculate the average value, namely the voltage and the reactive current of the second power frequency period can be obtained, and the initial V-I track of one period can be generated according to the voltage and the reactive current of the first power frequency period and the voltage and the reactive current of the second power frequency period.
Step S102, acquiring a time derivative of the voltage and a time derivative of the reactive current based on the initial V-I track.
Specifically, the voltage and the initial V-I track of the reactive current are respectively obtained according to the obtained voltage and the initial V-I track of the reactive currentAnd reactive currentTime derivative of (1)And。
step S103, mapping the initial V-I track to a color space according to the time derivative of the voltage and the time derivative of the reactive current to obtain a color V-I track.
In particular, according to voltageAnd reactive currentTime derivative of (1)Andthe initial V-I track is mapped to a color space according to a certain rule to obtain a color V-I track, so that the information remeasurement of the V-I track is improved, and the accuracy of a non-invasive load identification model based on the color V-I track as a characteristic is higher.
In some embodiments, mapping the initial V-I trajectory to color space according to the time derivative of the voltage and the time derivative of the reactive current to obtain a color V-I trajectory comprises: aiming at the trace points on the initial V-I trace, determining the R value, the G value and the B value of the trace points according to the time derivative of the voltage and the time derivative of the reactive current; and generating a color V-I track according to the R value, the G value and the B value of the track point.
Further, determining the R value, the G value and the B value of the trace point according to the time derivative of the voltage and the time derivative of the reactive current, wherein the steps comprise: when the time derivative of the voltage is greater than or equal to 0 and the time derivative of the reactive current is greater than or equal to 0, determining the R value of the trace point as a first value; when the time derivative of the voltage is greater than or equal to 0 and the time derivative of the reactive current is less than 0, determining the G value of the trace point as a first value; when the time derivative of the voltage is less than 0 and the time derivative of the reactive current is greater than or equal to 0, determining the B value of the trace point as a first value; and when the time derivative of the voltage is less than 0 and the time derivative of the reactive current is less than 0, determining that the R value and the B value of the trace point are both first values.
In particular, falseIs provided withIs the ith track point on the initial V-I track,、andthe mapping rules of the ith track point in the initial V-I track in three dimensions of R, G and B are shown as formula (7):
that is, when the time derivative of the voltage is equal to or greater than 0 and the time derivative of the reactive current is equal to or greater than 0,taking a first value of 1; when the time derivative of the voltage is equal to or greater than 0 and the time derivative of the reactive current is less than 0,taking a first value of 1; when the time derivative of the voltage is less than 0 and the time derivative of the reactive current is greater than or equal to 0,taking a first value of 1; when the time derivative of the voltage is less than 0 and the time derivative of the reactive current is less than 0,andthe first value 1 is adopted, the ith track point of the initial V-I track is mapped to R, G and B color space according to the time derivative values of voltage and reactive current, each track point in the initial V-I track can be mapped to R, G and B color space, a corresponding color V-I track is finally formed, the formed color V-I track can be used as training data of a neural network model, and the color V-I track trained by the neural network is beneficial to improving the accuracy of non-intrusive load identification.
In summary, according to the V-I trajectory generation method for non-intrusive load identification of the embodiment of the present invention, when it is determined that a switching event occurs, an initial V-I trajectory of voltage and reactive current changing with time is obtained, and based on a time derivative of voltage and a time derivative of reactive current obtained from the initial V-I trajectory, the initial V-I trajectory is mapped to a color space according to the time derivatives of voltage and reactive current to obtain a color V-I trajectory. Therefore, the initial V-I track is mapped to the color space based on the time derivative of the voltage and the reactive current to obtain the color V-I track, so that the information carrying capacity of the V-I track is improved, the V-I track can be suitable for various load identification scenes of electric appliances, and meanwhile, the formed color V-I track can be trained by utilizing a neural network model, and the accuracy of the V-I track on non-invasive load identification is improved.
Fig. 3 is a schematic structural diagram of a V-I trajectory generation device for non-intrusive load identification according to an embodiment of the present invention. As shown in fig. 3, the V-I trajectory generation apparatus 100 for non-intrusive load recognition includes: an acquisition module 110 and a mapping module 120.
The obtaining module 110 is configured to obtain an initial V-I trajectory when it is determined that a switching event occurs, where the initial V-I trajectory is a trajectory of voltage and reactive current changing with time; the mapping module 120 is configured to obtain a time derivative of the voltage and a time derivative of the reactive current based on the initial V-I trajectory, and map the initial V-I trajectory to the color space according to the time derivative of the voltage and the time derivative of the reactive current to obtain a color V-I trajectory.
In some embodiments, the obtaining module 110 is specifically configured to: acquiring voltage and reactive current of at least one power frequency period before a switching event occurs, and acquiring voltage and reactive current of at least one power frequency period after the switching event occurs; carrying out average processing on the voltage and the reactive current of at least one power frequency period before the switching event occurs to obtain the voltage and the reactive current of a first power frequency period, and carrying out average processing on the voltage and the reactive current of at least one power frequency period after the switching event occurs to obtain the voltage and the reactive current of a second power frequency period; and generating an initial V-I track according to the voltage and the reactive current of the first power frequency period and the voltage and the reactive current of the second power frequency period. In some embodiments, the generated power obtaining module 130 is further specifically configured to: acquiring initial generating power of the hybrid electric vehicle; acquiring a corresponding correction coefficient according to the working condition; and correcting the initial generating power according to the correction coefficient to obtain the target generating power.
In some embodiments, the mapping module 120 is specifically configured to: aiming at the trace points on the initial V-I trace, determining the R value, the G value and the B value of the trace points according to the time derivative of the voltage and the time derivative of the reactive current; and generating a color V-I track according to the R value, the G value and the B value of the track point.
In some embodiments, the mapping module 120 is specifically configured to: when the time derivative of the voltage is greater than or equal to 0 and the time derivative of the reactive current is greater than or equal to 0, determining the R value of the trace point as a first value; when the time derivative of the voltage is greater than or equal to 0 and the time derivative of the reactive current is less than 0, determining the G value of the trace point as a first value; when the time derivative of the voltage is less than 0 and the time derivative of the reactive current is greater than or equal to 0, determining the B value of the trace point as a first value; and when the time derivative of the voltage is less than 0 and the time derivative of the reactive current is less than 0, determining that the R value and the B value of the trace point are both the first values.
It should be noted that, for the description of the V-I trajectory generation device for non-intrusive load identification in the present application, please refer to the description of the V-I trajectory generation method for non-intrusive load identification in the present application, and detailed description thereof is omitted here.
According to the V-I track generation device for non-intrusive load identification, an initial V-I track of voltage and reactive current changing along with time is obtained through an obtaining module when a switching event is determined to occur; and acquiring a time derivative of the voltage and a time derivative of the reactive current based on the initial V-I track through a mapping module, and mapping the initial V-I track to a color space according to the time derivatives of the voltage and the reactive current to obtain a color V-I track. Therefore, the initial V-I track is mapped to the color space based on the time derivative of the voltage and the reactive current to obtain the color V-I track, so that the information carrying capacity of the V-I track is improved, the V-I track can be suitable for various load identification scenes of electric appliances, and meanwhile, the formed color V-I track can be trained by utilizing a neural network model, and the accuracy of the V-I track on non-invasive load identification is improved.
Embodiments of the present invention also provide a computer-readable storage medium having stored thereon a V-I trajectory generation program that, when executed by a processor, implements the V-I trajectory generation method as described above.
According to the computer-readable storage medium for non-intrusive load identification, the initial V-I track is mapped to the color space based on the time derivative of the voltage and the reactive current through the V-I track generation method for non-intrusive load identification, so that the color V-I track is obtained, the information carrying capacity of the V-I track is improved, the V-I track can be suitable for various load identification scenes of electric appliances, and meanwhile, the formed color V-I track can be trained by using a neural network model, so that the accuracy of the V-I track on non-intrusive load identification is improved.
An embodiment of the present invention further provides an electronic device, including: the device comprises a memory, a processor and a V-I track generation program which is stored on the memory and can run on the processor, wherein when the processor executes the program, the V-I track generation method is realized.
According to the electronic equipment provided by the embodiment of the invention, by the V-I track generation method for non-intrusive load identification, the initial V-I track is mapped to the color space based on the time derivative of the voltage and the reactive current to obtain the color V-I track, so that the information carrying capacity of the V-I track is improved, the V-I track can be suitable for various load identification scenes of electric appliances, and meanwhile, the formed color V-I track can be trained by utilizing a neural network model, and the accuracy of the V-I track on non-intrusive load identification is favorably improved.
An embodiment of the present invention further provides a convolutional neural network, configured to perform feature extraction on a color V-I trajectory generated by the V-I trajectory generation method in the first aspect, as shown in fig. 4, where the convolutional neural network 200 includes: a multi-level residual network 210, a reprocessing unit 220, and a depth separable convolution unit 230.
The multilevel residual network 210 is configured to perform multilevel feature extraction on an input image to obtain feature maps of different levels, and splice the feature maps of the different levels to obtain M1 feature maps; the reprocessing unit 220 is configured to perform random uniform rearrangement, pooling and splicing on the M1 feature maps to obtain M2 feature maps; the depth separable convolution unit 230 is configured to perform depth separable convolution processing on the M2 feature maps to obtain M3 feature maps, where M1, M2, and M3 are all positive integers.
Specifically, the generated color V-I trajectory graph is input to the multi-level residual network 210, multi-level feature extraction is performed on the color V-I trajectory graph through the multi-level residual network 210 to obtain feature graphs of different levels, the feature graphs of different levels are spliced to obtain M1 feature graphs, the reprocessing unit 220 performs random uniform rearrangement, pooling and splicing on the M1 feature graphs generated by the color V-I trajectory graph to obtain M2 feature graphs, the depth separable convolution unit 230 performs depth separable convolution on the obtained M2 feature graphs to finally obtain M3 feature graphs, and the extracted M3 feature graphs are used to train a classification model to obtain a non-intrusive load identification model.
According to the convolutional neural network disclosed by the embodiment of the invention, multi-level feature extraction and concatenation are carried out on input images through a multi-level residual network to obtain M1 feature maps, random uniform rearrangement, pooling and concatenation processing are carried out on M1 feature maps at a processing unit to obtain M2 feature maps, and depth separable convolution processing is carried out on M2 feature maps at a depth separable convolution unit to obtain M3 feature maps, so that more abundant feature maps can be extracted from an input color V-I track map, and the accuracy of non-intrusive load identification is further improved.
In some embodiments, the multi-level residual network 210 includes: the input end of the 1 st convolution unit is used as the input end of a convolution neural network, the input end of the (i + 1) th convolution unit is connected with the first output end of the ith convolution unit, and the N convolution units are used for performing multi-level feature extraction on an input image to obtain feature maps of different levels; the input end of the ith first maximum pooling layer is connected with the second output end of the ith convolution unit and is used for performing maximum pooling processing on the feature map of the corresponding convolution unit, N is an integer which is more than or equal to 2, and i is more than or equal to 1 and less than N; and the input end of the first splicing layer is respectively connected with the output end of the Nth convolution unit and the output end of each of the N-1 first maximum pooling layers, and is used for splicing the feature map output by the Nth convolution unit and the feature map subjected to maximum pooling to obtain M1 feature maps. Optionally, the pooling cores of each of the N-1 first maximum pooling layers are all different in size.
Specifically, the convolution units may include at least two or more convolution units, and may be selected according to actual requirements, as a specific example, as shown in fig. 5, if the multi-level residual network 210 includes 3 convolution units 211, then the multi-level residual network 210 includes 2 max pooling layers 212 and a first stitching layer 213, the input end of the 1 st convolution unit 211 is used as the input end of the convolution neural network and is used for receiving the color V-I track image, the first output end and the second output end of the 1 st convolution unit 211 are respectively connected with the input ends of the 2 nd convolution unit 211 and the 1 st first maximum pooling layer 212, the first output end and the second output end of the 2 nd convolution unit 211 are respectively connected with the input ends of the 3 rd convolution unit 211 and the 2 nd first maximum pooling layer 212, and the output end of the 3 rd convolution unit 211 and the output end of the 2 first maximum pooling layer 212 are respectively connected with the first stitching layer 213.
When the multi-level residual network 210 works, the 3 convolution units 211 are used for performing multi-level feature extraction on an input color V-I trace graph to obtain feature graphs of different levels, an input end of the 1 st first maximum pooling layer 212 is connected with a second output end of the 1 st convolution unit 211, and the feature graph formed by the 1 st convolution unit 211 can be subjected to maximum pooling, and similarly, the 2 nd first maximum pooling layer 212 can be subjected to maximum pooling on the feature graph formed by the 2 nd convolution unit 211, and since the 2 nd first maximum pooling layers 212 have different sizes of pooling kernels, local features at different positions in the feature graph output by the corresponding convolution layer can be integrated, so that the richness of output features is greatly increased, the first splicing layer 213 splices the feature graphs output by the 3 rd convolution unit 211, the 1 st first maximum pooling layer 212, and the 2 nd first maximum pooling layer 212, to form M1 feature maps so that a combination of feature maps extracted by different convolution units can be implemented.
In some embodiments, as shown in FIG. 6, each convolution unit 211 of the N convolution units includes: convolutional layer 2110, first normalized processing layer 2111, and ReLU activation layer 2112.
Specifically, as shown in fig. 6, when the convolution units 211 are operated, the convolution layer 2110 of each convolution unit 211 is used for acquiring an input image and performing convolution processing, the first normalization processing layer 2111 is used for performing normalization processing on the convolution-processed image, and finally, the normalization-processed image is activated by the ReLU activation layer 2112 to form a corresponding feature map.
In some embodiments, reprocessing unit 220 includes: the device comprises a shuffling layer, a partition layer, a spatial pyramid pooling layer and a second splicing layer, wherein the input end of the shuffling layer is connected with a multi-level residual error network and is used for shuffling M1 feature maps; the input end of the dividing layer is connected with the output end of the shuffling layer, the dividing layer comprises K output ends and is used for carrying out uniform random dividing processing on the M1 characteristic graphs after shuffling processing to obtain K groups of characteristic graphs and outputting the K groups of characteristic graphs through the K output ends, and K is an integer greater than or equal to 2; the spatial pyramid pooling layers comprise K-1 second maximum pooling layers, and input ends of the K-1 second maximum pooling layers are correspondingly connected with the 1 st to the K-1 st output ends in the segmentation layers and are used for performing maximum pooling on feature maps output by the 1 st to the K-1 st output ends; and the input end of the second splicing layer is respectively connected with the Kth output end of the segmentation layer and the output end of each second maximum pooling layer in the K-1 second maximum pooling layers, and is used for splicing the characteristic diagram output by the Kth output end and the characteristic diagram subjected to maximum pooling to obtain M2 characteristic diagrams. Optionally, the pooling cores of each of the K-1 second largest pooling layers are different in size.
Specifically, the splitting layer may include at least two or more output ends, and the setting may be selected according to actual requirements, as a specific example, as shown in fig. 7, if the splitting layer 222 includes 4 output ends, the corresponding spatial pyramid pooling layer 223 includes 3 second maximum pooling layers 2230, the input end of the shuffle layer 221 is connected to the multi-level residual network 210, the input end of the splitting layer 222 is connected to the output end of the shuffle layer 221, the 1 st to 3 rd output ends in the splitting layer 222 are respectively connected to the 3 second maximum pooling layers 2230 in the spatial pyramid pooling layer 223, and the input end of the second splicing layer 224 is respectively connected to the 4 th output end in the splitting layer 222 and the output end of each of the 3 second maximum pooling layers 2230.
When the reprocessing unit 220 works, the shuffling layer 221 is configured to shuffle M1 feature maps input by the multi-level residual error network 210 to disrupt an original feature map distributed in a hierarchical convolution order, and implement random and uniform rearrangement of the multi-level residual error network feature maps, so as to ensure that the feature maps obtained in a post-convolution process can implement relatively sufficient information exchange and feature fusion reconstruction, 4 output ends of the segmentation layer 222 are configured to perform uniform and random segmentation processing on the shuffled M1 feature maps to obtain 4 sets of feature maps, 3 second maximum pooling layers 2230 are configured to perform maximum pooling processing on the feature maps output by the 1 st to 3 rd output ends of the segmentation layer 222, and since pooling kernels of each second maximum pooling layer are different in size and each second maximum pooling operation corresponds to a feature from a different convolution unit or a first maximum pooling layer in the multi-level residual error network, therefore, the feature richness is greatly increased, meanwhile, since M1 feature maps are divided into 4 groups of uniform feature maps, the calculation amount of the structure is greatly reduced, and the second stitching layer 224 is used for stitching the feature map output by the kth output end of the division layer 222 and the feature map after the maximum pooling processing to obtain M2 feature maps.
In some embodiments, as shown in FIG. 8, depth separable convolution element 230 includes: the channel-by-channel convolution layer 231, the point-by-point convolution layer 232, the second normalization processing layer 233 and the ReLU activation layer 234, wherein the input end of the channel-by-channel convolution layer 231 is connected with the re-processing unit 220 and is used for performing channel-by-channel convolution processing on the M2 feature maps; the input end of the point-by-point convolution layer 232 is connected with the output end of the channel-by-channel convolution layer 231 and is used for performing point-by-point convolution processing on the feature map subjected to the channel-by-channel convolution processing; the input end of the second normalization processing layer 233 is connected with the output end of the point-by-point convolution layer 231, and is used for performing normalization processing on the feature map subjected to the point-by-point convolution processing; the input end of the ReLU activation layer 234 is connected to the output end of the second normalization processing layer 233, and is used for performing activation processing on the feature maps after normalization processing to obtain M3 feature maps.
Specifically, when the depth separable convolution unit 230 operates, the channel-by-channel convolution layer 231 performs channel-by-channel convolution processing on the M2 feature maps output by the processing unit 220, the point-by-point convolution layer 232 performs point-by-point convolution processing on the feature maps subjected to the channel-by-channel convolution processing, the second normalization processing layer 233 performs normalization processing on the feature maps subjected to the point-by-point convolution processing, finally, the ReLU activation layer 234 performs activation processing on the feature maps subjected to the normalization processing to obtain M3 feature maps, and the extracted M3 feature maps are used to train the classification model to obtain the non-intrusive load identification model.
It should be noted that, if the M3 finally extracted feature maps do not meet the use requirement due to insufficient expression of the feature extraction network, fig. 4 may be repeatedly stacked as a module for multiple times to extract features with higher semantic meaning and richer content.
In summary, according to the convolutional neural network of the embodiment of the present invention, the multi-level feature extraction and concatenation are performed on the input images through the multi-level residual network to obtain M1 feature maps, the processing unit performs random uniform rearrangement, pooling and concatenation on the M1 feature maps to obtain M2 feature maps, and the depth separable convolution unit performs depth separable convolution on the M2 feature maps to obtain M3 feature maps, so that it is possible to extract richer feature maps from the input color V-I track map, and further improve the accuracy of non-intrusive load identification.
It should be noted that the logic and/or steps shown in the flowcharts or otherwise described herein, such as an ordered listing of executable instructions that can be considered to implement logical functions, can be embodied in any computer-readable medium for use by or in connection with an instruction execution system, apparatus, or device, such as a computer-based system, processor-containing system, or other system that can fetch the instructions from the instruction execution system, apparatus, or device and execute the instructions. For the purposes of this description, a "computer-readable medium" can be any means that can contain, store, communicate, propagate, or transport the program for use by or in connection with the instruction execution system, apparatus, or device. More specific examples (a non-exhaustive list) of the computer-readable medium would include the following: an electrical connection (electronic device) having one or more wires, a portable computer diskette (magnetic device), a Random Access Memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory (EPROM or flash memory), an optical fiber device, and a portable compact disc read-only memory (CDROM). Additionally, the computer-readable medium could even be paper or another suitable medium upon which the program is printed, as the program can be electronically captured, via for instance optical scanning of the paper or other medium, then compiled, interpreted or otherwise processed in a suitable manner if necessary, and then stored in a computer memory.
It should be understood that portions of the present invention may be implemented in hardware, software, firmware, or a combination thereof. In the above embodiments, the various steps or methods may be implemented in software or firmware stored in memory and executed by a suitable instruction execution system. For example, if implemented in hardware, as in another embodiment, any one or combination of the following techniques, which are known in the art, may be used: a discrete logic circuit having a logic gate circuit for implementing a logic function on a data signal, an application specific integrated circuit having an appropriate combinational logic gate circuit, a Programmable Gate Array (PGA), a Field Programmable Gate Array (FPGA), or the like.
In the description herein, references to the description of the term "one embodiment," "some embodiments," "an example," "a specific example," or "some examples," etc., mean that a particular feature, structure, material, or characteristic described in connection with the embodiment or example is included in at least one embodiment or example of the invention. In this specification, the schematic representations of the terms used above do not necessarily refer to the same embodiment or example. Furthermore, the particular features, structures, materials, or characteristics described may be combined in any suitable manner in any one or more embodiments or examples.
Furthermore, the terms "first", "second" and "first" are used for descriptive purposes only and are not to be construed as indicating or implying relative importance or to implicitly indicate the number of technical features indicated. Thus, a feature defined as "first" or "second" may explicitly or implicitly include at least one of the feature. In the description of the present invention, "a plurality" means at least two, e.g., two, three, etc., unless specifically limited otherwise.
In the present invention, unless otherwise expressly stated or limited, the terms "mounted," "connected," "secured," and the like are to be construed broadly and can, for example, be fixedly connected, detachably connected, or integrally formed; can be mechanically or electrically connected; they may be directly connected or indirectly connected through intervening media, or they may be connected internally or in any other suitable relationship, unless expressly stated otherwise. The specific meanings of the above terms in the present invention can be understood by those skilled in the art according to specific situations.
Although embodiments of the present invention have been shown and described above, it is understood that the above embodiments are exemplary and should not be construed as limiting the present invention, and that variations, modifications, substitutions and alterations can be made to the above embodiments by those of ordinary skill in the art within the scope of the present invention.
Claims (16)
1. A V-I trajectory generation method for non-intrusive load recognition, the method comprising:
when a switching event is determined to occur, acquiring an initial V-I track, wherein the initial V-I track is a track of voltage and reactive current changing along with time;
obtaining a time derivative of the voltage and a time derivative of the reactive current based on the initial V-I trajectory;
mapping the initial V-I trajectory to a color space according to the time derivative of the voltage and the time derivative of the reactive current to obtain a color V-I trajectory, wherein the method comprises: for the trace points on the initial V-I trace, determining an R value, a G value and a B value of the trace points according to the time derivative of the voltage and the time derivative of the reactive current; generating the color V-I track according to the R value, the G value and the B value of the track point;
wherein the determining of the R, G and B values of the trace points from the time derivative of the voltage and the time derivative of the reactive current comprises: when the time derivative of the voltage is greater than or equal to 0 and the time derivative of the reactive current is greater than or equal to 0, determining that the R value of the trace point is a first value; when the time derivative of the voltage is greater than or equal to 0 and the time derivative of the reactive current is less than 0, determining that the G value of the trace point is the first value; when the time derivative of the voltage is smaller than 0 and the time derivative of the reactive current is larger than or equal to 0, determining the value B of the trace point as the first value; and when the time derivative of the voltage is less than 0 and the time derivative of the reactive current is less than 0, determining that the R value and the B value of the trace point are both the first value.
2. The V-I trajectory generation method of claim 1, further comprising:
collecting a voltage signal and a current signal;
carrying out Fourier transform on the voltage signal and the current signal to obtain the voltage and the current;
obtaining the reactive current according to the voltage and the current;
and when the change conditions of the voltage and the reactive current meet preset conditions, determining that a switching event occurs.
3. The V-I trace generation method of claim 2, wherein prior to fourier transforming the voltage signal and the current signal to obtain the voltage and current, the method further comprises:
and smoothing the voltage signal and the current signal.
4. The V-I trajectory generation method of claim 3, wherein the smoothing the voltage signal and the current signal comprises:
and performing one-dimensional convolution processing on the voltage signal and the current signal.
5. The V-I trajectory generation method according to any one of claims 1-4, wherein the obtaining an initial V-I trajectory includes:
acquiring voltage and reactive current of at least one power frequency period before a switching event occurs, and acquiring voltage and reactive current of at least one power frequency period after the switching event occurs;
carrying out average processing on the voltage and the reactive current of at least one power frequency period before the switching event occurs to obtain the voltage and the reactive current of a first power frequency period, and carrying out average processing on the voltage and the reactive current of at least one power frequency period after the switching event occurs to obtain the voltage and the reactive current of a second power frequency period;
and generating the initial V-I track according to the voltage and the reactive current of the first power frequency period and the voltage and the reactive current of the second power frequency period.
6. A V-I trajectory generation apparatus for non-intrusive load recognition, the apparatus comprising:
the device comprises an acquisition module, a switching module and a control module, wherein the acquisition module is used for acquiring an initial V-I track when a switching event is determined to occur, and the initial V-I track is a track of voltage and reactive current changing along with time;
the mapping module is used for acquiring a time derivative of the voltage and a time derivative of the reactive current based on the initial V-I track, mapping the initial V-I track to a color space according to the time derivative of the voltage and the time derivative of the reactive current to obtain a color V-I track, and specifically, for a track point on the initial V-I track, determining an R value, a G value and a B value of the track point according to the time derivative of the voltage and the time derivative of the reactive current, and generating the color V-I track according to the R value, the G value and the B value of the track point; the method is further specifically used for determining that the R value of the trace point is a first value when the time derivative of the voltage is greater than or equal to 0 and the time derivative of the reactive current is greater than or equal to 0, determining that the G value of the trace point is the first value when the time derivative of the voltage is greater than or equal to 0 and the time derivative of the reactive current is less than 0, determining that the B value of the trace point is the first value when the time derivative of the voltage is less than 0 and the time derivative of the reactive current is greater than or equal to 0, and determining that the R value and the B value of the trace point are both the first value when the time derivative of the voltage is less than 0 and the time derivative of the reactive current is less than 0.
7. The V-I trajectory generation device of claim 6, wherein the acquisition module is specifically configured to:
acquiring voltage and reactive current of at least one power frequency period before a switching event occurs, and acquiring voltage and reactive current of at least one power frequency period after the switching event occurs;
carrying out average processing on the voltage and the reactive current of at least one power frequency period before the switching event occurs to obtain the voltage and the reactive current of a first power frequency period, and carrying out average processing on the voltage and the reactive current of at least one power frequency period after the switching event occurs to obtain the voltage and the reactive current of a second power frequency period;
and generating the initial V-I track according to the voltage and the reactive current of the first power frequency period and the voltage and the reactive current of the second power frequency period.
8. A computer-readable storage medium, characterized in that a V-I trajectory generation program is stored thereon, which when executed by a processor implements the V-I trajectory generation method of any one of claims 1 to 5.
9. An electronic device, comprising: a memory, a processor and a V-I trajectory generation program stored on the memory and executable on the processor, the processor implementing the V-I trajectory generation method of any one of claims 1-5 when executing the program.
10. A convolutional neural network for performing feature extraction on a color V-I trajectory generated based on the V-I trajectory generation method of any one of claims 1-5, the network comprising:
the multi-level residual error network is used for performing multi-level feature extraction on the input image to obtain feature maps of different levels, and splicing the feature maps of the different levels to obtain M1 feature maps;
the reprocessing unit is used for carrying out random uniform rearrangement, pooling and splicing treatment on the M1 characteristic graphs to obtain M2 characteristic graphs;
and the depth separable convolution unit is used for performing depth separable convolution processing on the M2 feature maps to obtain M3 feature maps, wherein M1, M2 and M3 are all positive integers.
11. The convolutional neural network of claim 10, wherein the multi-level residual network comprises:
the input end of the 1 st convolution unit is used as the input end of the convolution neural network, the input end of the (i + 1) th convolution unit is connected with the first output end of the ith convolution unit, and the N convolution units are used for performing multi-level feature extraction on the input image to obtain feature maps of different levels;
the input end of the ith first maximum pooling layer is connected with the second output end of the ith convolution unit and is used for performing maximum pooling processing on the feature map of the corresponding convolution unit, N is an integer which is more than or equal to 2, and i is more than or equal to 1 and less than N;
and the input end of the first splicing layer is respectively connected with the output end of the Nth convolution unit and the output end of each first maximum pooling layer in the N-1 first maximum pooling layers, and the first splicing layer is used for splicing the feature map output by the Nth convolution unit and the feature map subjected to maximum pooling to obtain the M1 feature maps.
12. The convolutional neural network of claim 11, wherein each of the N convolutional units comprises: the device comprises a convolution layer, a first normalization processing layer and a ReLU activation layer.
13. The convolutional neural network of claim 11, wherein the pooling kernel size of each of the N-1 first maximum pooling layers is different.
14. The convolutional neural network of claim 10, wherein the reprocessing unit comprises:
the input end of the shuffling layer is connected with the multi-level residual error network and is used for carrying out shuffling processing on the M1 characteristic graphs;
the input end of the splitting layer is connected with the output end of the shuffling layer, the splitting layer comprises K output ends and is used for carrying out uniform random splitting processing on the M1 characteristic diagrams after shuffling processing to obtain K groups of characteristic diagrams, and the K groups of characteristic diagrams are output through the K output ends, and K is an integer greater than or equal to 2;
the spatial pyramid pooling layers comprise K-1 second maximum pooling layers, and input ends of the K-1 second maximum pooling layers are correspondingly connected with the 1 st to the K-1 st output ends in the segmentation layers and are used for performing maximum pooling on the feature maps output by the 1 st to the K-1 st output ends;
and the input end of the second splicing layer is respectively connected with the Kth output end of the segmentation layer and the output end of each second maximum pooling layer of the K-1 second maximum pooling layers, and the second splicing layer is used for splicing the characteristic diagram output by the Kth output end and the characteristic diagram after maximum pooling to obtain the M2 characteristic diagrams.
15. The convolutional neural network of claim 14, wherein the pooling kernel size of each of the K-1 second largest pooling layers is different.
16. The convolutional neural network of claim 10, wherein the deep separable convolution unit comprises:
the input end of the channel-by-channel convolution layer is connected with the reprocessing unit and is used for carrying out channel-by-channel convolution processing on the M2 feature maps;
the input end of the point-by-point convolution layer is connected with the output end of the channel-by-channel convolution layer and is used for performing point-by-point convolution processing on the feature map subjected to the channel-by-channel convolution processing;
the input end of the second normalization processing layer is connected with the output end of the point-by-point convolution layer and is used for performing normalization processing on the feature map subjected to the point-by-point convolution processing;
and the input end of the ReLU activation layer is connected with the output end of the second normalization processing layer and is used for performing activation processing on the feature maps after normalization processing to obtain the M3 feature maps.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN202210255087.1A CN114330461B (en) | 2022-03-16 | 2022-03-16 | V-I track generation method and device for non-invasive load identification and neural network |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN202210255087.1A CN114330461B (en) | 2022-03-16 | 2022-03-16 | V-I track generation method and device for non-invasive load identification and neural network |
Publications (2)
Publication Number | Publication Date |
---|---|
CN114330461A CN114330461A (en) | 2022-04-12 |
CN114330461B true CN114330461B (en) | 2022-06-24 |
Family
ID=81033017
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN202210255087.1A Active CN114330461B (en) | 2022-03-16 | 2022-03-16 | V-I track generation method and device for non-invasive load identification and neural network |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN114330461B (en) |
Families Citing this family (1)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN116359602B (en) * | 2023-03-07 | 2024-05-03 | 北京智芯微电子科技有限公司 | Non-invasive electric vehicle charging identification method, device, medium and intelligent ammeter |
Family Cites Families (5)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
GB2570890B (en) * | 2018-02-07 | 2020-05-06 | Green Running Ltd | Method and apparatus for monitoring electrical power consumption |
CN112418722B (en) * | 2020-12-08 | 2021-09-10 | 浙江大学 | Non-invasive load identification method based on V-I (velocity-amplitude) trajectory graph and neural network |
CN113036759B (en) * | 2021-03-31 | 2023-09-26 | 上海电力大学 | Fine granularity identification method and identification system for power consumer load |
CN113537330B (en) * | 2021-07-08 | 2024-05-14 | 中国海洋大学 | Non-invasive load identification method |
CN113901973A (en) * | 2021-09-29 | 2022-01-07 | 北京智芯微电子科技有限公司 | Power load identification method and device, storage medium and chip equipment |
-
2022
- 2022-03-16 CN CN202210255087.1A patent/CN114330461B/en active Active
Also Published As
Publication number | Publication date |
---|---|
CN114330461A (en) | 2022-04-12 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN110108992B (en) | Cable partial discharge fault identification method and system based on improved random forest algorithm | |
CN111650453A (en) | Power equipment diagnosis method and system based on windowing characteristic Hilbert imaging | |
CN114742097B (en) | Optimization method for automatically determining variation modal decomposition parameters based on bearing vibration signals | |
CN114330461B (en) | V-I track generation method and device for non-invasive load identification and neural network | |
CN115166563B (en) | Power battery aging state evaluation and retirement screening method and system | |
CN117674198A (en) | Novel frequency modulation method and system for super-capacitor coupled lithium battery | |
CN104569747A (en) | System and method for checking insulativity of power-off cable | |
CN117691631A (en) | Electric power frequency modulation method and system based on hybrid energy storage device | |
CN117748501B (en) | Wind power prediction method and system for energy storage auxiliary black start | |
CN115423013A (en) | Power system operation mode classification method, device, equipment and medium | |
CN110909302A (en) | Method and system for learning local disturbance characteristics of operating state parameters of alternating-current and direct-current power grid | |
CN117825998A (en) | Lithium ion battery remaining life prediction method, system, equipment and medium | |
CN114881225A (en) | Power transmission and transformation inspection model network structure searching method, system and storage medium | |
CN113299298A (en) | Residual error unit, network and target identification method, system, device and medium | |
CN113010985A (en) | Non-invasive load identification method based on parallel AANN | |
CN117527495A (en) | Modulation mode identification method and device for wireless communication signals | |
CN116298725A (en) | Fault arc detection method, system and storage medium | |
CN116381424A (en) | Shore power box insulation performance abnormality detection method based on multi-source data fusion | |
CN114332443B (en) | Non-intrusive load recognition device, method, storage medium, and electronic apparatus | |
CN117572161A (en) | Arc fault detection method and device and related components | |
CN112731080A (en) | Method for diagnosing rapid development type partial discharge oil paper insulation degradation state | |
CN112701683A (en) | Non-invasive load detection and decomposition method, system and storage medium | |
CN113673471B (en) | Transformer winding vibration signal feature extraction method | |
CN117092554B (en) | Inverter coupling fault analysis method and device, electronic equipment and storage medium | |
CN117390508B (en) | Hydroelectric generating set signal state identification method based on time-shifting multi-scale cosine similarity entropy |
Legal Events
Date | Code | Title | Description |
---|---|---|---|
PB01 | Publication | ||
PB01 | Publication | ||
SE01 | Entry into force of request for substantive examination | ||
SE01 | Entry into force of request for substantive examination | ||
GR01 | Patent grant | ||
GR01 | Patent grant |