Electricity load identification method and device, electronic equipment and storage medium
By decomposing the AC signal and generating feature maps, and combining a dual recognition method with multiple recognition models, the problem of low accuracy in identifying malicious loads in existing technologies has been solved, achieving higher recognition accuracy.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- LIANGYUN SMART ENERGY (GUANGDONG) CO LTD
- Filing Date
- 2024-10-18
- Publication Date
- 2026-04-21
AI Technical Summary
In existing technologies, the instantaneous power increase method is prone to misidentifying normally used household appliances as malicious loads when identifying electrical loads, resulting in low identification accuracy.
By decomposing the AC signal of the target line, voltage and current trajectory diagrams and frequency domain feature diagrams are generated. Multiple identification models (first identification model, second identification model and third identification model) are used to identify from multiple dimensions. Combined with a logistic regression model, it is determined whether there is a malicious load.
It improves the accuracy of malicious payload identification, avoids false positives, and enhances the accuracy of identification results.
Smart Images
Figure CN121901671A_ABST
Abstract
Description
Technical Field
[0001] This invention relates to the field of power facility technology, and more specifically, to a method, apparatus, electronic device, and storage medium for identifying electrical loads. Background Technology
[0002] As the demand for electricity increases, so too does the number of disasters caused by improper electricity use. For example, charging electric vehicles in densely populated areas or using high-power electrical appliances such as induction cookers, rice cookers, and hair dryers in student dormitories. Therefore, to reduce fire safety accidents, it is necessary to promptly identify and prevent "malicious loads" from being connected to the power grid in certain situations to avoid fires.
[0003] In related technologies, the main methods for identifying "malicious loads" include the instantaneous power increase method. Specifically, by measuring the instantaneous power in the line in real time, when a sudden and significant increase in the instantaneous power in the line is detected, and the change in the instantaneous power value is greater than a preset threshold, the load connected to the line is determined to be a "malicious load".
[0004] However, the instantaneous power increase method has a limitation: due to its principle, appliances with high instantaneous power cannot be present in the circuit. That is, when there are normally used loads with high instantaneous power in the circuit, such as household appliances like air conditioners and refrigerators, these normally used loads will be misjudged as malicious loads, resulting in low accuracy of the identification results. Summary of the Invention
[0005] The purpose of this invention is to address the shortcomings of the prior art by providing a method, apparatus, electronic device, and storage medium for identifying electrical loads, so as to improve the accuracy of identifying malicious loads.
[0006] To achieve the above objectives, the technical solutions adopted in the embodiments of this application are as follows:
[0007] In a first aspect, embodiments of this application provide a method for identifying electrical loads, the method comprising:
[0008] Acquire the AC signal of the target line to be identified over multiple cycles;
[0009] The AC signal within each of the aforementioned cycles is decomposed to obtain the decomposed signal corresponding to each of the aforementioned cycles.
[0010] Based on the decomposed signal corresponding to each period, a feature map corresponding to each period is generated, the feature map including: voltage and current trajectory map and frequency domain feature map;
[0011] Based on the feature maps corresponding to each cycle and multiple pre-trained recognition models, the recognition results corresponding to each cycle are obtained. The multiple recognition models include: a first recognition model, a second recognition model, and a third recognition model. The first recognition model is used to identify whether there is a malicious load in each cycle based on the voltage and current trajectory maps of each cycle. The second recognition model is used to identify whether there is a malicious load in each cycle based on the frequency domain feature maps of each cycle. The third recognition model is used to obtain the recognition results corresponding to each cycle based on the recognition results of the first recognition model and the recognition results of the second recognition model in each cycle.
[0012] Based on the identification results corresponding to each cycle, it is determined whether there is a malicious load on the target line.
[0013] Optionally, the alternating current signal includes: a current signal and a voltage signal;
[0014] The process of decomposing the AC signal within each of the aforementioned periods to obtain the decomposed signal corresponding to each of the aforementioned periods includes:
[0015] The current signal and voltage signal within the period are respectively subjected to a first decomposition process to obtain multiple sub-current signals and multiple sub-voltage signals corresponding to the period.
[0016] Correlation analysis is performed on the multiple sub-current signals and the current signal corresponding to the period, and a set of sub-current signals with a correlation greater than a first threshold is obtained.
[0017] Correlation analysis is performed on the multiple sub-voltage signals and the voltage signal within the period, and a set of sub-voltage signals with a correlation greater than the second threshold within the period is obtained.
[0018] The sub-current signal set is subjected to a second decomposition process to obtain multiple single-frequency current signals, and the sub-voltage signal set is subjected to a second decomposition process to obtain multiple single-frequency voltage signals. The multiple single-frequency current signals and the multiple single-frequency voltage signals are used as the decomposition signals corresponding to the period.
[0019] Optionally, generating a feature map corresponding to each of the periods based on the decomposed signals corresponding to each period includes:
[0020] Using a single-frequency current signal and a single-frequency voltage signal corresponding to the same moment as a set of coordinate points, the multiple single-frequency current signals and multiple single-frequency voltage signals corresponding to the period are divided into multiple sets of coordinate points, and a voltage and current trajectory diagram corresponding to the period is generated based on the multiple sets of coordinate points, wherein the horizontal and vertical axes of the voltage and current trajectory diagram represent current and the vertical axis represents voltage.
[0021] Fourier transforms are performed on the multiple single-frequency current signals corresponding to the period to obtain the amplitude of the multiple single-frequency current signals. Based on the multiple single-frequency current signals and their amplitudes, a frequency domain feature map corresponding to the period is generated, wherein the horizontal and vertical axes of the frequency domain feature map represent frequency and the vertical axis represents amplitude.
[0022] Optionally, the first recognition model is trained based on the following steps:
[0023] A first training sample set is constructed based on multiple sample voltage and current trajectory diagrams corresponding to malignant loads and multiple sample voltage and current trajectory diagrams corresponding to non-malignant loads.
[0024] The initial first classifier is iteratively trained based on the first training sample set until the latest iteration of the first classifier satisfies the preset convergence condition. The first classifier that satisfies the preset convergence condition is used as the first recognition model.
[0025] The second recognition model is trained based on the following steps:
[0026] The second training sample set is constructed based on the frequency domain feature maps of multiple samples corresponding to malignant loads and multiple samples corresponding to non-malignant loads.
[0027] The initial second classifier is iteratively trained based on the second training sample set until the latest iteration of the second classifier satisfies the preset convergence condition. The second classifier that satisfies the preset convergence condition is used as the second recognition model.
[0028] Optionally, the third recognition model is trained based on the following steps:
[0029] Based on the multiple first recognition results output by the first recognition model, the labels corresponding to each first recognition result, the multiple second recognition results output by the second recognition model, and the labels corresponding to each second recognition result, the first regression parameters and the second regression parameters of the logistic regression model are determined.
[0030] Substituting the first regression parameter and the second regression parameter into the logistic regression model, the third identification model is obtained.
[0031] Optionally, obtaining the recognition result corresponding to each of the periods based on the feature maps corresponding to each period and multiple pre-trained recognition models includes:
[0032] The voltage and current trajectory diagram corresponding to the cycle is input into the first identification model to obtain the first identification result output by the first identification model. The first identification result is used to indicate whether there is a malicious load.
[0033] The frequency domain feature map corresponding to the period is input into the second identification model to obtain the second identification result output by the second identification model. The second identification result is used to indicate whether there is a malicious load.
[0034] The third recognition model determines the first product result of the first recognition result and the first regression parameter, and determines the second product result of the second recognition result and the second regression parameter, and determines the summation result of the first product result and the second product result, and determines the recognition result corresponding to the period based on the summation result and a preset threshold.
[0035] Optionally, determining whether there is malicious load on the target line based on the identification results corresponding to each of the said periods includes:
[0036] If the identification result corresponding to at least one cycle is consistent with the preset result, then it is determined that there is a malicious load on the target line.
[0037] Secondly, embodiments of this application also provide an electrical load identification device, the device comprising:
[0038] The acquisition module is used to acquire the AC signal of the target line to be identified over multiple cycles.
[0039] The decomposition module is used to decompose the AC signal in each of the said cycles to obtain the decomposed signal corresponding to each of the said cycles.
[0040] The generation module is used to generate a feature map corresponding to each of the periods based on the decomposed signals corresponding to each period. The feature map includes a voltage and current trajectory map and a frequency domain feature map.
[0041] The identification module is used to obtain the identification result corresponding to each of the cycles based on the feature map corresponding to each cycle and multiple identification models pre-trained. The multiple identification models include: a first identification model, a second identification model and a third identification model. The first identification model is used to identify whether there is a malicious load in each cycle based on the voltage and current trajectory map of each cycle. The second identification model is used to identify whether there is a malicious load in each cycle based on the frequency domain feature map of each cycle. The third identification model is used to obtain the identification result corresponding to each cycle based on the identification results of the first identification model in each cycle and the identification results of the second identification model in each cycle.
[0042] The determination module is used to determine whether there is a malicious load on the target line based on the identification results corresponding to each of the said cycles.
[0043] Optionally, the alternating current signal includes: a current signal and a voltage signal;
[0044] The decomposition module is specifically used for:
[0045] The current signal and voltage signal within the period are respectively subjected to a first decomposition process to obtain multiple sub-current signals and multiple sub-voltage signals corresponding to the period.
[0046] Correlation analysis is performed on the multiple sub-current signals and the current signal corresponding to the period, and a set of sub-current signals with a correlation greater than a first threshold is obtained.
[0047] Correlation analysis is performed on the multiple sub-voltage signals and the voltage signal within the period, and a set of sub-voltage signals with a correlation greater than the second threshold within the period is obtained.
[0048] The sub-current signal set is subjected to a second decomposition process to obtain multiple single-frequency current signals, and the sub-voltage signal set is subjected to a second decomposition process to obtain multiple single-frequency voltage signals. The multiple single-frequency current signals and the multiple single-frequency voltage signals are used as the decomposition signals corresponding to the period.
[0049] Optionally, the generation module is specifically used for:
[0050] Using a single-frequency current signal and a single-frequency voltage signal corresponding to the same moment as a set of coordinate points, the multiple single-frequency current signals and multiple single-frequency voltage signals corresponding to the period are divided into multiple sets of coordinate points, and a voltage and current trajectory diagram corresponding to the period is generated based on the multiple sets of coordinate points, wherein the horizontal and vertical axes of the voltage and current trajectory diagram represent current and the vertical axis represents voltage.
[0051] Fourier transforms are performed on the multiple single-frequency current signals corresponding to the period to obtain the amplitude of the multiple single-frequency current signals. Based on the multiple single-frequency current signals and their amplitudes, a frequency domain feature map corresponding to the period is generated, wherein the horizontal and vertical axes of the frequency domain feature map represent frequency and the vertical axis represents amplitude.
[0052] Optionally, the device further includes:
[0053] The training module is used for:
[0054] A first training sample set is constructed based on multiple sample voltage and current trajectory diagrams corresponding to malignant loads and multiple sample voltage and current trajectory diagrams corresponding to non-malignant loads.
[0055] The initial first classifier is iteratively trained based on the first training sample set until the latest iteration of the first classifier satisfies the preset convergence condition. The first classifier that satisfies the preset convergence condition is used as the first recognition model.
[0056] The second recognition model is trained based on the following steps:
[0057] The second training sample set is constructed based on the frequency domain feature maps of multiple samples corresponding to malignant loads and multiple samples corresponding to non-malignant loads.
[0058] The initial second classifier is iteratively trained based on the second training sample set until the latest iteration of the second classifier satisfies the preset convergence condition. The second classifier that satisfies the preset convergence condition is used as the second recognition model.
[0059] Optionally, the training module is also used for:
[0060] Based on the multiple first recognition results output by the first recognition model, the labels corresponding to each first recognition result, the multiple second recognition results output by the second recognition model, and the labels corresponding to each second recognition result, the first regression parameters and the second regression parameters of the logistic regression model are determined.
[0061] Substituting the first regression parameter and the second regression parameter into the logistic regression model, the third identification model is obtained.
[0062] Optionally, the identification module is specifically used for:
[0063] The voltage and current trajectory diagram corresponding to the cycle is input into the first identification model to obtain the first identification result output by the first identification model. The first identification result is used to indicate whether there is a malicious load.
[0064] The frequency domain feature map corresponding to the period is input into the second identification model to obtain the second identification result output by the second identification model. The second identification result is used to indicate whether there is a malicious load.
[0065] The third recognition model determines the first product result of the first recognition result and the first regression parameter, and determines the second product result of the second recognition result and the second regression parameter, and determines the summation result of the first product result and the second product result, and determines the recognition result corresponding to the period based on the summation result and a preset threshold.
[0066] Optionally, the determining module is specifically used for:
[0067] If the identification result corresponding to at least one cycle is consistent with the preset result, then it is determined that there is a malicious load on the target line.
[0068] Thirdly, embodiments of this application also provide an electronic device, including: a processor, a storage medium, and a bus, wherein the storage medium stores machine-readable instructions executable by the processor, and when the electronic device is running, the processor communicates with the storage medium via the bus, and the processor executes the machine-readable instructions to perform the steps of the method provided in the first aspect above.
[0069] Fourthly, embodiments of this application also provide a computer-readable storage medium storing a computer program, which, when executed by a processor, performs the method described in the first aspect above.
[0070] The beneficial effects of this application are:
[0071] This application provides a method, apparatus, electronic device, and storage medium for identifying electrical loads. Considering that multiple electrical devices may exist simultaneously on a target line, this solution proposes to decompose the AC signal of the target line over multiple cycles to obtain the decomposed signal corresponding to each cycle. This decomposes a complex multi-component signal into multiple single-component signals, transforming the identification of mixed signals in the application scenario into a single-signal identification problem, thus improving the accuracy of the identification results. Then, based on the decomposed signals corresponding to each cycle, feature maps of each cycle in different dimensions are generated, enabling the identification of malicious loads using image recognition methods to improve identification accuracy. Next, based on the feature maps corresponding to each cycle and multiple pre-trained identification models, the identification results corresponding to each cycle are obtained. This involves dual identification from multiple dimensions based on multiple models to determine whether a malicious load exists, improving the accuracy of malicious load identification. Finally, based on the identification results corresponding to each cycle, the existence of a malicious load on the target line is determined, avoiding the possibility of misjudgment due to relying solely on the identification results of a single cycle. This improves the accuracy of the identification results and solves the problem of low identification accuracy in existing technologies. Attached Figure Description
[0072] To more clearly illustrate the technical solutions of the embodiments of the present invention, the accompanying drawings used in the embodiments will be briefly introduced below. It should be understood that the following drawings only show some embodiments of the present invention and should not be regarded as a limitation on the scope. For those skilled in the art, other related drawings can be obtained based on these drawings without creative effort.
[0073] Figure 1This is a schematic diagram of the structure of an electronic device provided in an embodiment of this application;
[0074] Figure 2 A flowchart illustrating a method for identifying electrical loads provided in an embodiment of this application;
[0075] Figure 3 A flowchart illustrating another method for identifying electrical loads provided in an embodiment of this application;
[0076] Figure 4 A flowchart illustrating another method for identifying electrical loads provided in this application embodiment;
[0077] Figure 5 A flowchart illustrating another method for identifying electrical loads provided in an embodiment of this application;
[0078] Figure 6 A flowchart illustrating another method for identifying electrical loads provided in this application embodiment;
[0079] Figure 7 A flowchart illustrating another method for identifying electrical loads provided in an embodiment of this application;
[0080] Figure 8 A flowchart illustrating another method for identifying electrical loads provided in this application embodiment;
[0081] Figure 9 This is a schematic diagram of the structure of an electrical load identification device provided in an embodiment of this application. Detailed Implementation
[0082] To make the objectives, technical solutions, and advantages of the embodiments of this application clearer, the technical solutions of the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. It should be understood that the drawings in this application are for illustrative and descriptive purposes only and are not intended to limit the scope of protection of this application. Furthermore, it should be understood that the schematic drawings are not drawn to scale. The flowcharts used in this application illustrate operations implemented according to some embodiments of this application. It should be understood that the operations in the flowcharts may not be implemented in sequence, and steps without logical contextual relationships may be reversed or implemented simultaneously. In addition, those skilled in the art, guided by the content of this application, may add one or more other operations to the flowcharts, or remove one or more operations from the flowcharts.
[0083] Furthermore, the described embodiments are merely some, not all, of the embodiments of this application. The components of the embodiments of this application described and illustrated herein can typically be arranged and designed in various different configurations. Therefore, the following detailed description of the embodiments of this application provided in the accompanying drawings is not intended to limit the scope of the claimed application, but merely to illustrate selected embodiments of the application. All other embodiments obtained by those skilled in the art based on the embodiments of this application without inventive effort are within the scope of protection of this application.
[0084] It should be noted that the term "comprising" will be used in the embodiments of this application to indicate the presence of the features declared thereafter, but does not exclude the addition of other features.
[0085] Figure 1 This is a schematic diagram of the structure of an electronic device provided in an embodiment of this application; the electronic device may be a computer, server, or other processing device with processing capabilities, used to implement the electrical load identification method provided in this application. Figure 1 As shown, the electronic device includes: a memory 101 and a processor 102.
[0086] The memory 101 and the processor 102 are electrically connected directly or indirectly to enable data transmission or interaction. For example, the electrical connection can be achieved through one or more communication buses or signal lines.
[0087] The memory 101 may be, but is not limited to, random access memory (RAM), read-only memory (ROM), programmable read-only memory (PROM), erasable programmable read-only memory (EPROM), electrically erasable programmable read-only memory (EEPROM), etc.
[0088] Processor 102 can be an integrated circuit chip with signal processing capabilities. The aforementioned processor 102 can be a general-purpose processor, including a Central Processing Unit (CPU), a Network Processor (NP), etc. It can implement or execute the methods, steps, and logic block diagrams disclosed in the embodiments of this invention. The general-purpose processor can be a microprocessor or any conventional processor.
[0089] Understandable. Figure 1 The structure described is for illustrative purposes only; the electronic device may also include components that are more advanced than those described above. Figure 1 The more or fewer components shown, or having the same Figure 1 The different configurations shown. Figure 1 The components shown can be implemented using hardware, software, or a combination thereof.
[0090] The memory 101 is used to store programs, and the processor 102 calls the programs stored in the memory 101 to execute the electrical load identification method provided in the following embodiments.
[0091] The following will illustrate the method for identifying electrical loads provided in this application and the corresponding beneficial effects through multiple embodiments.
[0092] Figure 2 This is a flowchart illustrating a method for identifying electrical loads provided in an embodiment of this application. Optionally, the execution entity of this method may be... Figure 1 The electronic device shown has data processing capabilities.
[0093] It should be understood that in other embodiments, the order of some steps in the electrical load identification method can be interchanged according to actual needs, or some steps can be omitted or deleted. For example... Figure 2 As shown, the method includes:
[0094] S201. Acquire the AC signal of the target line to be identified in multiple cycles.
[0095] The target line can be any line to be identified in the power grid. For example, the target line could be the power line in a student dormitory or the power line in a residential building.
[0096] For example, by using a 1-second cycle, one can obtain the AC signal of the target line per second, such as the effective value of voltage, the effective value of current, active power, reactive power, apparent power, power factor, power factor angle, and harmonic power, etc.
[0097] Optionally, an AC signal with at least m cycles is acquired. To ensure the recognition effect, m>=5. At the same time, to ensure the recognition speed, m should not be too large. The value of m can be determined according to the specific situation.
[0098] In one feasible approach, for example, an electrical signal acquisition device installed on the target line can periodically acquire the electrical signal of the target line, and the acquired electrical signal of the target line over multiple periods can be transmitted to the aforementioned... Figure 1The electronic equipment in the circuit is designed to identify whether there is a malicious load on the power line 1 based on the electrical signals of the target line over multiple cycles.
[0099] Among them, toxic loads, also known as resistive loads, generally refer to purely resistive direct-heating electrical appliances. For example, in student dormitories, toxic loads include appliances such as soldering irons, hair dryers, and electric stoves; in residential buildings, toxic loads include electric vehicles. That is, the type of toxic load to be identified varies depending on the situation.
[0100] S202. Decompose the AC signal in each cycle to obtain the decomposed signal corresponding to each cycle.
[0101] In one feasible approach, for example, in a student dormitory, a computer and a rice cooker are simultaneously connected to power line 1. In this case, with multiple electrical devices on the same power line, the AC signal obtained from power line 1 is a mixed signal. Therefore, a pre-defined decomposition algorithm (such as empirical mode decomposition or local mean decomposition) can be used to decompose the AC signal within each cycle, obtaining the decomposed signal corresponding to each cycle. This decomposes a complex multi-component signal into multiple single-component signals, thus transforming the identification of mixed signals in the application scenario into a single-signal identification problem, improving the accuracy of the identification results.
[0102] S203. Generate the feature map corresponding to each period based on the decomposed signal corresponding to each period.
[0103] The feature maps include voltage and current trajectory maps and frequency domain feature maps.
[0104] Optionally, in this embodiment, considering that directly identifying the existence of malignant loads based on the decomposed signals corresponding to each period could easily overlook some potential key information, leading to low accuracy of the identification results, it is proposed that feature maps corresponding to each period be generated based on the decomposed signals corresponding to each period. This yields images of each period in different dimensions, enabling the identification of malignant loads to be solved using image recognition methods, thereby improving identification accuracy.
[0105] S204. Based on the feature maps corresponding to each period and the multiple recognition models obtained in the pre-training, the recognition results corresponding to each period are obtained.
[0106] The multiple identification models include: a first identification model, a second identification model, and a third identification model. The first identification model is used to identify whether there is a malicious load based on the voltage and current trajectory diagram. The second identification model is used to identify whether there is a malicious load based on the frequency domain feature diagram. The third identification model is used to obtain the identification results corresponding to each cycle based on the identification results of the first identification model and the identification results of the second identification model.
[0107] In this embodiment, a first identification model determines whether a malicious load exists on the target line within each cycle based on the voltage and current trajectory diagrams corresponding to each cycle, thus obtaining the identification result of the first identification model for each cycle. A second identification model, based on the frequency domain characteristics corresponding to each cycle and the second identification model, determines whether a malicious load exists on the target line within each cycle, thus obtaining the identification result of the second identification model for each cycle. Then, a third identification model, based on the identification results of the first and second identification models for each cycle, obtains the corresponding identification result for each cycle. Therefore, this solution uses multiple models to perform dual identification from multiple dimensions to determine whether a malicious load exists, improving the accuracy of malicious load identification.
[0108] S205. Based on the identification results corresponding to each cycle, determine whether there is a malicious load on the target line.
[0109] Optionally, in one feasible approach, for example, a logical truth table is pre-constructed, recording the mapping relationship between the identification results for each period and the presence of malicious load on the target line. Then, after determining the identification results for each period, the presence of malicious load on the target line can be determined by querying the logical truth table. For example, if the identification result for the first period is 1, the identification result for the second period is 0, and the identification result for the third period is 0, then querying the logical truth table can determine that malicious load exists on the target line. Here, 0 represents the absence of malicious load, and 1 represents the presence of malicious load. Therefore, in this scheme, the presence of malicious load on the target line is determined based on the identification results for each period, avoiding misjudgments and improving the accuracy of the identification results.
[0110] Optionally, if there is a malicious load on the target line, the target line is disconnected from the malicious load to prevent the malicious load from entering the target line, thereby avoiding a fire.
[0111] In summary, this application provides a method for identifying electrical loads. Considering that multiple electrical devices may exist simultaneously on a target line, this method proposes to decompose the AC signal of the target line over multiple cycles to obtain the decomposed signal corresponding to each cycle. This decomposes a complex multi-component signal into multiple single-component signals, transforming the identification of mixed signals in the application scenario into a single-signal identification problem, thus improving the accuracy of the identification results. Then, based on the decomposed signals corresponding to each cycle, feature maps of each cycle in different dimensions are generated, enabling the identification of malicious loads using image recognition methods to improve identification accuracy. Next, based on the feature maps corresponding to each cycle and multiple pre-trained identification models, the identification results corresponding to each cycle are obtained. This involves dual identification from multiple dimensions based on multiple models to determine whether a malicious load exists, improving the accuracy of malicious load identification. Finally, based on the identification results corresponding to each cycle, it is determined whether a malicious load exists on the target line, avoiding the possibility of misjudgment due to relying solely on the identification results of a single cycle. This improves the accuracy of the identification results and solves the problem of low identification accuracy in existing technologies.
[0112] Optionally, the alternating current signal includes: a current signal and a voltage signal;
[0113] refer to Figure 3 As shown, step S202 above includes:
[0114] S301. Perform the first decomposition process on the current signal and voltage signal within the period respectively to obtain multiple sub-current signals and multiple sub-voltage signals corresponding to the period.
[0115] For example, the first decomposition process can be Empirical Mode Decomposition (EMD), and the second decomposition process can be Independent Component Analysis (ICA). That is, the first decomposition process and the second decomposition process are two different processing and analysis methods. In this way, the corresponding electrical signals in each period can be decomposed based on different decomposition methods to obtain individual signals and noise signals.
[0116] In this embodiment, taking the first cycle as an example, the current signal within the first cycle is decomposed using EMD to obtain multiple sub-current signals corresponding to the first cycle; and the voltage signal within the first cycle is also decomposed using EMD to obtain multiple sub-voltage signals corresponding to the first cycle. Therefore, through EMD decomposition, noise signals in the corresponding electrical signals of each cycle can be removed, resulting in multiple single electrical signals corresponding to each cycle.
[0117] S302. Perform correlation analysis on the multiple sub-current signals and the current signal corresponding to each period, and select the set of sub-current signals with a correlation greater than the first threshold within the period; and perform correlation analysis on the multiple sub-voltage signals and the voltage signal corresponding to each period, and select the set of sub-voltage signals with a correlation greater than the second threshold within the period.
[0118] Optionally, correlation analysis is performed on multiple sub-current signals and the current signal corresponding to the period to obtain the correlation between each sub-current signal and the current signal. The correlation between each sub-current signal and the current signal is compared with a first threshold, and multiple highly correlated sub-current signals with a correlation greater than the first threshold are selected. These multiple highly correlated sub-current signals are then used as a set of sub-current signals with a correlation greater than the first threshold within the period. Similarly, correlation analysis is performed on multiple sub-voltage signals and the voltage signal corresponding to the period to obtain the correlation between each sub-voltage signal and the voltage signal. The correlation between each sub-voltage signal and the voltage signal is then compared with a first threshold, and multiple highly correlated sub-voltage signals with a correlation greater than a second threshold are selected. These multiple highly correlated sub-voltage signals are then used as a set of sub-voltage signals with a correlation greater than the first threshold within the period. Therefore, sub-current signals and sub-voltage signals with low correlation within the period are removed.
[0119] S303. Perform a second decomposition process on the sub-current signal set to obtain multiple single-frequency current signals, and perform a second decomposition process on the sub-voltage signal set to obtain multiple single-frequency voltage signals, and use the multiple single-frequency current signals and multiple single-frequency voltage signals as decomposition signals corresponding to the period.
[0120] Optionally, the obtained sub-current signal set can be further decomposed using ICA to obtain multiple single-frequency current signals; and the obtained sub-voltage signal set can be decomposed using ICA to obtain multiple single-frequency voltage signals, that is, each sub-current signal and each sub-voltage signal can be decomposed into multiple independent frequency component signals; therefore, the multiple single-frequency current signals and multiple single-frequency voltage signals are used as the decomposed signals corresponding to the first cycle. Thus, the decomposition processing of the corresponding electrical signals in each cycle is achieved, resulting in multiple single signals.
[0121] Optionally, refer to Figure 4 As shown, step S203 above includes:
[0122] S401. Using the single-frequency current signal and single-frequency voltage signal corresponding to the same moment as a set of coordinate points, divide the multiple single-frequency current signals and multiple single-frequency voltage signals corresponding to the period into multiple sets of coordinate points, and generate the voltage and current trajectory diagram corresponding to the period based on the multiple sets of coordinate points.
[0123] In the voltage and current trajectory diagram, the horizontal axis represents current and the vertical axis represents voltage.
[0124] It should be noted that the voltage and current trajectory diagram is a characteristic diagram in the time domain. Therefore, in this scheme, time is used as the correlation object. A single-frequency current signal and a single-frequency voltage signal corresponding to the same moment are used as a set of coordinate points. The single-frequency current signals and single-frequency voltage signals corresponding to multiple moments within the first cycle are correlated to obtain multiple sets of coordinate points. Then, with current as the horizontal and vertical axis and voltage as the vertical axis, these multiple sets of coordinate points are plotted on the XOY plane, thus generating the voltage and current trajectory diagram corresponding to the first cycle. Therefore, the voltage and current trajectory diagrams corresponding to each cycle can be obtained.
[0125] S402. Perform Fourier transform on the multiple single-frequency current signals corresponding to the period to obtain the amplitude of the multiple single-frequency current signals, and generate the frequency domain feature map corresponding to the period based on the multiple single-frequency current signals and their amplitudes.
[0126] In the frequency domain feature graph, the horizontal and vertical axes represent frequency, and the vertical axis represents amplitude.
[0127] Optionally, in this embodiment, a Fourier transform is performed on the multiple single-frequency current signals corresponding to the first cycle to obtain the amplitude of the multiple single-frequency current signals; then, the multiple single-frequency current signals and their amplitudes are plotted on the XOY plane with each single-frequency current signal as the horizontal axis and the amplitude of each single-frequency current signal as the vertical axis, thus generating the frequency domain feature map corresponding to the first cycle.
[0128] Therefore, this approach involves multi-dimensional analysis of multiple single-frequency current signals and multiple single-frequency voltage signals corresponding to each cycle to obtain voltage and current trajectory diagrams and frequency domain feature diagrams for each cycle. This allows for the uncovering of potential key information, thereby improving the accuracy of the identification results.
[0129] The following examples will explain in detail how to train multiple recognition models.
[0130] Optionally, refer to Figure 5 As shown, the first recognition model is trained based on the following steps:
[0131] S501. Construct the first training sample set based on multiple sample voltage and current trajectory diagrams corresponding to malignant loads and multiple sample voltage and current trajectory diagrams corresponding to non-malignant loads.
[0132] S502. Based on the first training sample set, iteratively train the initial first classifier until the latest iteration of the first classifier satisfies the preset convergence condition. The first classifier that satisfies the preset convergence condition is taken as the first recognition model.
[0133] Optionally, in this embodiment, in order to train the first recognition model, it is necessary to first obtain multiple sample voltage and current trajectory diagrams figure1_UI corresponding to malignant loads and multiple sample voltage and current trajectory diagrams figure0_UI corresponding to non-malignant loads. The multiple sample voltage and current trajectory diagrams (i.e., figure1_UI, figure0_UI) and the labels of each sample voltage and current trajectory diagram are used as the first training sample set. Then, an initial first classifier, such as an SVM classifier, is selected, and the multiple sample voltage and current trajectory diagrams included in the first training sample set are input into the initial first classifier. The initial first classifier is iteratively trained multiple times until the latest iteration of the first classifier meets the preset convergence condition. This indicates that the accuracy of the new model trained at this time can no longer be significantly improved, and the iteration can be stopped. At this time, the first classifier that meets the preset convergence condition is used as the first recognition model. If the first classifier trained in the current iteration does not meet the preset convergence condition, the iterative training process continues.
[0134] Therefore, in this scheme, the first recognition model model_ui is obtained through multiple iterations of training based on multiple sample voltage and current trajectory diagrams (i.e., figure1_UI, figure0_UI) and the labels (i.e., y_true) of each sample voltage and current trajectory diagram.
[0135] When there is no load on the target line, it is considered to be unloaded. In this case, unloaded can also be considered to have a non-malicious load, that is, the label of the voltage and current trajectory diagram when unloaded is 0.
[0136] And, reference Figure 6 As shown, the second recognition model is trained based on the following steps:
[0137] S601. Construct a second training sample set based on the frequency domain feature maps of multiple samples corresponding to malignant loads and the frequency domain feature maps of multiple samples corresponding to non-malignant loads.
[0138] S602. Based on the second training sample set, iteratively train the initial second classifier until the latest iteration of the second classifier satisfies the preset convergence condition. The second classifier that satisfies the preset convergence condition is used as the second recognition model.
[0139] Optionally, in this embodiment, in order to train the second recognition model, it is necessary to first obtain multiple sample frequency domain feature maps figure1_fft corresponding to malignant loads and multiple sample frequency domain feature maps figure0_fft corresponding to non-malignant loads. The multiple sample frequency domain feature maps (i.e., figure1_fft, figure0_fft) and the labels of each sample frequency domain feature map are used as the second training sample set. Then, an initial second classifier, such as an SVM classifier, is selected, and the multiple sample frequency domain feature maps included in the second training sample set are input into the initial second classifier. The initial second classifier is iteratively trained multiple times until the latest iteration of the second classifier meets the preset convergence condition. At this time, the second classifier that meets the preset convergence condition is used as the second recognition model. If the second classifier obtained by the current iteration training does not meet the preset convergence condition, the iterative training process continues.
[0140] Therefore, in this scheme, the second recognition model model_fft is obtained through multiple iterations of training based on multiple sample frequency domain feature maps (i.e., figure1_fft, figure0_fft) and the labels (i.e., y_true) of each sample frequency domain feature map.
[0141] Optionally, refer to Figure 7 As shown, the third recognition model is trained based on the following steps:
[0142] S701. Based on the multiple first recognition results output by the first recognition model, the labels corresponding to each first recognition result, the multiple second recognition results output by the second recognition model, and the labels corresponding to each second recognition result, determine the first regression parameters and the second regression parameters of the logistic regression model.
[0143] The label corresponding to the first recognition result y_pred_ui is the label of the sample voltage and current trajectory map input to the first recognition model, and the label corresponding to the second recognition result y_pred_fft is the label of the sample frequency domain feature map input to the second recognition model.
[0144] S702. Substitute the first regression parameter and the second regression parameter into the logistic regression model to obtain the third identification model.
[0145] Optionally, in order to further improve the accuracy of the recognition results, it is proposed that a logistic regression model model_combined, which inherits the discrimination results of the two models, can be constructed based on the multiple first recognition results output by the first recognition model, the labels corresponding to each first recognition result, the multiple second recognition results output by the second recognition model, and the labels corresponding to each second recognition result.
[0146] Specifically, a logistic regression model, model_combined, is constructed, i.e., y_true = w1*y_pred_ui + w2*y_pred_fft. This involves performing least squares fitting on the multiple first recognition results y_pred_ui output by the first recognition model, the multiple second recognition results y_pred_fft output by the second recognition model, and the labels of each recognition result to obtain the first regression parameter w1 and the second regression parameter w2. Then, the first regression parameter w1 and the second regression parameter w2 are substituted into the logistic regression model to obtain the third recognition model.
[0147] Optionally, refer to Figure 8 As shown, step 204 above includes:
[0148] S801. Input the voltage and current trajectory diagram corresponding to the period into the first recognition model to obtain the first recognition result output by the first recognition model.
[0149] The first identification result is used to indicate whether a malicious payload exists. The first identification result includes: 1 or 0, where 1 represents the presence of a malicious payload and 0 represents the absence of a malicious payload.
[0150] S802. Input the frequency domain feature map corresponding to the period into the second recognition model to obtain the second recognition result output by the second recognition model.
[0151] The second identification result is used to indicate whether a malicious payload exists. The second identification result includes: 1 or 0, where 1 represents the presence of a malicious payload and 0 represents the absence of a malicious payload.
[0152] In this embodiment, for example, taking the first cycle as an example, the voltage and current trajectory diagram corresponding to the first cycle is input into the first identification model to obtain the first identification result corresponding to the first cycle output by the first identification model, that is, based on the voltage and current trajectory diagram, it is determined whether there is a malignant load on the target line in each cycle; and the frequency domain feature diagram corresponding to the first cycle is input into the second identification model to obtain the second identification result corresponding to the first cycle output by the second identification model, that is, based on the frequency domain feature diagram, it is determined whether there is a malignant load on the target line in each cycle.
[0153] S803. The third identification model determines the first product result of the first identification result and the first regression parameter, and determines the second product result of the second identification result and the second regression parameter, and determines the summation result of the first product result and the second product result, and determines the identification result corresponding to the period based on the summation result and the preset threshold.
[0154] For example, the preset threshold is 0.6. Optionally, considering that the summation result of the logistic regression model output is a value less than zero, it is proposed that the summation result of the logistic regression model output be compared with the preset threshold. If the summation result is greater than or equal to the preset threshold, the recognition result output by the third recognition model is determined to be 1, that is, there is a malicious load; if the summation result is less than the preset threshold, the recognition result output by the third recognition model is determined to be 0, that is, there is no malicious load.
[0155] In this embodiment, for example, if the first regression parameter w1 is 0.7 and the second regression parameter w2 is 0.3, the first identification result y_pred_ui_1 corresponding to the first period output by the first identification model is 1, and the second identification result y_pred_fft_1 corresponding to the first period output by the second identification model is 0, then y_true_1 = 0.7 * 1 + 0.3 * 0 = 0.7, that is, the summation result is 0.7. Therefore, the summation result is greater than the preset threshold, i.e., 0.7 > 0.6, and the identification result y_pred_1 corresponding to the first period is determined to be 1, meaning that the target line has a malicious load in the first period. Similarly, the above method can also be used to determine whether the target line has a malicious load in other periods.
[0156] Optionally, step S205 above includes:
[0157] If the identification result for at least one cycle matches the preset result, then it is determined that there is a malicious load on the target line.
[0158] For example, the preset result value is 1.
[0159] Optionally, in one feasible approach, if the identification result y_pred_i corresponding to any given period is 1, where i is the period number (i = 1, 2, ..., N), then it is determined that there is a malicious load on the target line; otherwise, it is determined that there is no malicious load. Therefore, in this embodiment, the determination of whether there is a malicious load on the target line is based on the identification results corresponding to multiple periods, avoiding misjudgments and improving the accuracy of the identification results.
[0160] The apparatus and storage medium used to implement the electrical load identification method provided in this application will be described below. The specific implementation process and technical effects are described above and will not be repeated below.
[0161] Optionally, refer to Figure 9 As shown, this is an electrical load identification device provided in an embodiment of this application. The device includes:
[0162] The acquisition module 901 is used to acquire the AC signal of the target line to be identified in multiple cycles;
[0163] The decomposition module 902 is used to decompose the AC signal in each cycle to obtain the decomposed signal corresponding to each cycle.
[0164] The generation module 903 is used to generate feature maps corresponding to each cycle based on the decomposed signals corresponding to each cycle. The feature maps include voltage and current trajectory maps and frequency domain feature maps.
[0165] The identification module 904 is used to obtain the identification result for each cycle based on the feature map corresponding to each cycle and multiple identification models pre-trained. The multiple identification models include: a first identification model, a second identification model and a third identification model. The first identification model is used to identify whether there is a malignant load in each cycle based on the voltage and current trajectory map of each cycle. The second identification model is used to identify whether there is a malignant load in each cycle based on the frequency domain feature map of each cycle. The third identification model is used to obtain the identification result for each cycle based on the identification results of the first identification model and the second identification model in each cycle.
[0166] The determination module 905 is used to determine whether there is a malicious load on the target line based on the identification results corresponding to each cycle.
[0167] Optionally, the alternating current signal includes: a current signal and a voltage signal;
[0168] Decomposition module 902 is specifically used for:
[0169] The current signal and voltage signal within the period are decomposed separately to obtain multiple sub-current signals and multiple sub-voltage signals corresponding to the period.
[0170] Correlation analysis is performed on multiple sub-current signals and the current signal within the period, and a set of sub-current signals with a correlation greater than a first threshold within the period is obtained; and correlation analysis is performed on multiple sub-voltage signals and the voltage signal within the period, and a set of sub-voltage signals with a correlation greater than a second threshold within the period is obtained.
[0171] The sub-current signal set is subjected to a second decomposition process to obtain multiple single-frequency current signals, and the sub-voltage signal set is subjected to a second decomposition process to obtain multiple single-frequency voltage signals. The multiple single-frequency current signals and multiple single-frequency voltage signals are used as decomposed signals corresponding to the period.
[0172] Optionally, module 903 is specifically used for:
[0173] Using the single-frequency current signal and single-frequency voltage signal corresponding to the same moment as a set of coordinate points, the multiple single-frequency current signals and multiple single-frequency voltage signals corresponding to the period are divided into multiple sets of coordinate points, and a voltage and current trajectory diagram corresponding to the period is generated based on the multiple sets of coordinate points. In the voltage and current trajectory diagram, the horizontal axis represents current and the vertical axis represents voltage.
[0174] Fourier transforms are performed on multiple single-frequency current signals corresponding to the period to obtain the amplitude of multiple single-frequency current signals. Based on the multiple single-frequency current signals and their amplitudes, a frequency domain feature map corresponding to the period is generated, where the horizontal and vertical axes of the frequency domain feature map represent frequency and the vertical axis represents amplitude.
[0175] Optionally, the device further includes:
[0176] The training module is used for:
[0177] A first training sample set is constructed based on multiple sample voltage and current trajectory diagrams corresponding to malignant loads and multiple sample voltage and current trajectory diagrams corresponding to non-malignant loads.
[0178] The initial first classifier is iteratively trained based on the first training sample set until the latest iteration of the first classifier satisfies the preset convergence condition. The first classifier that satisfies the preset convergence condition is taken as the first recognition model.
[0179] The second training sample set is constructed based on the frequency domain feature maps of multiple samples corresponding to malignant loads and multiple samples corresponding to non-malignant loads.
[0180] The initial second classifier is iteratively trained based on the second training sample set until the latest iteration of the second classifier meets the preset convergence condition. The second classifier that meets the preset convergence condition is used as the second recognition model.
[0181] Optionally, the training module is also used for:
[0182] Based on the multiple first recognition results output by the first recognition model, the labels corresponding to each first recognition result, the multiple second recognition results output by the second recognition model, and the labels corresponding to each second recognition result, the first regression parameters and the second regression parameters of the logistic regression model are determined.
[0183] Substituting the first and second regression parameters into the logistic regression model yields the third identification model.
[0184] Optionally, the identification module 904 is specifically used for:
[0185] The voltage and current trajectory diagram corresponding to the cycle is input into the first identification model to obtain the first identification result output by the first identification model. The first identification result is used to indicate whether there is a malicious load.
[0186] The frequency domain feature map corresponding to the period is input into the second recognition model to obtain the second recognition result output by the second recognition model. The second recognition result is used to indicate whether there is a malicious load.
[0187] The third identification model determines the first product of the first identification result and the first regression parameter, the second product of the second identification result and the second regression parameter, the sum of the first product and the second product, and the identification result corresponding to the period based on the sum and the preset threshold.
[0188] Optionally, module 905 is defined, specifically for:
[0189] If the identification result for at least one cycle matches the preset result, then it is determined that there is a malicious load on the target line.
[0190] The above-described device is used to execute the method provided in the foregoing embodiments, and its implementation principle and technical effect are similar, so they will not be described again here.
[0191] These modules can be one or more integrated circuits configured to implement the above methods, such as one or more Application Specific Integrated Circuits (ASICs), one or more digital signal processors (DSPs), or one or more Field Programmable Gate Arrays (FPGAs). Alternatively, when a module is implemented using processing element scheduler code, the processing element can be a general-purpose processor, such as a Central Processing Unit (CPU) or other processor capable of calling program code. Furthermore, these modules can be integrated together as a system-on-a-chip (SOC).
[0192] Optionally, the present invention also provides a program product, such as a computer-readable storage medium, including a program that, when executed by a processor, is used to perform the above-described method embodiments.
[0193] In the several embodiments provided by this invention, it should be understood that the disclosed apparatus and methods can be implemented in other ways. For example, the apparatus embodiments described above are merely illustrative; for instance, the division of units is only a logical functional division, and in actual implementation, there may be other division methods. For example, multiple units or components may be combined or integrated into another system, or some features may be ignored or not executed. Furthermore, the coupling or direct coupling or communication connection shown or discussed may be through some interfaces; the indirect coupling or communication connection between apparatuses or units may be electrical, mechanical, or other forms.
[0194] The units described as separate components may or may not be physically separate. The components shown as units may or may not be physical units; that is, they may be located in one place or distributed across multiple network units. Some or all of the units can be selected to achieve the purpose of this embodiment according to actual needs.
[0195] Furthermore, the functional units in the various embodiments of the present invention can be integrated into one processing unit, or each unit can exist physically separately, or two or more units can be integrated into one unit. The integrated unit can be implemented in hardware or in the form of hardware plus software functional units.
[0196] The integrated units implemented as software functional units described above can be stored in a computer-readable storage medium. These software functional units, stored in a storage medium, include several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) or processor to execute some steps of the methods described in the various embodiments of the present invention. The aforementioned storage medium includes various media capable of storing program code, such as USB flash drives, portable hard drives, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical disks.
Claims
1. A method for identifying electrical loads, characterized in that, The method includes: Acquire the AC signal of the target line to be identified over multiple cycles; The AC signal within each of the aforementioned cycles is decomposed to obtain the decomposed signal corresponding to each of the aforementioned cycles. Based on the decomposed signal corresponding to each period, a feature map corresponding to each period is generated, the feature map including: voltage and current trajectory map and frequency domain feature map; Based on the feature maps corresponding to each cycle and multiple pre-trained recognition models, the recognition results corresponding to each cycle are obtained. The multiple recognition models include: a first recognition model, a second recognition model, and a third recognition model. The first recognition model is used to identify whether there is a malicious load in each cycle based on the voltage and current trajectory maps of each cycle. The second recognition model is used to identify whether there is a malicious load in each cycle based on the frequency domain feature maps of each cycle. The third recognition model is used to obtain the recognition results corresponding to each cycle based on the recognition results of the first recognition model and the recognition results of the second recognition model in each cycle. Based on the identification results corresponding to each cycle, it is determined whether there is a malicious load on the target line.
2. The method according to claim 1, characterized in that, in, The alternating current signal includes: current signal and voltage signal; The process of decomposing the AC signal within each of the aforementioned periods to obtain the decomposed signal corresponding to each of the aforementioned periods includes: The current signal and voltage signal within the period are respectively subjected to a first decomposition process to obtain multiple sub-current signals and multiple sub-voltage signals corresponding to the period. Correlation analysis is performed on multiple sub-current signals and the current signal within the period, and a set of sub-current signals with a correlation greater than a first threshold is obtained; and correlation analysis is performed on multiple sub-voltage signals and the voltage signal within the period, and a set of sub-voltage signals with a correlation greater than a second threshold is obtained. The sub-current signal set is subjected to a second decomposition process to obtain multiple single-frequency current signals, and the sub-voltage signal set is subjected to a second decomposition process to obtain multiple single-frequency voltage signals. The multiple single-frequency current signals and the multiple single-frequency voltage signals are used as the decomposition signals corresponding to the period.
3. The method according to claim 2, characterized in that, The step of generating a feature map corresponding to each of the periods based on the decomposed signals corresponding to each period includes: Using a single-frequency current signal and a single-frequency voltage signal corresponding to the same moment as a set of coordinate points, the multiple single-frequency current signals and multiple single-frequency voltage signals corresponding to the period are divided into multiple sets of coordinate points, and a voltage and current trajectory diagram corresponding to the period is generated based on the multiple sets of coordinate points, wherein the horizontal and vertical axes of the voltage and current trajectory diagram represent current and the vertical axis represents voltage. Fourier transforms are performed on the multiple single-frequency current signals corresponding to the period to obtain the amplitude of the multiple single-frequency current signals. Based on the multiple single-frequency current signals and their amplitudes, a frequency domain feature map corresponding to the period is generated, wherein the horizontal and vertical axes of the frequency domain feature map represent frequency and the vertical axis represents amplitude.
4. The method according to claim 1, characterized in that, The first recognition model is trained based on the following steps: A first training sample set is constructed based on multiple sample voltage and current trajectory diagrams corresponding to malignant loads and multiple sample voltage and current trajectory diagrams corresponding to non-malignant loads. The initial first classifier is iteratively trained based on the first training sample set until the latest iteration of the first classifier satisfies the preset convergence condition. The first classifier that satisfies the convergence condition is taken as the first recognition model. The second recognition model is trained based on the following steps: The second training sample set is constructed based on the frequency domain feature maps of multiple samples corresponding to malignant loads and multiple samples corresponding to non-malignant loads. The initial second classifier is iteratively trained based on the second training sample set until the latest iteration of the second classifier satisfies the preset convergence condition. The second classifier that satisfies the preset convergence condition is used as the second recognition model.
5. The method according to claim 1, characterized in that, The third recognition model is trained based on the following steps: Based on the multiple first recognition results output by the first recognition model, the labels corresponding to each first recognition result, the multiple second recognition results output by the second recognition model, and the labels corresponding to each second recognition result, the first regression parameters and the second regression parameters of the logistic regression model are determined. Substituting the first regression parameter and the second regression parameter into the logistic regression model, the third identification model is obtained.
6. The method according to claim 5, characterized in that, The step of obtaining the recognition result corresponding to each of the aforementioned periods based on the feature maps corresponding to each period and multiple pre-trained recognition models includes: The voltage and current trajectory diagram corresponding to the cycle is input into the first identification model to obtain the first identification result output by the first identification model. The first identification result is used to indicate whether there is a malicious load. The frequency domain feature map corresponding to the period is input into the second identification model to obtain the second identification result output by the second identification model. The second identification result is used to indicate whether there is a malicious load. The third recognition model determines the first product result of the first recognition result and the first regression parameter, and determines the second product result of the second recognition result and the second regression parameter, and determines the summation result of the first product result and the second product result, and determines the recognition result corresponding to the period based on the summation result and a preset threshold.
7. The method according to claim 1, characterized in that, The step of determining whether there is malicious load on the target line based on the identification results corresponding to each of the aforementioned periods includes: If the identification result corresponding to at least one cycle is consistent with the preset result, then it is determined that there is a malicious load on the target line.
8. A device for identifying electrical loads, characterized in that, The device includes: The acquisition module is used to acquire the AC signal of the target line to be identified over multiple cycles. The decomposition module is used to decompose the AC signal in each of the said cycles to obtain the decomposed signal corresponding to each of the said cycles. The generation module is used to generate a feature map corresponding to each of the periods based on the decomposed signals corresponding to each period. The feature map includes a voltage and current trajectory map and a frequency domain feature map. The identification module is used to obtain the identification result corresponding to each of the cycles based on the feature map corresponding to each cycle and multiple identification models pre-trained. The multiple identification models include: a first identification model, a second identification model and a third identification model. The first identification model is used to identify whether there is a malicious load in each cycle based on the voltage and current trajectory map of each cycle. The second identification model is used to identify whether there is a malicious load in each cycle based on the frequency domain feature map of each cycle. The third identification model is used to obtain the identification result corresponding to each cycle based on the identification results of the first identification model in each cycle and the identification results of the second identification model in each cycle. The determination module is used to determine whether there is a malicious load on the target line based on the identification results corresponding to each of the said cycles.
9. An electronic device, characterized in that, include: The device includes a processor, a storage medium, and a bus, wherein the storage medium stores machine-readable instructions executable by the processor, and when the electronic device is in operation, the processor communicates with the storage medium via the bus, and the processor executes the machine-readable instructions to perform the steps of the method as described in any one of claims 1-7.
10. A computer-readable storage medium, characterized in that, The storage medium stores a computer program, which is executed by a processor to perform the method as described in any one of claims 1-7.