CN-122026342-A - Non-invasive high-power electric appliance access detection and risk early warning method, system, equipment and medium

CN122026342ACN 122026342 ACN122026342 ACN 122026342ACN-122026342-A

Abstract

The invention discloses a non-invasive high-power electrical appliance access detection and risk early warning method, system, equipment and medium, which comprise the steps of detecting a differential threshold based on a total surface side instantaneous power sequence, identifying an on/off event, screening high-power electrical appliance candidate events by a first preset threshold, carrying out voltage zero-crossing phase alignment on the on event, extracting steady-state current waveforms before and after the on event, calculating event current components, inputting the event current components into a time sequence convolution network for processing, inputting the event current components into a two-dimensional convolution neural network for classification after learning image features, outputting load types of the events, acquiring a running time self-adaptive threshold by a first clustering algorithm, and triggering an early warning mechanism if the high-power electrical appliance candidate events do not detect paired off events within the running time self-adaptive threshold. The invention realizes the high-efficiency detection and risk prevention and control of the high-power electric appliance access, and has stronger practicability and popularization value.

Inventors

LUAN WENPENG
LI KEKE
LIU BO
Cai Hanju
HAN YANG
ZHAO BOCHAO
HUANG YAJUAN

Assignees

天津大学

Dates

Publication Date: 20260512
Application Date: 20251201

Claims (10)

1. A non-invasive high-power electrical appliance access detection and risk early warning method is characterized by comprising the following steps: Identifying an opening event and a closing event based on differential threshold detection of a total surface instantaneous power sequence, and screening high-power electrical appliance candidate events by a first preset threshold; Performing voltage zero crossing phase alignment on the starting event, extracting steady-state current waveforms before and after the starting event, and calculating event current components; Inputting the event current component into a time sequence convolution network for processing to obtain a learnable image characteristic; Inputting the characteristic of the leachable image into a two-dimensional convolutional neural network for classification, and outputting the load category to which the event belongs; constructing a duration set of each electric appliance based on the historical identification result, and acquiring a running time self-adaptive threshold value through a first clustering algorithm; and if the high-power electrical appliance candidate event does not detect the matched closing event within the running time self-adaptive threshold, triggering an early warning mechanism.
2. The method for non-invasive high-power appliance access detection and risk pre-warning according to claim 1, wherein the screening the high-power appliance candidate event with the first preset threshold comprises: collecting a voltage sequence and a current sequence from the total surface end of a user, and calculating an instantaneous power sequence; performing sliding differential operation on the instantaneous power sequence, and detecting the occurrence time of a load switching event; when the sliding difference operation result exceeds a set event detection threshold value, judging that a load switching event exists; If the steady-state power corresponding to the load switching event exceeds a first preset threshold, recording the load switching event as a high-power electrical appliance candidate event, otherwise, discarding the high-power electrical appliance candidate event; And according to the principle of conservation of the power steps, matching the opening event with the closing event in a time neighborhood to obtain a complete event pair or an independent opening event marked as to-be-confirmed closing.
3. The method for non-invasive high-power electrical access detection and risk early warning according to claim 2, wherein the calculating the event current component comprises: Selecting a plurality of power frequency periods before the starting event occurs, and calculating a steady-state baseline current; Selecting a plurality of power frequency periods after the starting event occurs, and calculating the steady-state current after the event; And calculating an event current component based on the steady-state baseline current and the post-event steady-state current, and performing first processing on the event current component to generate a standardized event current waveform.
4. The method for non-invasive high-power electrical access detection and risk early warning according to claim 3, wherein said inputting the event current component into a time-series convolution network for processing, obtaining the learnable image features comprises: Inputting the standardized event current waveform into a time sequence convolution network, and extracting a time sequence feature matrix; carrying out channel number alignment treatment on the time sequence feature matrix through point-by-point convolution to obtain an aligned feature matrix; calculating correlation matrixes of different characteristic channels on a time axis based on the aligned characteristic matrixes; and normalizing the correlation matrix into a single-channel gray level image, and outputting the single-channel gray level image as a characteristic of a leachable image.
5. The method for detecting and warning risk of access to a high-power electrical appliance according to claim 4, wherein the load category to which the output event belongs comprises: Inputting the leachable image characteristics into a multi-layer two-dimensional convolutional neural network, and processing the leachable image characteristics layer by layer through batch normalization, nonlinear activation and pooling operation; And inputting the finally extracted image features into a full-connection layer, obtaining load category probability distribution through Softmax function processing, selecting a load category with highest probability as a recognition result, and outputting corresponding classification confidence.
6. The method for non-invasive high-power electrical access detection and risk early warning according to claim 5, wherein the obtaining the running time adaptive threshold by the first clustering algorithm comprises: Collecting operation duration samples of all electric appliances in a history period, forming a duration set, performing robustness preprocessing on the duration set, and eliminating the influence of abnormal data points; performing cluster analysis on the preprocessed duration time set by adopting a first clustering algorithm, determining the optimal cluster quantity, and selecting a data cluster corresponding to a normal working condition from a clustering result based on posterior responsibility and business priori knowledge; Calculating a running time self-adaptive threshold according to the distribution characteristics of the normal working condition data cluster and adding a safety margin; And carrying out online updating and self-adaptive adjustment on the self-adaptive threshold value of the running time so as to adapt to the change of the electricity consumption behavior of the user.
7. The method for detecting access and early warning risk of a non-invasive high-power electrical appliance according to claim 6, wherein the triggering early warning mechanism comprises: if the matched closing event is detected within the self-adaptive threshold value of the running time, the registration record of the event is cleared, and if the current moment exceeds the self-adaptive threshold value of the running time and the closing event is not detected, the overtime early warning is triggered; comprehensively determining an overtime early warning level according to the classification confidence level, the overtime time length and the continuous power; And writing back the actually-occurring running duration to the historical data set, and periodically updating the running time adaptive threshold value.
8. A non-invasive high-power electrical appliance access detection and risk early warning system, applying the non-invasive high-power electrical appliance access detection and risk early warning method as claimed in any one of claims 1 to 7, comprising: the load event detection module is used for identifying an opening event and a closing event based on differential threshold detection of the total surface instantaneous power sequence and screening high-power electrical appliance candidate events by a first preset threshold; the characteristic extraction module is used for carrying out voltage zero crossing phase alignment on the starting event, extracting steady-state current waveforms before and after the starting event and calculating event current components; The image feature extraction module is used for inputting the event current component into a time sequence convolution network for processing to obtain a learnable image feature; The load feature classification module is used for inputting the learnable image features into a two-dimensional convolutional neural network for classification and outputting the load category to which the event belongs; The threshold value acquisition module is used for constructing a duration time set of each electric appliance based on the historical identification result and acquiring a running time self-adaptive threshold value through a first clustering algorithm; And the risk early warning module is used for triggering an early warning mechanism if the high-power electrical appliance candidate event does not detect a matched closing event within the running time self-adaptive threshold.
9. An electronic device comprising a memory and a processor, wherein the memory is configured to store computer executable instructions, and the processor implements the steps of a non-invasive high-power electrical access detection and risk early warning method as claimed in any one of claims 1 to 7 when executing the computer executable instructions.
10. A computer readable storage medium having stored thereon computer executable instructions, wherein the computer executable instructions when executed by a processor implement the steps of a non-invasive method for detecting access to and risk pre-warning of high power electrical appliances as claimed in any one of claims 1 to 7.

Description

Non-invasive high-power electric appliance access detection and risk early warning method, system, equipment and medium Technical Field The invention relates to the technical field of power load monitoring, in particular to a non-invasive high-power electrical appliance access detection and risk early warning method, system, equipment and medium. Background The existing high-power electric appliance monitoring method mostly depends on an invasive monitoring means, namely, energy consumption data acquisition is carried out by installing special sensors on all electric appliance ports. Although the existing method can acquire the operation information of the electric appliance with higher precision, the method faces the problems of complex installation, high cost, low user acceptance and the like in actual popularization, and is not beneficial to large-scale application in resident families and office environments. In contrast, the Non-invasive load monitoring (Non-Intrusive Load Monitoring, NILM) technology can realize the identification and decomposition of various load operation states and energy consumption under the condition that a single electric appliance is not required to be additionally provided with a sensor by only collecting voltage, current or power signals of the total surface, so that the Non-invasive load monitoring (Non-Intrusive Load Monitoring, NILM) technology has more practicability and economy. In the conventional NILM method, the mainstream load recognition technology mainly includes a signal processing method based on event detection and feature extraction, a statistical modeling method based on a hidden markov model, and a pattern recognition method based on a deep learning method. The signal processing method depends on the characteristics of manual design, can realize better recognition effect under specific scenes, but has insufficient adaptability to complex loads or scenes in which multiple electric appliances are operated simultaneously. The statistical modeling method deduces the state of the electric appliance by establishing an implicit relation between state transition and observed power, has the advantages of intuitive modeling and higher reasoning efficiency, and is limited in the process of multi-state electric appliances and high noise scenes. The deep learning method can automatically learn the features, has higher recognition accuracy, but has serious dependence on a large amount of annotation data, and has easily limited generalization capability. In addition, the non-invasive detection and risk early warning of the current high-power electrical appliance are obviously insufficient, wherein (1) load identification depends on traditional artificial features or shallow models, the identification capability is limited, (2) the lack of a running time threshold model constructed by combining user historical behavior data is difficult to effectively monitor abnormal running behaviors, and (3) the existing research neglects the safety risk characteristics of the high-power electrical appliance, and a systematic risk early warning mechanism is not formed. Therefore, a new method is needed to be capable of combining an appliance identification means based on the characteristic of a leachable image, fully utilizing historical electricity consumption data of a user to establish normal operation time thresholds of various appliances, and constructing a detection and early warning mechanism of a high-power appliance on the basis, so that accurate identification of high-power appliance access, effective monitoring of overtime operation and timely early warning of risk hidden danger are realized. Disclosure of Invention The present invention has been made in view of the above-described problems occurring in the prior art. Therefore, the invention provides a non-invasive high-power electric appliance access detection and risk early warning method, system, equipment and medium, which solve the problems that the current non-invasive load monitoring technology lacks a special risk early warning mechanism for the high-power electric appliance and is difficult to combine with the historical behavior of a user to self-adaptively judge the abnormal running state. In order to solve the technical problems, the invention provides the following technical scheme: in a first aspect, the present invention provides a method for detecting access and early warning risk of a non-invasive high-power electrical appliance, including: Identifying an opening event and a closing event based on differential threshold detection of a total surface instantaneous power sequence, and screening high-power electrical appliance candidate events by a first preset threshold; Performing voltage zero crossing phase alignment on the starting event, extracting steady-state current waveforms before and after the starting event, and calculating event current components; Inputting the event current component into a time se