Voltage fluctuation time range mining method

Abstract

The invention relates to a voltage fluctuation time range mining method, and belongs to the technical field of Internet of Things and power data analysis, and the method comprises the steps: obtaining asynchronous voltage data of a plurality of Internet of Things devices in a target region; identifying time points when the voltage is higher or lower than a threshold value in an anti-interference manner through a continuous verification mechanism; calculating the offset duration of each low-voltage event by taking the first reliable high-voltage event as the starting time; counting the equipment number distribution of the offset time length and forming a histogram; performing frequency domain transformation on the distribution information to extract periodic features; according to the method, the effect of performing high-robustness and accurate mining on the voltage fluctuation period in an asynchronous and noisy data environment of the Internet of Things is achieved, and the technical problem of low period identification accuracy caused by dependence on data synchronism and poor anti-interference capability of a traditional method is solved.

Description

Technical Field

[0001] This invention relates to the fields of Internet of Things and power data analysis technology, and in particular to a method for mining the time range of voltage fluctuations. Background Technology

[0002] With the rapid development of smart grid and Internet of Things (IoT) technologies, it has become possible to monitor the voltage quality at the end of the distribution network in real time through a large number of user-side devices (such as smart meters and electricity terminals). Accurately grasping voltage fluctuations, especially their periodic patterns, is of great significance for ensuring the safety of electrical equipment, optimizing grid operation, and implementing demand-side response.

[0003] Currently, the main approaches to mining voltage fluctuation cycles from massive amounts of equipment data can be divided into two categories: The first category is based on synchronous sampling and waveform analysis. This type of method requires monitoring devices to report voltage data at a fixed, synchronous sampling rate, forming a regular time series. Subsequently, classic signal processing techniques (such as Fourier transform and autocorrelation analysis) are used to directly extract the periodic component from the voltage waveform. However, this method has significant limitations: in real-world IoT environments, massive numbers of devices typically report data via wireless networks in an event-driven or periodic but asynchronous manner. Network latency, device clock differences, and communication protocol limitations cause the data to be inherently asynchronous and non-uniformly spaced on the time axis. Forcibly resampling or aligning this type of data introduces errors, destroys the original signal characteristics, and makes traditional waveform analysis methods ineffective or even impossible to apply directly.

[0004] The second category is based on simple thresholding and event statistics. To accommodate asynchronous data, some methods attempt to simplify the model, for example, by setting a threshold for voltage (e.g., 220V) and statistically analyzing the temporal distribution of alarm events where the voltage exceeds the threshold. However, these methods are too crude, and their core problem lies in their weak anti-interference capability. Voltage data collected by IoT devices inevitably contains measurement noise, instantaneous pulse interference, etc. Simple thresholding methods may misclassify a single noise point as a valid voltage exceedance event, generating a large number of "pseudo-events." These pseudo-events are mixed with real voltage fluctuation events, severely polluting the event sequence, leading to chaotic and unreliable periodic characteristics based on them, resulting in low accuracy.

[0005] In summary, existing technologies face a dilemma when dealing with voltage cycle mining in the Internet of Things (IoT) environment: waveform analysis methods, which are demanding in terms of data regularity, cannot adapt to the asynchronous nature of IoT data; while simple thresholding methods that attempt to adapt to asynchronous data suffer from low accuracy due to poor noise resistance. Therefore, there is an urgent need for a new method for accurately mining voltage fluctuation cycles that can simultaneously accommodate the asynchronous nature of IoT data and possess strong noise resistance. Summary of the Invention

[0006] To address the aforementioned technical problems, this invention proposes a method for mining the time range of voltage fluctuations. This invention achieves high robustness and accuracy in mining voltage fluctuation cycles in an asynchronous and noisy IoT data environment, solving the technical problem of low cycle identification accuracy caused by the reliance on data synchronization and poor anti-interference ability of traditional methods.

[0007] The technical solution to achieve the objective of this invention is: a method for mining the time range of voltage fluctuations, comprising the following steps: Acquire voltage data samples from multiple sample devices in the target area within the target time period; Based on the voltage data samples, a first time point is determined within the target time period when the voltage reported by the sample device is greater than a first threshold, and a second time point is determined when the reported voltage is lower than a second threshold; wherein, the first threshold is greater than or equal to the second threshold; Taking the first occurrence of the first time point as the starting time, calculate the offset time of each second time point relative to the starting time; Based on all the aforementioned offset durations, statistically analyze the distribution information of the number of devices along the offset duration dimension; Based on the periodic characteristics of the device quantity distribution information, the voltage fluctuation period of the target area is determined.

[0008] In some embodiments, after the acquisition step, the voltage data samples are preprocessed, including deleting duplicate reported data and / or removing abnormal data that exceeds a preset reasonable voltage range.

[0009] In some embodiments, the step of determining the first time point and the second time point includes: for a single sample device, only when the voltage data it reports multiple times consecutively meets the condition of being greater than the first threshold or lower than the second threshold is the device determined to have entered the corresponding voltage state, and the reporting time point of the first data in the data group that meets the condition is determined as the corresponding first time point or second time point. In some embodiments, the step of statistically generating device quantity distribution information includes: dividing the offset duration into multiple consecutive time intervals and counting the number of offset durations falling into each time interval to form a distribution histogram.

[0010] In some embodiments, the step of determining the voltage fluctuation period based on the periodic characteristics of the device quantity distribution information includes: performing frequency domain transformation processing on the device quantity distribution information, and determining the voltage fluctuation period based on the peak frequency in the transformed spectrum.

[0011] In some embodiments, the method further includes: when there are differences in voltage fluctuation characteristics within the target time period, dividing the target time period into at least two sub-time periods, and executing the mining method to determine the voltage fluctuation cycle of each sub-time period.

[0012] In some embodiments, the method further includes: sending the determined voltage fluctuation period parameter to a service server; wherein the sending process is performed using an encrypted communication protocol.

[0013] In some embodiments, the method is performed by a mining server that is in communication with a message server, and the voltage data samples are obtained from the sample device through the message server.

[0014] In some embodiments, the sample device is an electrical terminal device that has undergone IoT transformation within the target area, including industrial production equipment, residential appliances, and public facility electrical equipment; the sample device reports voltage data at a preset frequency via the MQTT communication protocol, the preset frequency being once per minute, and the reported voltage data includes at least a unique device identifier, a high-precision timestamp, and a voltage measurement value.

[0015] In some embodiments, the sub-time periods include daytime periods and nighttime periods, with the daytime period defined as 08:00 - 20:00 and the nighttime period defined as 20:00 - 08:00 the next day; the statistical accuracy of the voltage fluctuation cycle corresponding to each sub-time period is different, with the statistical error of the daytime period controlled within ±5 seconds and the statistical error of the nighttime period controlled within ±10 seconds.

[0016] Compared with existing technologies, the significant advantages of this invention are: 1. This invention fundamentally enhances the robustness of its method through a dual innovative design: a "continuous verification mechanism" and "distribution statistics based on offset duration." The principle is as follows: The "continuous verification mechanism" simulates the continuous physical characteristics of real voltage state transitions. By requiring consistent measurements across multiple consecutive measurements, it filters out instantaneous noise, ensuring the reliability of state transition event identification and eliminating the possibility of misjudging noise as the start of a periodic signal from the data source. Subsequently, the offset duration of all reliable events relative to the same starting point is calculated, and the analysis object is transformed from dispersed absolute time points to a unified relative time interval, eliminating interference from asynchronous device reporting. Finally, the group distribution of offset durations of a large number of devices is statistically analyzed. Based on the law of large numbers, the real periodic signal caused by fluctuations in the common power grid is amplified in the statistical distribution, while random individual noise is weakened by averaging. Therefore, through the above synergistic effect, this invention can stably and accurately mine the real period of voltage fluctuations in real-world IoT environments containing significant noise and asynchronous data, achieving the core objective of "higher mining accuracy."

[0017] 2. This invention perfectly adapts to the asynchronous and non-uniformly spaced data characteristics of the Internet of Things (IoT), possessing strong versatility across various scenarios. In IoT power monitoring scenarios, massive numbers of devices independently report data, exhibiting inherent asynchronicity and non-uniformity. This makes traditional signal periodicity detection algorithms (such as directly applying FFT to the original time series) that rely on fixed sampling rates or require strict time alignment difficult to apply. This invention cleverly solves this fundamental adaptation problem by introducing a step of "offset duration calculation based on the starting event." The principle is that this method shifts the focus of analysis from the precise alignment of voltage data on the absolute time axis to the time delay of each device's state transition event relative to a common reference event. This design completely eliminates the dependence on data synchronization, enabling direct and efficient processing of raw IoT data reported from different devices at different frequencies and at different times. Therefore, this invention can be seamlessly integrated into existing IoT architectures without modifying the data acquisition layer, possessing strong versatility for application in a wide range of practical scenarios.

[0018] 3. This invention provides refined and adaptive insight into the time-varying characteristics of voltage fluctuations: Power grid load exhibits significant time-varying characteristics (such as diurnal variations), which can lead to changes in voltage fluctuation cycles. A single global cycle estimate cannot reflect this dynamic pattern. This invention provides refined analytical capabilities through an optional "time-segmented differential analysis" step. The principle is that this method automatically divides a long-term analysis window into multiple sub-time periods with more stable internal characteristics (such as day and night) based on preset rules or data characteristics, and independently executes a complete cycle mining process. Through this "segmented stabilization" process, the method can reveal the independent periodic characteristics of voltage fluctuations within different time periods, thus providing a refined "portrait" of voltage fluctuations that changes over time. This makes the output of this invention not only a macroscopic cycle value, but also able to support advanced application needs such as differentiated dispatching of the power grid, refined management, and preventative maintenance of electrical equipment.

[0019] 4. This invention achieves full automation and high objectivity in the periodicity identification process, improving the reliability of the results: Traditional methods rely on manual observation of waveforms or experience to judge the period, resulting in strong subjectivity, low efficiency, and difficulty in quantification and reproduction. This invention automates and objectifies the periodicity identification process through the "period determination based on frequency domain transformation" step. The device quantity distribution information is treated as a discrete signal, and standard algorithms such as Fast Fourier Transform (FFT) are used to transform it to the frequency domain. In the frequency domain, the periodicity is manifested as concentrated energy spectral peaks, the corresponding frequencies of which can be automatically and accurately detected by the algorithm and converted into periodic values. This process is entirely algorithm-driven, eliminating subjective biases introduced by human intervention, ensuring the repeatability, scientific rigor, and high efficiency of the results. This allows the method to be integrated as a reliable decision support module into an automated smart grid analysis platform.

[0020] 5. This invention reduces system implementation complexity and deployment costs, facilitating technology promotion: At the engineering implementation level, this invention directly utilizes the widely deployed "device-message server" IoT communication architecture as the data source, eliminating the need for additional deployment of dedicated, expensive synchronous data acquisition systems or large-scale modifications to terminal devices for periodic data mining. The method's design follows the principle of "adapting to the data source rather than modifying it," accepting native, heterogeneous IoT data formats and solving analytical challenges through upper-layer algorithm innovation. This means that this mining method can be quickly and cost-effectively deployed on existing IoT power monitoring systems, significantly lowering the barrier to technology implementation and facilitating the rapid promotion and large-scale application of this advanced analytical technology. Attached Figure Description

[0021] The present invention will be further explained below with reference to the accompanying drawings and embodiments: Figure 1 This is a schematic diagram of the steps of a method for mining the time range of voltage fluctuations provided in one embodiment of the present invention; Figure 2 This is a schematic diagram of determining the voltage fluctuation period based on the periodic characteristics of distribution information in one embodiment of the present invention; Figure 3 This is a flowchart of voltage fluctuation time range mining provided in one embodiment of the present invention; Figure 4 This is a schematic diagram of voltage data sample preprocessing and state transition time point identification provided in one embodiment of the present invention; Figure 5 This is a schematic diagram of offset duration calculation and device quantity distribution information provided in one embodiment of the present invention. Detailed Implementation

[0022] The present invention will now be described in detail, and the technical solutions in the embodiments of the present invention will be clearly and completely described. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0023] This invention provides an improved method for mining the time range of voltage fluctuations. The technical solution of this invention is as follows: like Figure 1 - Figure 5 As shown in the figure, the voltage fluctuation time range mining method provided by the embodiment of the present invention specifically includes the following steps: S100: Acquire voltage data samples from multiple sample devices in the target area within the target time period. Specifically, the target area can be a power supply area such as a residential community, industrial park, or commercial building. The target time period can be flexibly set according to the analysis needs. For example, to study the daily cycle characteristics, it can be set to a continuous 24 hours (0:00 to 24:00), and to study the peak electricity consumption pattern, it can be set to 9:00 am to 5:00 pm on weekdays.

[0024] In a specific implementation, the sample devices are electrical terminal devices within the target area that have undergone IoT transformation, including industrial production equipment, residential appliances, and public utility electrical equipment. These devices have built-in voltage sensors and follow IoT communication protocols (such as MQTT), reporting their collected mains voltage values ​​to a centralized message server at a preset reporting frequency (such as once per minute). The data mining system subscribes to relevant data topics and asynchronously retrieves all voltage reporting data packets from all sample devices within the target time period from the message server. Each data packet contains at least a unique device identifier, a high-precision timestamp, and a voltage measurement value, thus forming an initial voltage data sample set. The voltage data reporting frequency of the sample devices is fixed at once per minute, and the core fields of the reported data are consistent with the claims, ensuring the standardization of data collection and the reliability of subsequent analysis.

[0025] By setting step S100, raw voltage data is acquired from real and complex IoT power monitoring scenarios. The key is that this invention directly utilizes the widely deployed "device-message server" communication architecture in IoT systems as the data source, eliminating the need for a dedicated synchronous data acquisition system. The principle and benefits of this setup are that IoT device data reporting inherently possesses discreteness, asynchronicity, and non-uniform interval characteristics, making traditional synchronous acquisition methods based on fixed sampling rates difficult to apply directly. The method of this invention is designed from the outset to adapt to these data characteristics, enabling seamless integration into existing IoT infrastructure. It boasts strong implementation convenience and scenario universality, laying the foundation for application in a wide range of real-world scenarios.

[0026] S200: Based on the voltage data samples, determine the first time point within the target time period where the voltage reported by the sample device is greater than a first threshold, and the second time point where the reported voltage is lower than a second threshold. This step is the core of state transition event identification. The first and second thresholds can typically be set to standard mains voltage values ​​(e.g., 220V). In practical applications, to improve data quality and the reliability of event identification, this step can further include two sub-steps: preprocessing and robust judgment.

[0027] Preprocessing sub-step: The raw voltage data samples obtained in step S100 are cleaned. First, the data is compared based on three fields: device identifier, timestamp, and voltage value, and data records that are completely duplicated due to network latency or other reasons are deleted. Second, a reasonable voltage range is set (e.g., based on national power standards and historical data statistics, set to 180V to 260V), and measurements outside this range are discarded as obvious outliers to form a valid voltage data sequence.

[0028] Robust Judgment Sub-step: To effectively resist misjudgments caused by single measurement noise or instantaneous interference, this embodiment introduces a continuous verification mechanism to identify reliable voltage state transition events. Taking a single sample device as an example, the system reads its cleaned voltage data sequence in chronological order. Only when the reported voltage values ​​are consistently greater than the first threshold (e.g., 220V) for a preset number of consecutive times (e.g., 3 times), the device is determined to be in a "voltage too high" state, and the timestamp of the first report in this set of continuous data is recorded as a first time point. Similarly, only when the reported voltage values ​​are consistently lower than the second threshold (e.g., 220V) for a preset number of consecutive times (e.g., 3 times), the device is determined to be in a "voltage too low" state, and the first timestamp is recorded as a second time point. Data fluctuations that do not meet the continuity condition are considered noise and ignored.

[0029] By setting up step S200, especially the introduction of "preprocessing" and "continuous verification mechanism," noise-resistant and highly reliable extraction of voltage state transition time points is achieved. The technical principle is that real grid voltage state transitions (such as those caused by the start-up and shutdown of large equipment or grid dispatching) typically have a certain degree of continuity and are continuously measured within a short time window; while sensor noise and transient pulse interference manifest as isolated or short-lived outliers. Preprocessing directly filters out obviously erroneous data; the continuous verification mechanism utilizes the continuous characteristic of real state transitions, distinguishing real events from transient noise by requiring consistency in multiple consecutive measurements. This design offers two beneficial effects: first, it ensures the high reliability of each extracted state transition time point, avoiding misjudging noise points as event starting points; second, it provides a robust analytical benchmark for the entire method, significantly reducing the impact of noise on the accuracy of subsequent cycle mining from the data source.

[0030] It is important to note that the aforementioned "continuous verification mechanism" is a refinement and optimization of the concept of "first" exceeding the threshold. Directly using the absolute "first" high voltage point in time is highly susceptible to transient noise interference. This invention uses the "first voltage spike event that satisfies the continuous stability condition" as the starting benchmark, essentially pursuing an optimal balance between early timing and event reliability. This definition is more in line with engineering practice and is a key design feature for improving the accuracy of the entire method.

[0031] S300: Taking the first occurrence of the first time point as the starting time, calculate the offset time of each of the second time points relative to the starting time. Based on step S200, the system sorts all identified first time points according to their timestamps and determines the earliest timestamp as the start time T0 of the entire analysis. This start time T0 is the moment when the "first reliable high voltage event" occurs.

[0032] Subsequently, all second time points identified in step S200 (i.e., the times when reliable "low voltage" events occur) are iterated. For each such time point Tn, the difference between it and the starting time T0 is calculated, resulting in: offset duration ΔTn = Tn - T0. All calculated ΔTn constitute an offset duration dataset. High-precision timing standards (such as UNIX timestamps, accurate to milliseconds) are used for timestamp processing to ensure calculation accuracy.

[0033] By setting step S300, discrete events occurring asynchronously and at absolute times are transformed into a sequence of relative time intervals starting from the same reference event. The underlying principle lies in a shift in analytical perspective. Voltage fluctuations in the power grid have a common impact on devices within the same power supply area. When a voltage spike occurs, the affected devices will gradually enter a voltage drop state. Due to differences in their characteristics and load conditions, the absolute times of state transitions for each device are dispersed and asynchronous. However, their time delays (i.e., offset durations) relative to the common starting event (voltage spike) contain the periodic information of the voltage fluctuation. The core benefit of this transformation is that it completely eliminates the dependence on the synchronization of device data reporting, perfectly adapting to the asynchronous and non-uniform interval characteristics of IoT data. Simultaneously, it shifts the focus of analysis from the specific waveform of voltage amplitude to the "event interval" relationship, which better reflects the essence of the fluctuation, providing a clean and well-organized data foundation for subsequent statistical analysis.

[0034] S400: Based on all the aforementioned offset durations, statistically analyze the device quantity distribution information along the offset duration dimension. This step aims to uncover group patterns from individual offset duration data of a large number of devices. Specifically, the system performs statistical analysis on the offset duration dataset generated in step S300. First, the overall range of offset durations (from minimum to maximum) is determined. Then, this range is divided into multiple consecutive, non-overlapping time intervals (e.g., 0-60 seconds, 60-120 seconds) at fixed time intervals (e.g., 60 seconds). Next, the number of offset durations ΔTn falling within each time interval is counted. This number represents the number of devices that undergo a "voltage high to low" state transition at that specific time offset.

[0035] Ultimately, the system generates a distribution histogram with the offset duration interval as the x-axis and the number of devices in the corresponding interval as the y-axis. This histogram is an intuitive visualization of the "distribution information of the number of devices along the offset duration dimension".

[0036] By setting step S400, statistical distribution characteristics reflecting group behavior are extracted from discrete individual time interval data. The technical principle is based on the law of large numbers and the central limit concept in probability statistics. The offset duration of a single device is influenced by its own random factors and may not exhibit a clear pattern. However, when the number of sample devices is sufficient, state transition events driven by voltage fluctuations in the common power grid will form statistical clusters around specific time offsets. The beneficial effect of this step is that it transforms individual data that may contain random noise into a distribution graph that clearly shows the potential cyclical pattern through aggregated statistics. The peak position in the distribution histogram intuitively indicates the typical time delay of voltage state transitions experienced by most devices, providing a powerful data representation for the next step of automatically and accurately identifying fluctuation cycles.

[0037] S500: Based on the periodic characteristics of the device quantity distribution information, determine the voltage fluctuation period of the target area. This step is the final extraction of periodic information. Based on the distribution histogram generated in step S400, the system uses a periodicity identification algorithm from the field of signal processing for analysis. In a preferred embodiment, a frequency domain transformation method is used.

[0038] The specific implementation process is as follows: The vertical axis sequence of the distribution histogram (i.e., the number of devices in each interval) is treated as a discrete time-domain signal. A Fast Fourier Transform (FFT) is performed on this signal to transform it from the time domain to the frequency domain, obtaining the corresponding spectrum. In the spectrum, the peak frequency f with a significantly higher amplitude than the background noise is searched. The period T = 1 / f corresponding to this peak frequency f is the main voltage fluctuation period extracted from the data. For example, if the peak frequency obtained from analyzing 24-hour data is approximately 0.001157 Hz (i.e., 1 / 86400 Hz), then the corresponding period T is approximately 24 hours.

[0039] To further ensure the reliability of the results, cross-validation can be performed by combining time-domain analysis to check whether the distribution histogram also shows accumulation peaks near integer multiples of the period such as T, 2T, and 3T.

[0040] By setting steps S500, particularly employing frequency domain transformation algorithms such as FFT, objective, quantitative, and accurate identification of voltage fluctuation periods is achieved. The underlying principle is that any signal with periodicity or near-periodicity will concentrate its energy in the frequency domain at the fundamental frequency and its harmonics, manifesting as obvious spectral peaks. FFT analysis of the device quantity distribution transforms potentially less obvious periodic clustering patterns in the time domain into easily identifiable and quantifiable peak features in the frequency domain. The significant benefit of this method is that it avoids the subjectivity and inaccuracy of manually interpreting distribution histograms, automatically and accurately locating core period values, greatly improving the automation and scientific rigor of period mining, and enabling this method to be integrated into smart grid management systems as a reliable decision support tool.

[0041] S600: Time-based Differentiation Analysis Considering that the power grid load varies significantly throughout the day, the voltage fluctuation characteristics may exhibit time-varying characteristics (for example, the load is light at night and the fluctuation period is long; the load is heavy during the day and the fluctuation period is short). The method of this invention can be further extended to have the ability to perform refined analysis in different time periods.

[0042] In one specific implementation, when the system determines, through preliminary analysis or based on preset rules (such as dividing a day into daytime and nighttime periods based on historical experience), that there may be significant differences in voltage fluctuation characteristics within a target time period, it automatically initiates a time-segmented analysis mode.

[0043] Specifically, the system divides the original target time period into two or more sub-time periods, including daytime and nighttime periods. The daytime period is specifically defined as 08:00 - 20:00, and the nighttime period is specifically defined as 20:00 - 08:00 the next day. The statistical precision of the voltage fluctuation cycle varies for each sub-time period; the statistical error for the daytime period is controlled within ±5 seconds, and the statistical error for the nighttime period is controlled within ±10 seconds. This precision setting matches the sensitivity of the power grid load to voltage fluctuations at different times. Then, each sub-time period is treated as an independent target time period, and steps S100 to S500 are executed again completely. That is, data for each sub-time period is acquired separately, and state event identification, offset duration calculation, distribution statistics, and cycle identification are performed independently to obtain the voltage fluctuation cycle (e.g., T_day and T_night) corresponding to each sub-time period.

[0044] By extending the settings in step S600, adaptive and refined capture of the time-varying characteristics of voltage fluctuations is achieved. The underlying principle is the "segmented stabilization" approach used in signal processing for non-stationary signals. A long-term data sequence with potentially changing characteristics is divided into multiple short-term segments with relatively stable internal characteristics for separate analysis, thus enabling a more accurate description of the signal's local features. A key benefit of this extension is that it allows the method to not only provide an overall, average period estimate but also reveal and quantify the dynamic patterns of period variation over time (e.g., day / night, weekday / weekend). The resulting insights are deeper and more refined, offering greater practical value for refined power grid scheduling, demand-side response strategy development, and preventative maintenance of electrical equipment.

[0045] System Deployment and Data Application

[0046] The aforementioned method for mining the time range of voltage fluctuations can be deployed on a dedicated mining server. This server connects to an IoT message server via a network to acquire data and is equipped with the necessary computing and storage resources to efficiently run the algorithm. After completing the cycle calculation, the mining server formats and encapsulates the obtained voltage fluctuation cycle parameters (including overall cycle value, time-segmented cycle value, confidence level, etc.) and sends them to the power grid company's business server via a secure communication link (e.g., encrypted transmission using the TLS protocol), providing critical data input for advanced applications such as load forecasting, voltage and reactive power control, and equipment operation and maintenance management.

[0047] Summary of the working principle and process of this invention

[0048] This invention begins with the acquisition of raw data adapted to the heterogeneous and asynchronous environment of the Internet of Things (IoT) (S100). By cleaning the data and designing a continuous verification mechanism based on the principle of physical persistence (S200), key event moments characterizing power grid state transitions are robustly identified. By calculating the offset duration of all "low voltage" events relative to the first "high voltage" event (S300), the analytical perspective is shifted from an asynchronous absolute time axis to a unified relative time interval. Utilizing the law of large numbers, the distribution of offset durations across a large number of devices is statistically analyzed to highlight group patterns (S400). Finally, using signal processing techniques such as frequency domain transformation (S500), the periodic characteristics of voltage fluctuations are objectively and accurately extracted from the statistical distribution. The entire process transforms the noise and asynchronicity issues in IoT power big data into solvable statistical and signal processing problems through a series of innovative steps, thereby achieving highly reliable, adaptive, and accurate mining of mains voltage fluctuation cycles in complex real-world environments.

[0049] The technical means disclosed in this invention are not limited to those described above, but also include technical solutions composed of equivalent substitutions of the above technical features. Matters not covered in this invention are common knowledge to those skilled in the art.

Claims

1. A method for mining the time range of voltage fluctuations, characterized in that, Includes the following steps: Acquire voltage data samples from multiple sample devices in the target area within the target time period; Based on the voltage data samples, a first time point is determined within the target time period when the voltage reported by the sample device is greater than a first threshold, and a second time point is determined when the reported voltage is lower than a second threshold; wherein, the first threshold is greater than or equal to the second threshold; Taking the first occurrence of the first time point as the starting time, calculate the offset time of each second time point relative to the starting time; Based on all the aforementioned offset durations, statistically analyze the distribution information of the number of devices along the offset duration dimension; Based on the periodic characteristics of the device quantity distribution information, the voltage fluctuation period of the target area is determined.

2. The method for mining the time range of voltage fluctuations according to claim 1, characterized in that: After the acquisition step, the voltage data sample is preprocessed. The preprocessing includes deleting duplicate reported data and / or removing abnormal data that exceeds a preset reasonable voltage range.

3. The method for mining the time range of voltage fluctuations according to claim 1, characterized in that, The steps for determining the first time point and the second time point include: for a single sample device, when the voltage data reported by the device multiple times in a row meets the condition of being greater than the first threshold or lower than the second threshold, the device is determined to enter the corresponding voltage state, and the reporting time point of the first data in the data group that meets the condition is determined as the corresponding first time point or second time point.

4. The method for mining the time range of voltage fluctuations according to claim 1, characterized in that, The step of statistically generating the distribution information of the number of devices includes: dividing the offset duration into multiple consecutive time intervals, and counting the number of offset durations falling into each time interval to form a distribution histogram.

5. The method for mining the time range of voltage fluctuations according to claim 1, characterized in that, The step of determining the voltage fluctuation period based on the periodic characteristics of the device quantity distribution information includes: performing frequency domain transformation processing on the device quantity distribution information, and determining the voltage fluctuation period based on the peak frequency in the transformed spectrum.

6. The method for mining the time range of voltage fluctuations according to claim 1, characterized in that, The method further includes: when there are differences in voltage fluctuation characteristics within the target time period, dividing the target time period into at least two sub-time periods, and executing the mining method separately to determine the voltage fluctuation cycle of each sub-time period.

7. The method for mining the time range of voltage fluctuations according to claim 1, characterized in that, The method further includes: sending the determined voltage fluctuation period parameter to the service server; wherein the sending process is performed using an encrypted communication protocol.

8. The method for mining the time range of voltage fluctuations according to claim 1, characterized in that, The method is executed by a mining server that communicates with a message server, and the voltage data samples are obtained from the sample device through the message server.

9. The method for mining the time range of voltage fluctuations according to claim 1, characterized in that, The sample device is an electrical terminal device that has undergone IoT transformation within the target area, including industrial production equipment, residential appliances and public facility electrical equipment; the sample device reports voltage data at a preset frequency via the MQTT communication protocol, the preset frequency being once per minute, and the reported voltage data includes at least a unique device identifier, a high-precision timestamp and a voltage measurement value.

10. The method for mining the time range of voltage fluctuations according to claim 6, characterized in that, The sub-time periods include daytime periods and nighttime periods. The daytime period is defined as 08:00-20:00, and the nighttime period is defined as 20:00-08:00 the next day. The statistical accuracy of the voltage fluctuation cycle corresponding to each sub-time period is different. The statistical error of the daytime period is controlled within ±5 seconds, and the statistical error of the nighttime period is controlled within ±10 seconds.

Images