Load monitoring method of electric energy meter

Abstract

The invention relates to the technical field of electrical measurement, in particular to a load monitoring method for an electric energy meter, and the method comprises the steps: synchronously collecting a power grid voltage signal and a load current signal; performing phase difference estimation processing on the power grid voltage signal and the load current signal to obtain a fundamental wave phase difference; calculating a power factor according to the fundamental wave phase difference, the fundamental wave current effective value of the load current signal and the total current effective value; calculating active power based on the power grid voltage signal and the load current signal; determining the operation state of the target monitoring equipment based on the active power and the power factor; wherein the running state comprises a standby state or an equipment turn-off event. The practicability and the accuracy of the ammeter load monitoring function can be improved.

Description

Technical Field

[0001] This application relates to the field of electrical measurement technology, and in particular to a load monitoring method for an electricity meter. Background Technology

[0002] As a key metering device at the end of the power grid, the basic function of an electricity meter is to accurately record the cumulative electricity consumption of users. With the development of smart grids and advanced metering systems, the function of modern smart electricity meters has gradually expanded from simple electricity measurement to online sensing and analysis of user load characteristics, in order to support advanced applications such as demand-side response, electricity safety early warning, and refined energy management.

[0003] Currently, some advanced electricity meters can collect instantaneous electrical quantities such as voltage and current through built-in metering chips, and calculate derived parameters such as active power, reactive power, and power factor. Based on these parameters, some solutions attempt to identify the macroscopic operating status of downstream electrical equipment, for example, by setting a fixed low-power threshold to determine whether the equipment is in standby or off state.

[0004] However, the above methods face significant challenges in practical applications. Due to the vast differences in the electrical characteristics of various electrical devices, their standby power consumption can range from a few tenths of a watt to tens of watts. Furthermore, many nonlinear loads (such as switching power supplies) generate severe current harmonics during standby, resulting in extremely low power factors. This makes it difficult to rely solely on a single, fixed power threshold or simple power factor criterion in complex and ever-changing real-world power consumption scenarios, easily leading to misjudgments and thus limiting the practicality and accuracy of load monitoring functions. Summary of the Invention

[0005] Therefore, it is necessary to provide a load monitoring method for smart energy meters that improves the accuracy of the load monitoring function, in order to address the aforementioned technical problems.

[0006] This application provides a load monitoring method for an electricity meter, the method comprising: Simultaneously acquire grid voltage and load current signals; The phase difference between the grid voltage signal and the load current signal is estimated to obtain the fundamental phase difference. The power factor is calculated based on the fundamental phase difference, the effective value of the fundamental current of the load current signal, and the effective value of the total current. Calculate the active power based on the grid voltage signal and the load current signal; Based on the active power and the power factor, the operating status of the target monitoring device is determined; wherein, the operating status includes standby state or device shutdown event.

[0007] The beneficial effects of this invention are as follows: 1) By introducing power factor as a second criterion, this invention takes into account the phase characteristics (inductive / capacitive / resistive) and waveform distortion (harmonic content) of the load. Standby devices (such as switching power supplies) typically exhibit "low power consumption + low power factor," while a true shutdown event exhibits "zero power consumption + meaningless power factor." This two-dimensional criterion, combining "quantity" and "quality," more comprehensively characterizes the difference between the two states from a physical perspective, significantly improving the theoretical completeness and discriminative power of the judgment.

[0008] 2) This invention effectively captures current waveform distortion caused by harmonics by accurately calculating the true power factor (rather than a simple displacement power factor). This allows the system to move away from relying on a universal but fragile fixed threshold and instead dynamically evaluate based on the instantaneous and real power consumption characteristics of each device, thereby enhancing its adaptability and robustness to diverse and complex nonlinear loads.

[0009] 3) This invention establishes the final state determination based on two independent but complementary parameters: active power and power factor, and sets clear logical rules. This cross-validation mechanism effectively filters out false alarms caused by transient interference (such as brief power fluctuations due to grid voltage dips) that may occur with a single parameter. Furthermore, since both parameters are calculated based on synchronously acquired raw voltage and current signals, the data source is consistent, ensuring the consistency and stability within the determination logic, thus providing highly reliable monitoring results in actual operation. Attached Figure Description

[0010] To more clearly illustrate the technical solutions in the embodiments of this application or related technologies, the accompanying drawings used in the description of the embodiments of this application or related technologies will be briefly introduced below.

[0011] Figure 1 This is a schematic diagram of the hardware architecture of a smart energy meter used in a load monitoring method for an energy meter provided in an embodiment of the present invention. Figure 2 This is a schematic diagram of the process for determining the basic state based on active power and power factor in a load monitoring method for an electricity meter provided in an embodiment of this application. Figure 3 This is a schematic diagram of the process for accurately calculating the effective value of the fundamental current and the effective value of the total current based on frequency domain decomposition in a load monitoring method for an energy meter provided in an embodiment of this application. Figure 4 This is a flowchart illustrating the introduction of dynamic harmonic distribution characteristics to enhance the ability to identify shutdown events in a load monitoring method for an electricity meter provided in this application embodiment. Figure 5This is a flowchart illustrating a high-robust phase difference estimation method based on multi-cycle confidence weighting in a load monitoring method for an energy meter provided in an embodiment of this application. Figure 6 This is a schematic diagram of the frequency domain focusing process used to improve the quality of the single-cycle correlation function in a load monitoring method for an energy meter provided in an embodiment of this application; Figure 7 This is a schematic diagram of the process of deeply fusing device identification based on harmonic fingerprint and status determination in a load monitoring method for an electricity meter provided in an embodiment of this application. Figure 8 This is a schematic diagram of a refined judgment process for distinguishing standby and shutdown events using information entropy in a load monitoring method for an electricity meter provided in an embodiment of this application. Figure 9 This is a schematic diagram of the process of adaptively correcting the fundamental frequency using the centroid of the harmonic spectrum in a load monitoring method for an energy meter provided in an embodiment of this application. Figure 10 This is a flowchart illustrating an efficient phase difference estimation alternative based on orthogonal decomposition method in a load monitoring method for an electricity meter provided in this application embodiment. Figure 11 This is a flowchart illustrating the grid fundamental frequency tracking module based on zero-crossing detection and recursive filtering in a load monitoring method for an electricity meter provided in this application embodiment. Detailed Implementation

[0012] To make the objectives, technical solutions, and advantages of this application clearer, the following detailed description is provided in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative and not intended to limit the scope of this application.

[0013] In smart electricity monitoring scenarios, accurately identifying the true operating state of electrical equipment (such as distinguishing between low-power standby mode and complete power-off events) is fundamental to achieving refined energy management and user behavior analysis. Traditional electricity meters only provide cumulative electricity consumption or instantaneous power data, making it difficult to effectively distinguish between these two macroscopically "low-power" states. Therefore, there is an urgent need for a method that can deeply mine the inherent characteristics of voltage and current signals and make intelligent judgments based on them. This embodiment provides a load monitoring method for electricity meters, which is applied to... Figure 1 The present invention relates to a smart energy meter. The specific implementation steps of this invention will be described in detail below, using this smart energy meter as an example.

[0014] In one exemplary embodiment, such as Figure 2 As shown, this embodiment provides a load monitoring method for an electricity meter, including the following steps S101 to S105: Step S101: Synchronously acquire grid voltage signal and load current signal.

[0015] The grid voltage signal refers to the potential difference signal obtained from the public power grid that changes continuously over time. Its nominal frequency is 50Hz or 60Hz in most countries, and its ideal waveform is a sine wave. The load current signal refers to the current signal flowing through the electrical equipment to be monitored (i.e., the target monitoring equipment). Its waveform characteristics directly reflect the electrical characteristics and operating status of the equipment.

[0016] Synchronous acquisition is the physical foundation for ensuring the accuracy of all subsequent calculations. Voltage and current signals must be strictly aligned on the time axis; any tiny deviation in sampling timing will be significantly amplified in phase difference and power calculations, leading to incorrect final state determination. Therefore, using the same highly stable crystal oscillator as the sampling clock source for the analog-to-digital converter (ADC) is crucial for achieving high-precision synchronous acquisition.

[0017] In a specific hardware implementation, a smart meter integrates a dedicated metering chip. This chip contains at least two high-resolution (e.g., 16-bit or higher) ADC channels. One channel is connected to a voltage divider network to acquire the grid voltage signal; the other channel is connected to a current sensor (such as a manganese-copper shunt or current transformer) to acquire the load current signal. Both ADC channels are triggered by the same internal clock, ensuring that at every sampling moment... Both can simultaneously obtain a pair of precisely corresponding instantaneous voltage values. and instantaneous current value The sampling frequency is usually set to an integer multiple of the fundamental frequency. For example, for a 50Hz power grid, 6.4kHz (i.e., 128 sampling points per cycle) can be selected to meet the needs of signal reconstruction and analysis.

[0018] In a specific application scenario, suppose the target monitoring device is a household air conditioner. When the air conditioner is in cooling mode, its compressor motor is an inductive load, and the current waveform lags behind the voltage waveform, possibly containing certain harmonics. Through the aforementioned synchronous acquisition mechanism, the electricity meter can completely and without distortion capture this voltage-current signal pair, providing raw data for subsequent accurate analysis of its operating status.

[0019] Step S102: Perform phase difference estimation processing on the grid voltage signal and the load current signal to obtain the fundamental phase difference.

[0020] The fundamental phase difference refers to the phase offset angle between the fundamental component of the grid voltage signal and the fundamental component of the load current signal, usually represented by the Greek letter φ. This parameter is a core indicator characterizing the load nature (resistive, inductive, or capacitive) and a key input for calculating the power factor.

[0021] The fundamental purpose of phase difference estimation is to remove DC components, noise, and high-order harmonics from voltage and current signals, focusing on the fundamental frequency component, which is the most important part of energy transmission, thereby obtaining a stable and reliable measure of phase relationship. Directly using the zero-crossing points of the original signal to calculate the phase difference will result in drastic fluctuations when there are harmonics or noise pollution in the power grid, making it unsuitable for accurate state identification.

[0022] In a typical digital signal processing flow of this embodiment, the synchronously acquired voltage signal is first processed. and current signal Perform digital bandpass filtering separately, and set the passband center frequency of the filter to the nominal frequency of the power grid (e.g., ...). or To extract the pure voltage fundamental component. and the fundamental component of the current .

[0023] Subsequently, the phase difference was estimated using the zero-crossing detection method, the calculation logic of which is as follows: the fundamental voltage component was recorded respectively. and the fundamental component of the current The moment when a voltage value changes from negative to positive within a complete power grid cycle is denoted as the voltage zero-crossing moment. and the moment when the current crosses zero First, calculate the time difference between the two. As shown in equation (1): (1) Then, the fundamental phase difference is calculated based on the time difference value and the current fundamental frequency. As shown in equation (2): (2) in, Indicates the current fundamental frequency of the power grid (unit: Hertz). ).

[0024] In another alternative robust implementation, the cross-correlation method can also be used: by calculating... and The cross-correlation function is used to search for the discrete offset corresponding to the peak position of the cross-correlation function, which is then converted into the time delay. And the phase difference is obtained.

[0025] Taking a real-world application scenario of air conditioning equipment monitoring as an example: When the air conditioner starts up and enters a stable cooling state, due to the significant inductive load characteristics of the compressor motor, its fundamental current phase will lag significantly behind the voltage phase, resulting in the measured... Usually distributed in to (correspond to Between (radians). When the air conditioner stops running and enters standby mode, although its internal switching power supply may produce severe waveform distortion, the phase relationship between its fundamental component and the voltage is usually close to in phase ( It may exhibit weak capacitive characteristics. By accurately estimating the fundamental phase difference, a preliminary perception and judgment of the physical characteristics of the load can be achieved.

[0026] Step S103: Calculate the power factor based on the fundamental phase difference, the effective value of the fundamental current of the load current signal, and the effective value of the total current.

[0027] The fundamental current effective value refers to the effective value (RMS value) of the fundamental component in the load current signal, denoted as The total current RMS value refers to the effective value (RMS value) of the entire load current signal, including the fundamental frequency and all harmonic components, denoted as . Power factor (PF) is an important indicator for measuring the efficiency of electrical energy utilization. It is defined as the ratio of active power to apparent power.

[0028] Traditional power factor calculations only consider the cosine of the fundamental phase difference ( This can lead to significant errors when dealing with modern nonlinear loads (such as switching power supplies and frequency converters) because these devices introduce a large number of harmonics, causing the effective value of the total current to deteriorate. Much larger than the effective value of the fundamental current This step involves introducing... This distortion factor calculates the true power factor, which can more comprehensively and accurately reflect the actual power consumption characteristics of the device. This is crucial for distinguishing different types of low-power states (such as standby and shutdown).

[0029] In a specific computational logic, the load current signal is first analyzed using a Fast Fourier Transform (FFT) or a digital filter bank. The fundamental component is separated from the middle. .

[0030] The effective value of the fundamental current The calculation formula is shown in equation (3): (3) The effective value of the total current The calculation formula is shown in equation (4): (4) Then the fundamental phase difference obtained in step S102 is... Substitute in, and the final power factor The calculation formula is shown in equation (5): (5) in, Indicates the fundamental frequency period of the power grid (unit: seconds). Formula (5) clearly shows that the power factor is determined by the displacement factor ( ) and distortion factors ( (To be decided jointly)

[0031] In a specific application scenario, the active power of an LCD TV in standby mode may only be... However, due to the characteristics of its switching power supply, the effective value of the total current is... Possibly reach The effective value of the fundamental current Maybe only ,at the same time Approximately According to the above formula (5), its true power factor is... This is a very low value. In comparison, a purely resistive nightlight, even with the same wattage, would... ,That and The power factor is close to This significant difference provides a key criterion for subsequent state determination.

[0032] Step S104: Calculate the active power based on the grid voltage signal and the load current signal.

[0033] Active power refers to the average power actually consumed by the load and converted into other forms of energy (such as heat, light, and mechanical energy) in an AC circuit, denoted as P. It is the most direct and reliable physical quantity for determining whether a device is "doing work".

[0034] Calculating active power is fundamental to load monitoring. Regardless of the equipment's operational state, its actual energy consumption is reflected in its active power. While standby mode consumes less power, After the equipment is turned off, Therefore, accurate and real-time calculations are essential. It is the cornerstone for distinguishing between these two states.

[0035] In a standard digital calculation method, the instantaneous voltage value sequence synchronously acquired in step S101 is used. and instantaneous current value sequence First, calculate the instantaneous power sequence. The calculation formula is shown in equation (6): (6) Then, in one or more complete grid fundamental cycles Within the timeframe, the active power can be obtained by taking the arithmetic mean of the instantaneous power sequence. The calculation formula is shown in equation (7): (7) in, This represents the total number of sampling points within a calculation period.

[0036] This method is directly derived from the physical definition of active power and has high accuracy.

[0037] In a specific application scenario, when a user unplugs a TV that is in standby mode, the load current... It will instantly decay to zero. The electricity meter will detect this in the next calculation cycle. Almost all zero, resulting in It was almost all zero, and the final calculated value was... The value will change from the previous The power level drops sharply to the milliwatt level (primarily due to measurement noise). This The mutation is the most definitive signal for identifying a "device shutdown event".

[0038] Step S105: Determine the operating status of the target monitoring equipment based on active power and power factor.

[0039] The operating state here specifically refers to two macroscopic states that need to be distinguished: standby state (the device is powered on but does not perform main functions, maintaining low power consumption operation) and device shutdown event (the device is physically disconnected from the power supply, stopping all power-consuming activities).

[0040] Active power With power factor Combining these parameters for judgment is to overcome the limitations of relying on a single parameter. It cannot distinguish between a low-power resistive load (such as a nightlight) and a standby non-linear load; relying solely on... It also cannot distinguish between a highly distorted standby device and a circuit that is turned off but has a weak leakage current. The combination of the two constitutes a two-dimensional criterion space, which can effectively improve the robustness and accuracy of state recognition.

[0041] In a specific decision-making logic, two thresholds are preset: a first power threshold. (For example ) and second power factor threshold (For example The judgment rules are as follows: If the parameters monitored by the system in real time meet the judgment condition shown in equation (8), then the device is determined to be in standby mode: (8) like It continuously decreases from any level and stabilizes in a range close to zero (e.g., below) within a short period of time (such as within one fundamental frequency cycle). If this occurs, a device shutdown event is determined to have taken place. At this time, The value becomes meaningless because the current is too small, therefore... The absolute value is used as the main criterion.

[0042] In a specific application scenario, a smart home system aims to automatically identify television usage. When it detects… from (Viewing status) dropped to ,and from Down to In this way, it can accurately determine that the user has turned off the TV and entered standby mode, and trigger automated scenarios such as "entering power saving mode" accordingly. Subsequently, if it detects... from sudden drop In this way, it can be immediately determined that the TV plug has been unplugged, triggering a shutdown event and thus ending the power tracking of the device.

[0043] In the aforementioned embodiment, step S103 calculates the power factor using the fundamental phase difference, the effective value of the fundamental current, and the effective value of the total current. This is one of the key criteria for distinguishing between standby and off states. However, the accuracy of this calculation depends entirely on the precision of obtaining the effective values ​​of the fundamental current and the total current. If the calculation is based solely on a single grid cycle or the raw, unprocessed current signal, the result will be severely limited by noise, DC offset, and abundant harmonic components in the signal, leading to power factor distortion and subsequent misjudgment. To accurately separate and quantify the energy contributions of the fundamental and harmonics from the complex load current signal, it is necessary to introduce a systematic frequency domain analysis method. Figure 3 As shown, the specific implementation process includes steps S201 to S204: Step S201: Extract current data corresponding to multiple consecutive fundamental cycles of the power grid from the load current signal.

[0044] Multiple consecutive fundamental frequency cycles of a power grid refer to complete power frequency cycles that are consecutive in time and number more than one. For example, in a 50Hz power grid, one fundamental frequency cycle is 20 milliseconds. Extracting "four consecutive cycles" means extracting a continuous current signal segment with a duration of 80 milliseconds. The current data is the discrete instantaneous current value sequence obtained by sampling through an analog-to-digital converter (ADC) within this extracted time window.

[0045] Extracting data from multiple consecutive cycles is a prerequisite for accurate spectral analysis. A single cycle has a limited data length, insufficient spectral resolution, and is susceptible to random noise and transient interference, leading to unstable harmonic analysis results. By extracting multiple consecutive cycles, more information can be accumulated in the time domain, resulting in higher resolution and stronger noise resistance in the frequency domain, thus enabling more accurate separation of the fundamental frequency and its harmonic components.

[0046] Within a specific execution logic, the signal processing unit maintains a sliding time window. The length of this window is set to... One fundamental frequency cycle ( greater than integers, for example or Whenever new synchronous sampling data flows in, the window slides forward to always keep it containing the latest data. A complete cycle of load current signal sequence This load current signal sequence This will be used as input for subsequent frequency domain decomposition. Window length The choice requires a trade-off between computational complexity and analytical accuracy: The larger the value, the higher the frequency resolution, but the longer the computational latency.

[0047] Step S202: Perform frequency domain decomposition on each current data to obtain the amplitude of the fundamental component and the amplitude of each harmonic component of the load current signal.

[0048] Frequency domain decomposition is a mathematical process that converts a time-domain signal into its frequency components. In this context, it specifically refers to the extraction of intercepted time-domain current data. It is converted into complex numbers or amplitude-phase information at a series of discrete frequency points. The amplitude of the fundamental component is the signal amplitude at a frequency equal to the fundamental frequency of the power grid (e.g., 50Hz); the amplitudes of each harmonic component are the signal amplitudes at frequency points such as 100Hz (2nd), 150Hz (3rd), and 200Hz (4th).

[0049] The fundamental purpose of performing frequency domain decomposition is to decouple the complex and potentially severely distorted time-domain current waveform into a series of simple sinusoidal components. This decoupling allows us to independently quantify and analyze the fundamental and harmonic energies, which is the only accurate way to calculate the RMS values ​​of the fundamental and total currents. Directly processing the distorted waveform in the time domain cannot separate these two energy components with different properties.

[0050] In a typical algorithm implementation, the Fast Fourier Transform (FFT) is used as the core tool for frequency domain decomposition. First, the truncated... Current data for each cycle Windowing (such as Hanning windowing) is applied to reduce spectral leakage. Then, the windowed data is processed... Point FFT ( Usually The power of, and The output of the FFT is a complex array, whose... Each element corresponds to a frequency. spectral components at ( (For sampling frequency). (via index) ( By finding the fundamental frequency, the complex value of the fundamental component can be obtained, and its magnitude is the amplitude of the fundamental component. Similarly, through indexing... ( (where the harmonic order is ), can be obtained sequentially. Second-rate, Second-rate Second-rate( The preset maximum number of analyses, such as Harmonic component amplitude .

[0051] Step S203: Calculate the effective value of the fundamental current based on the amplitude of the fundamental component of the load current signal.

[0052] Effective value of fundamental current ( The effective value of a fundamental current is a physical quantity that measures the ability of a fundamental current to do work. It is defined as follows: If an equivalent direct current generates the same amount of heat through a resistor in one cycle as the fundamental alternating current generates in the same time period, then the value of this direct current is the effective value of the fundamental alternating current.

[0053] Based on the inherent relationship between the effective value and peak value (or amplitude) of a sinusoidal alternating current, for a pure sine wave, its effective value is equal to its amplitude divided by 1 / 2. Since step S202 has accurately separated the fundamental component through frequency domain decomposition and obtained its amplitude, Therefore, this relationship can be directly applied for calculation.

[0054] In a well-defined calculation formula, the effective value of the fundamental current... The calculation expression is:

[0055] in, This is the amplitude of the fundamental component obtained in step S202. The calculation process is simple and direct, and its accuracy depends entirely on the accuracy of the previous frequency domain decomposition.

[0056] Step S204: Calculate the effective value of the total current based on the amplitude of the fundamental component and the amplitude of each harmonic component.

[0057] RMS value of total current ( The energy of a current signal is a physical quantity that measures the energy of the entire current signal, including all frequency components (fundamental frequency + harmonics). According to Parseval's theorem, the total energy of a signal in the time domain is equal to its total energy in the frequency domain. This means that the square of the total effective value of the current is equal to the sum of the squares of the effective values ​​of each harmonic current (including the fundamental frequency).

[0058] Calculating the total effective value using the amplitudes of each component obtained in step S202 is an efficient and accurate method. It avoids the computational overhead of integrating or summing the original distorted waveform in the time domain, and naturally decomposes the total current into contributions from the fundamental and harmonic frequencies, which is highly consistent with the overall technical concept of this invention.

[0059] In a well-defined calculation logic, the amplitude of each harmonic component is first determined... ( from arrive ,in Calculate the effective value of each harmonic current (representing the fundamental frequency). Then, according to the principle of superposition of effective current values, the total effective current value is... The calculation formula is:

[0060] Will Substituting into the above formula, it can be simplified to:

[0061] This formula shows that it is only necessary to square, sum, and take the square root of the amplitudes of all analyzed harmonic components (including the fundamental frequency), and then multiply by a coefficient. This will give you the accurate RMS value of the total current.

[0062] In the aforementioned embodiment, step S105 effectively distinguishes between the standby and shutdown states of a device using a combination of active power and power factor criteria. However, in practical applications, when some devices enter deep sleep or execute a soft shutdown process, their active power gradually decreases to an extremely low level, and their power factor may also become low due to internal circuit characteristics. Their macroscopic performance is very similar to a physical shutdown event, making it difficult to reliably distinguish them based solely on these two static indicators. Furthermore, instantaneous fluctuations in the power grid can also cause brief jumps in power readings, leading to false alarms. To capture the transient dynamic characteristics during state transitions and establish a criterion more sensitive to sudden power outages, it is necessary to introduce time-series analysis of harmonic component stability. Figure 4 As shown, the specific implementation process includes steps S301 to S303: Step S301: For each harmonic component, calculate the rate of change of the harmonic component amplitude between adjacent fundamental periods of the power grid to obtain the rate of change sequence.

[0063] The rate of change of harmonic component amplitude refers to how quickly the amplitude of a specific harmonic (such as the 3rd or 5th harmonic) changes relative to the fundamental frequency of the power grid between two (or more) consecutive cycles. For example, the 3rd harmonic... The amplitude of each cycle is In the The amplitude of each cycle is Then its rate of change can be defined as Or its absolute value. A rate of change sequence is a sequence of rate of change values ​​calculated over a continuous time window for a specific harmonic.

[0064] The core purpose of calculating the rate of change of harmonic amplitude is to capture the dynamic stability of harmonic components. When an electrical device is operating stably (whether in operation or standby), its harmonic spectrum is usually relatively stable, with minimal variation in the amplitude of each harmonic between adjacent cycles. However, at the instant of a device shutdown event, the current rapidly decays to zero, causing a sharp and synchronous drop in the amplitude of all harmonic components within a very short time. By quantifying the severity of this change, a criterion highly sensitive to shutdown events can be constructed, effectively distinguishing between stable low-power states and sudden power outages.

[0065] In a specific computational logic, the signal processing unit continuously maintains a historical buffer queue of harmonic amplitudes. For each preset harmonic... (For example ), record its past A continuous cycle ( Amplitude within ) Each time a new cycle is completed... Harmonic analysis and obtained Then, calculate the latest rate of change of that harmonic:

[0066] in, It is a very small positive number (e.g.) This is used to prevent the denominator from being zero. Performing this operation for each monitored harmonic yields a set of parallel rate-of-change sequences. .

[0067] Step S302: Based on each rate of change sequence, generate dynamic harmonic distribution characteristics that reflect the stability of the harmonic components of the load current signal.

[0068] Dynamic harmonic distribution characteristics are a comprehensive index or vector that integrates the rate of change information of all monitored harmonics to globally characterize the overall stability or turbulence of the current load current harmonic components. This characteristic is no longer isolated information about a single harmonic, but rather a macroscopic manifestation of the coordinated changes of multiple harmonics.

[0069] The fundamental reason for generating this feature is that changes in a single harmonic can be affected by specific disturbances (such as instantaneous fluctuations in the power grid), leading to false alarms. However, a real equipment shutdown event will indiscriminately and synchronously affect all harmonic components generated by that equipment. Therefore, by aggregating the rates of change of multiple harmonics, the robustness and anti-interference capability of the criterion can be greatly improved, effectively filtering out occasional noise.

[0070] In a typical feature generation algorithm, various aggregation strategies can be employed. A simple and effective method is to calculate the root mean square (RMS) value of all rate-of-change sequences:

[0071] in It is the total number of harmonic orders monitored.

[0072] Another approach is to take the maximum of all rates of change:

[0073] Regardless of the aggregation method used, the final result is The smaller the value, the more stable the harmonics; the larger the value, the more volatile the harmonic components are, which is very likely caused by equipment shutdown. Step S303: Based on the active power, power factor and dynamic harmonic distribution characteristics, determine whether the target monitoring equipment has experienced an equipment shutdown event.

[0074] This step will apply the three core criteria obtained in the previous steps—active power. Power factor And the newly introduced dynamic harmonic distribution characteristics — The data is then integrated to make a final determination of the "device shutdown event". This is a multi-dimensional, cross-validated decision-making process.

[0075] Introduction This is to solve the problem of dependency and Potential gray areas exist. For example, some devices experience a brief transition phase before entering deep sleep, during which power consumption and harmonics rapidly decrease, which might be misinterpreted as a shutdown. A true physical shutdown, however, is characterized by instantaneous and complete harmonic oscillations. By requiring... It must exceed a high threshold to effectively eliminate interference from this type of transition state.

[0076] In a specific decision-making logic, a third threshold is preset. (For example, 0.8). The complete shutdown event determination rules are as follows: First, the active power is detected. A significant downward trend has emerged.

[0077] Then, examine the dynamic harmonic distribution characteristics. Does it exceed the preset third threshold? .

[0078] At the same time, confirm the active power at this time. It has fallen below the first power threshold. .

[0079] A device shutdown event is only determined to have occurred when all three conditions above are met simultaneously. Power factor This is for supplementary reference only, because at the moment of shutdown, It can become unstable due to insufficient current, and its value itself is not very meaningful, but its rapid deterioration trend can serve as evidence.

[0080] In the aforementioned embodiments, step S105 determines the equipment operating status based on active power and power factor, and step S303 further incorporates the dynamic stability characteristics of harmonics to improve the accuracy of shutdown event identification. However, the reliability of these determinations fundamentally depends on the estimation accuracy of the fundamental phase difference. In actual power grid environments, the phase difference estimated for a single cycle is highly susceptible to instantaneous interference such as high-frequency noise, harmonic contamination, or minute frequency fluctuations, leading to severe fluctuations in the results. Although step S201, which extracts data from multiple consecutive cycles, effectively improves the stability of the current RMS value calculation, this operation itself does not solve the problem of inconsistent phase estimation results across cycles. If such disturbed phase values ​​are simply averaged, outliers will still contaminate the final result. Therefore, a mechanism capable of intelligently identifying and suppressing abnormal phase estimations is urgently needed. Figure 5 As shown, the specific implementation process includes steps S401 to S402: Step S401: Perform phase difference estimation processing on the grid voltage signal and load current signal in multiple consecutive grid fundamental cycles to obtain multiple original time delay correlation functions and the preliminary phase difference corresponding to each original time delay correlation function.

[0081] The original time delay correlation function is a time-domain function obtained by analyzing the cross-correlation between voltage and current signals within a single grid fundamental cycle. Its main peak position directly corresponds to the time delay between the two. The preliminary phase difference is the phase angle calculated based on this time delay and a reference frequency (usually nominal 50Hz). In this step, instead of processing only an isolated cycle, multiple consecutive cycles (e.g., those extracted in step S201) are used. (Number) synchronized voltage-current data blocks, for each complete cycle of which a phase difference estimation sub-process is executed independently once (the specific implementation of this sub-process is defined by subsequent steps S501-S504), thereby outputting... One original time delay correlation function and its corresponding Initial phase difference .

[0082] The core purpose of multi-cycle independent estimation is to obtain data redundancy to cope with transient disturbances. Within any single power grid cycle, switching operations, lightning strikes, or the starting and stopping of other electrical equipment can introduce strong electromagnetic noise, causing the phase estimation results for that cycle to deviate significantly from the true value. By processing multiple consecutive cycles in parallel, multiple observation samples of the same physical quantity (true phase difference) are obtained, making it possible to subsequently identify and eliminate outlier observations.

[0083] In a specific execution logic, the signal processing unit prepares the signal in step S201. Each period of data is considered as a queue. It sequentially retrieves the data from the first period. One cycle ( from arrive The data is used to call a standardized phase estimation operator (i.e., the process in steps S501-S504), which returns the phase estimation data specific to that period. and All results are temporarily stored, awaiting the next step of confidence analysis.

[0084] Step S402: For each preliminary phase difference, determine the confidence weight of the preliminary phase difference based on the degree of deviation between the preliminary phase difference and the preliminary phase difference of the adjacent period.

[0085] Among them, the smaller the degree of deviation, the higher the confidence weight.

[0086] The confidence weight is a value between 0.0 and 1.0 used to quantify the reliability or credibility of a preliminary phase difference estimate. Its judgment logic is based on a simple yet effective assumption: the true fundamental phase difference is relatively stable over a short period (several power frequency cycles), therefore, the estimates for adjacent cycles should be highly consistent. Any estimate that significantly deviates from its neighboring values ​​is highly likely an outlier caused by transient interference.

[0087] The introduction of a confidence weighting mechanism is a key innovation in achieving high robustness in this scheme. It abandons the traditional simple averaging method (which includes outliers) and instead adopts an intelligent, adaptive weighting strategy that makes stable data "speak louder" and outliers "speak softer," thereby effectively filtering out random noise while preserving the true dynamics of the signal.

[0088] In a typical weight calculation algorithm, for non-boundary periods (Right now First, calculate its absolute deviation from the previous and next periods:

[0089]

[0090] Take the maximum of the two values ​​as the measure of instability at that point: Then, it is transformed into weights using a monotonically decreasing mapping function (such as an exponential function): .in, It is a positive scaling factor used to adjust the weights' sensitivity to bias. When Approaching hour, Approaching ;when When it is very large, Approaching For boundary periods ( or ( ), can be compared only with its only neighbor.

[0091] Step S403: Weight the multiple original time delay correlation functions according to their respective confidence weights to obtain the weighted time delay correlation function.

[0092] The weighted delay correlation function is obtained through step S401. The original time delay correlation function is used to obtain a comprehensive function by linearly weighting and summing the confidence weights calculated in step S402. Its mathematical expression is:

[0093] in: Represents a time delay variable; Representative periodic index ( ); The first value calculated in step S402 represents the value of the second value. Confidence weights for each period; The representative step S401 is the first The original time delay correlation function calculated for each cycle.

[0094] Choosing to fuse in the time-delay correlation function domain rather than the phase angle domain is another important design feature of this scheme. The time-delay correlation function contains richer information than a single phase value, such as the sharpness of the main peak and the size of the side lobes. Directly weighting the correlation functions preserves these detailed features. High-weighted correlation functions enhance the main peak, while the influence of low-weighted correlation functions on the final result is effectively suppressed. The final synthesized... It will have one more than any single The sharper and clearer main peaks provide a solid foundation for accurately extracting time delay.

[0095] Continuing with the refrigerator example, due to abnormal cycles... Its weight is extremely low; it has very little impact on the final result. It contributed almost nothing. And two normal cycles... and The correlation function is significantly strengthened due to its high weight. As a result, The main summit pointed very cleanly and accurately to the corresponding The actual time delay position of the phase difference is completely unaffected by startup interference.

[0096] Step S404: Extract the time delay corresponding to the main peak value from the weighted time delay correlation function, and convert the time delay into the fundamental phase difference based on the current fundamental frequency.

[0097] The current fundamental frequency is used as the frequency reference for phase difference calculation.

[0098] This step is the final output of high-precision phase estimation. First, in... The peak search algorithm is executed to locate the position of the global maximum value (main peak). The coordinates of this position on the horizontal axis (delay axis) are the desired comprehensive delay. Subsequently, a separate, high-precision frequency tracking module is invoked (its implementation can be referenced in zero-crossing detection and recursive filtering) to obtain the precise fundamental frequency at the current moment. Finally, using basic physical formulas... The final, robust fundamental phase difference was calculated. .

[0099] Use dynamic updates Instead of a fixed nominal value of 50Hz, this is to compensate for the actual frequency offset of the power grid (typically between 49.5-50.5Hz). Even a small frequency error (such as 0.1Hz) can lead to significant phase accumulation errors over a long period of integration. Therefore, [the following text is incomplete and requires further context: "…"] As a frequency reference, it is an indispensable part of ensuring long-term, high-precision phase estimation. This high precision... The value will be sent back to step S103 to calculate a more reliable power factor, thereby comprehensively improving the performance of the entire load monitoring.

[0100] In the aforementioned robust phase difference estimation scheme (steps S401-S404), the interference from abnormal periods is effectively suppressed by performing confidence-weighted fusion of the "original time delay correlation functions" for multiple consecutive cycles. However, the effectiveness of this scheme is highly dependent on the quality of the "original time delay correlation function" generated in each single cycle. If the main peak of the correlation function in a single cycle is blurred, the side lobes are too high, or there are spurious peaks, the accuracy of the fusion result will be greatly reduced even after subsequent weighting. The main reason for the poor quality of the single-cycle correlation function is that the high-order harmonic energy contained in the original voltage and current signals will seriously pollute the cross-correlation calculation, causing the time delay information to be submerged in noise. Therefore, in order to improve the signal-to-noise ratio of the correlation function in each cycle from the source, it is necessary to perform targeted purification of the signal spectrum before generating the correlation function, such as... Figure 6 As shown, the specific implementation process includes steps S501 to S504: Step S501: Using the nominal fundamental frequency of the power grid as the initial center, construct a continuous weighting function for focusing the fundamental frequency band.

[0101] A continuous weighting function is a real-valued, non-negative smooth window function W(f) defined in the frequency domain, whose main lobe center is precisely aligned with the nominal fundamental frequency of the power grid (e.g., 50Hz in China). The main lobe width of this function is designed to be a narrow band, typically covering only the fundamental frequency itself and the immediately adjacent very low harmonics (e.g., the third harmonic at 150Hz), while its side lobes decay rapidly to near zero. Common function forms include Gaussian windows, Hanning windows, or Kaiser windows.

[0102] The fundamental purpose of constructing this weighting function is to achieve frequency domain focusing. In real load currents, although the energy of higher harmonics (such as the 5th, 7th, and higher) is relatively small compared to the fundamental frequency, their absolute value is sufficient to form significant sidelobes in the cross-power spectrum. These sidelobes, after inverse transform, severely interfere with the main peak of the time delay correlation function, leading to phase estimation errors. By applying this narrowband weighting function, it is equivalent to constructing a high-Q bandpass filter in the frequency domain, selectively "amplifying" the fundamental frequency information and "suppressing" irrelevant harmonic noise, thereby creating a high signal-to-noise ratio environment for subsequent time delay estimation.

[0103] In a specific construction logic, assume the sampling frequency is... Performing a 128-point FFT on a 20ms (50Hz period) signal segment yields the following frequency resolution: At this point, the 50Hz fundamental frequency corresponds exactly to the first frequency point of the FFT (index). It is possible to construct a... A discrete Gaussian weighted sequence centered at a frequency of 3 points (covering 0-150Hz): Other frequency points For higher resolution FFTs, continuous Gaussian functions can be used. Perform interpolation calculations, where 50Hz Control bandwidth.

[0104] Step S502: Calculate the cross power spectrum of the grid voltage signal and the load current signal within the fundamental period of the grid.

[0105] Cross power spectrum describes the cross power spectrum of two random signals (in this case, voltage). and current The complex function of linear correlation in the frequency domain is denoted as . Its calculation method is based on the voltage signal spectrum. With current signal spectrum complex conjugate The product of, i.e. Phase angle of cross power spectrum It is directly equal to the phase difference between voltage and current at that frequency.

[0106] Calculating the cross-power spectrum is a core mathematical tool connecting time-domain observations with frequency-domain phase information. It transforms the cross-correlation relationship between two time-domain signals into the frequency domain, allowing us to directly read phase information at a specific frequency (such as the fundamental frequency). This is an indispensable intermediate step for achieving accurate phase estimation.

[0107] In a standard digital signal processing workflow, the synchronous sampling data within a single fundamental frequency cycle of the current power grid (e.g., 20ms) is first processed. and Execute separately Point Fast Fourier Transform (FFT) is used to obtain the discrete spectrum. and Then, regarding Taking the complex conjugate yields Finally, for each frequency point Performing complex multiplication: This yields the discretized cross-power spectrum. It fully contains amplitude and phase coupling information for all frequency components from DC to the Nyquist frequency.

[0108] Step S503: Use a continuous weighting function to weight the cross power spectrum to obtain the weighted cross power spectrum.

[0109] The weighted cross-power spectrum is the continuous weighting function of step S501. The original cross-power spectrum obtained in step S502 is applied. The result is denoted as This is a point-by-point multiplication operation that "shapes" the cross-power spectrum in the frequency domain.

[0110] The principle behind this weighting operation lies in the redistribution of frequency domain energy. Original cross-power spectrum. In this process, energy is distributed across multiple frequencies, including the fundamental frequency and harmonics. This is achieved through interaction with narrowband... Multiplication, only those located in The energy of frequencies within the main lobe (mainly the fundamental frequency) is preserved, while the energy of frequencies outside the main lobe (mainly higher harmonics) is significantly attenuated or even reduced to zero. This greatly simplifies the spectral structure, allowing the time-domain signal after the inverse transform to more purely reflect the time delay characteristics of the fundamental frequency.

[0111] In a well-defined computational logic, for discrete spectrum and discrete weighted sequences The weighted operation simplifies to: For all Execution. Due to It is zero or a very small value at most frequency points. It will become very "sparse", with almost all of its energy concentrated on the fundamental frequency and a few nearby frequencies. This sparsity is key to improving the accuracy of the subsequent inverse transform.

[0112] Step S504: Perform an inverse Fourier transform on the weighted cross-power spectrum to obtain the original time delay correlation function of the fundamental period of the power grid.

[0113] The original time delay correlation function is obtained by analyzing the weighted cross-power spectrum. The time-domain function obtained by performing the inverse Fourier transform (IFFT) .

[0114] The function's independent variable The cross-power spectrum represents the time delay, and its function value reflects the similarity between the voltage and current signals under that time delay. According to an extension of the Wiener-Khinchin theorem, the inverse transform of the cross-power spectrum is the cross-correlation function. Theoretically, The time delay corresponding to the global maximum value (main peak value) This is the actual time delay between the fundamental voltage wave and the fundamental current wave.

[0115] Performing an IFFT is a crucial step in remapping the phase information from the frequency domain back to the time domain. Due to the input... The IFFT results have been weighted and filtered out (high-order harmonics and broadband noise are removed). It will have a very sharp, prominent main peak and extremely low sidelobes. This makes The positioning becomes extremely accurate and robust, maintaining good performance even under conditions of low signal-to-noise ratio or nonlinear load interference.

[0116] In a typical IFFT implementation, the discrete weighted cross-power spectrum is... implement Point IFFT yields the discrete time-delay correlation function. ,in It is a time-delay index ( ). By searching maximum position and combined with the system sampling interval The overall delay can then be calculated. .this This will be used in subsequent steps, combined with the currently measured fundamental frequency, to ultimately convert it into a high-precision fundamental phase difference.

[0117] In the aforementioned embodiments, the macroscopic operating status of the device can be effectively determined through high-precision fundamental phase difference (step S404), reliable power factor (step S103), and dynamic stability characteristics of harmonics (step S302). However, this determination is geared towards a "black box" load and cannot identify the specific device in operation. In smart home or advanced energy management scenarios, if the device's identity can be further identified (e.g., "refrigerator of brand A" or "TV of model B"), more personalized services can be provided, such as customizing more precise standby / off thresholds for specific devices or generating detailed breakdowns of electricity consumption reports. Therefore, there is an urgent need for a mechanism that can utilize harmonic "fingerprints" for device identification and deeply integrate the identification results with status determination, such as... Figure 7 As shown, the specific implementation includes S601~S604: Step S601: Based on the amplitude of the fundamental component and the amplitude of each harmonic component of the load current signal, extract the complex component of the fundamental component and the complex component of the preset harmonic.

[0118] The fundamental complex component and harmonic complex components refer to a complete frequency domain representation that includes not only the amplitude information of each harmonic (including the fundamental wave) but also its phase information. During the frequency domain decomposition (such as FFT) in step S202, each frequency component... The output is a complex number. ,in The amplitude at that frequency. The phase angle is the fundamental frequency. This step extracts the fundamental frequency from the FFT result. ) and several preset harmonics (such as The complete complex number (secondary) .

[0119] The significance of extracting complex components lies in preserving the "vector" characteristics of the current signal at various frequency points. Through complex number operations, subsequent power feature extraction (such as the decomposition of active, reactive, and distorted power) can be performed more conveniently, which is the basis for achieving accurate equipment fingerprint modeling in non-intrusive load monitoring (NILM).

[0120] Extracting complex components, rather than just amplitude, is to construct a more informative and discriminative device feature vector. Even appliances of different brands and models, which produce similar harmonic amplitude spectra, may exhibit subtle but stable differences in the phase relationships of their various harmonics. This phase information constitutes the device's unique "harmonic fingerprint," which is crucial for achieving high-precision identification.

[0121] In a specific extraction logic, after completing the FFT in step S202, the signal processing unit directly reads the indexes of each characteristic frequency (such as the indices corresponding to the fundamental frequency and the 3rd, 5th, 7th, and 9th harmonics) from the output complex array. The complex values ​​at () are collectively represented by these complex numbers. dimension( The eigenvector of the total number of selected characteristic frequency components. This vector will serve as input for matching analysis against the template library.

[0122] Step S602: Match the fundamental complex component and harmonic complex components with the pre-established harmonic feature template library of typical electrical equipment, and calculate the similarity score of each equipment template.

[0123] The typical electrical equipment harmonic characteristic template library is a database that is pre-set before the smart terminal leaves the factory or dynamically built through cloud learning. Each record in the library (i.e., a "template") corresponds to a standard harmonic complex characteristic vector of a known device under its typical operating conditions. The matching process aims to quantify the feature vectors observed in real time. With each template in the library The degree of similarity between them.

[0124] The fundamental purpose of template matching is to achieve non-intrusive device identification. By comparing the frequency domain "fingerprint" of an unknown load with the "fingerprint profile" of a known device, the type of equipment operating in the current power circuit can be accurately inferred. This is the core prerequisite for achieving refined, itemized load monitoring.

[0125] In a standard matching algorithm implementation, cosine similarity is typically used as the metric. Considering that the feature vector is in complex form, for the th feature in the library... Each device template has a similarity score. The optimized calculation formula is as follows:

[0126] in, and They represent the real and imaginary parts of the extracted vector, respectively. " represents the dot product operation of vectors, Representing vectors Norm. The range of values ​​for cosine similarity is... The closer its value is to This indicates that the more consistent the directions of the two feature vectors are in the multidimensional space, the more similar the harmonic characteristics of the load to be identified are to the template.

[0127] Step S603: If the similarity score of a certain device template exceeds the preset similarity threshold, and the deviation between the theoretical fundamental phase and the fundamental phase difference corresponding to the device template is less than the preset phase tolerance, then the device is marked as a candidate operating device.

[0128] The first layer of verification is the similarity verification of harmonic features, which requires the currently calculated similarity score to meet the following conditions: (in, For example, take a preset similarity threshold. ).

[0129] The second layer of verification is the consistency verification of the fundamental phase characteristics: when each device template is entered into the database, in addition to recording the harmonic complex vector, its typical theoretical fundamental phase angle is also recorded. For example, motor-type loads typically exhibit inductive behavior, with a theoretical phase difference... Approximately (Current lags behind voltage); Switching power supply loads are approximately resistive under power factor correction (PFC) conditions. near .

[0130] The system will check the high-precision real-time fundamental phase difference obtained in step S404. Phase with the theoretical fundamental wave in the template The deviation between them is checked to determine whether its absolute value is less than the preset phase tolerance. (For example ), that is, judging inequalities Whether it is valid or not.

[0131] Through the above two verifications, the device is confirmed as the target device only when the frequency domain "fingerprint" is similar and the physical impedance characteristics (phase relationship) are consistent. This mechanism effectively reduces the false positive rate caused by relying solely on spectral similarity.

[0132] The dual verification is implemented to prevent mismatches. For example, some nonlinear loads may accidentally generate harmonic spectra similar to those of a certain device, but their fundamental phase characteristics (inductive / capacitive / resistive) must match those of the actual device. Only when both verifications pass is the device marked as a candidate for operation and proceeds to the next stage of cross-verification.

[0133] Step S604: Based on the information entropy value and dynamic harmonic distribution characteristics, cross-validate the candidate operating equipment; only when the cross-validation is passed, determine the operating status of the candidate operating equipment based on the active power and power factor, and use the operating status of the candidate operating equipment as the operating status of the target monitoring equipment.

[0134] Cross-validation here serves as the final safety valve. It utilizes two independent dynamic characteristics to verify the reasonableness of the candidate device's state: Information entropy value: reflects the regularity of the waveform of the weighted time delay correlation function.

[0135] Dynamic harmonic distribution characteristics (from step S302): reflect the stability of harmonic components.

[0136] Example of cross-validation rule: If the candidate device is a TV, and the current information entropy value is very low (indicating a stable signal) and the dynamic harmonic distribution characteristics also show stability, then the validation passes; if the information entropy value is very high (indicating a chaotic signal) but the dynamic harmonic distribution shows stability, then there may be a contradiction, and the validation fails.

[0137] Only when cross-validation passes will the device identification result be trusted, and the decision logic optimized for that specific candidate device will be invoked (for example, the standby power threshold of this TV is 1.2W, instead of the general 1W). Based on its active power P and power factor PF, a final, high-confidence operational status determination will be made. This marks a successful upgrade from "monitoring an unknown load" to "precisely managing a known device".

[0138] Based on the aforementioned device identification and cross-validation mechanism (step S604), the state determination can be performed for specific candidate operating devices. However, when a device is in an extremely low power consumption state (e.g., active power is below the first power threshold and power factor is below the second power factor threshold), its internal circuit operating mode may have subtle differences: in standby mode, the switching power supply typically turns on / off periodically at a fixed frequency, and the current waveform exhibits regularity; while when a device shutdown event occurs, the discharge of residual charge or contact arcing causes the current waveform to become random and chaotic. To capture this inherent regularity difference at the waveform level, it is necessary to introduce a feature that can quantify the "orderliness" of the signal. Information entropy, as a mature measurement tool, perfectly meets this need. Figure 8 As shown, the specific implementation process includes steps S701 to S703: Step S701: When the active power is lower than the first power threshold and the power factor is lower than the second power factor threshold, calculate the regularity information entropy value used to characterize the waveform of the weighted time delay correlation function.

[0139] This specifically refers to the weighted delay correlation function generated in step S403. After normalization, it is treated as a probability mass function, and its Shannon entropy is calculated. This entropy value directly reflects... The degree of energy concentration or the regularity of the geometric structure of the waveform.

[0140] The fundamental purpose of calculating information entropy is to distinguish between the "ordered" and "disordered" states of waveforms using mathematical methods. When a device is in standby mode, its internal switching power supply logic is stable, and the voltage and current signals exhibit a very strong cross-correlation. It will exhibit a very sharp, highly concentrated main peak, corresponding to a low entropy value. However, at the moment the device is turned off or during abnormal fluctuations, the current decays rapidly or is dominated by random noise, and the waveform of the correlation function becomes flat and diffuse, with the energy distribution tending to be uniform, thus leading to a significant increase in the calculated entropy value.

[0141] In a standard computational logic, the discrete weighted time delay correlation function is first processed. (in Normalize the time-delay index and construct a probability distribution sequence. :

[0142] Then, the information entropy is calculated according to the Shannon entropy definition formula. :

[0143] In order to ensure the robustness of the calculation, for The terms are selected according to mathematical conventions. The final scalar obtained This is the desired information entropy value, and its unit is bits.

[0144] Furthermore, if the information entropy value is lower than the third entropy threshold, then step S702 is executed; if the information entropy value is higher than the fourth entropy threshold, then step S703 is executed.

[0145] Step S702: Determine that the candidate operating device is in standby mode.

[0146] Step S703: Determine that a device shutdown event has occurred in the candidate operating device.

[0147] Among them, the third entropy threshold is less than the fourth entropy threshold.

[0148] The system presets two key judgment thresholds: the third entropy threshold. (e.g., taking 2.0 bits) and the fourth entropy threshold (For example, take 5.0 bits), and satisfy the following conditions: The specific judgment logic is as follows: Stable state determination: when the information entropy is calculated in real time When the cross-correlation function is in a highly ordered and concentrated state, it indicates that the target device is currently in a stable operating or standby state.

[0149] Shutdown status determination: when When the waveform distribution is extremely disordered and diffuse (with strong randomness), it indicates that the target device is in the shutdown process or has already been shut down.

[0150] Blurred area processing: If In between (i.e.) If the condition is not met, the result is "uncertain". In this case, the system can adopt a conservative strategy, such as maintaining the state determination of the previous moment, or triggering a longer period of observation logic to accumulate more feature samples.

[0151] By setting a non-overlapping threshold range that includes a "fuzzy zone," this decision logic possesses strong engineering robustness and can effectively filter out frequent jumps in critical values ​​(false judgments) caused by instantaneous power grid pulse noise or measurement errors. This refined decision result will serve as one of the key inputs for cross-validation in step S604 and will ultimately be used to output an accurate operating status report for the target monitoring equipment.

[0152] In the aforementioned high-precision phase difference estimation scheme (steps S401-S404), a precise "current fundamental frequency" is relied upon as the frequency reference for converting the time delay into a phase difference. This frequency is typically obtained through zero-crossing detection of the grid voltage. However, when the grid load fluctuates drastically or strong harmonic interference exists, the voltage waveform may be distorted, leading to errors in zero-crossing detection and consequently, deviations in frequency estimation. If this biased frequency is used for phase calculation, it will introduce systematic errors. Although the confidence weighting mechanism in step S402 can suppress some of the influence, fundamentally improving the accuracy of the frequency reference is more effective. Therefore, an adaptive mechanism capable of using harmonic dynamic information to sense and correct fundamental frequency drift is urgently needed. Figure 9 As shown, the specific implementation process includes steps S801 to S803: Step S801: For each fundamental frequency cycle of a power grid in a series of consecutive power grid fundamental frequency cycles, calculate the spectral centroid offset of each fundamental frequency cycle based on the dynamic harmonic distribution characteristics, and obtain the spectral centroid offset sequence.

[0153] The spectral centroid (SC) is the location of the "centroid" of the energy distribution in a signal's spectrum. For discrete spectra, its calculation formula is: ,in It is the first The physical frequency value of each frequency component. It is its corresponding spectral amplitude. Spectral centroid offset. This reflects the deviation between the measured energy centroid and the theoretically expected position.

[0154] In this embodiment, to reduce computational complexity, the process is not based on full-band frequency sweep, but cleverly reuses the harmonic amplitudes extracted in step S301. The core idea behind the calculation is that a tiny drift in the power grid frequency causes all higher harmonic frequencies to scale proportionally (i.e., the 1st harmonic frequency). The frequency shift of the subharmonic is the fundamental frequency shift. This (multiplied by 100) causes a significant shift in the center of gravity of the harmonic spectrum energy.

[0155] For the Each fundamental frequency cycle of the power grid, utilizing harmonic amplitude Its spectral centroid shift The approximate calculation logic is as follows: First, calculate the measured weighted average frequency, and then map it back to the fundamental frequency using the deviation value. The formula is expressed as follows: For the One grid fundamental frequency period, spectral centroid offset The calculation formula is revised as follows:

[0156] : No. Within the period, the The amplitude of the subharmonic is taken from the harmonic characteristic value extracted in step S301. : No. The measured center frequency of the second harmonic is obtained from the FFT transformation in step S202. The actual frequency value is calculated from the frequency bin where the subharmonic component is located and its adjacent energy distribution.

[0157] : No. The theoretical nominal frequency of the subharmonic. The nominal fundamental frequency of the power grid (in this embodiment, it is...) Therefore, the third harmonic should theoretically be in And so on.

[0158] : Calculate the highest harmonic order involved, which is a preset integer (e.g., or This determines the spectral width involved in the center of gravity evaluation.

[0159] This is the weighted total order, weighted by energy (the square of the amplitude). It is introduced to "normalize" the overall offset of the harmonic group back to the offset scale of the fundamental wave.

[0160] This formula essentially quantifies the degree to which the harmonic group deviates from the theoretical harmonic sequence. For continuous... Perform this operation every cycle to obtain the spectral centroid offset sequence. .

[0161] Step S802: If multiple consecutive spectral centroid offsets in the spectral centroid offset sequence exceed the preset offset threshold and have the same offset direction, then calculate the average value of the multiple consecutive spectral centroid offsets.

[0162] This step aims to distinguish between genuine frequency drift and random measurement noise. (Single cycle) Fluctuations may occur due to sampling jitter. This scheme uses trend consistency determination logic: only when the sequence is continuous... One (e.g.) The offsets all exceed a small preset offset threshold. (For example The system only confirms that a real frequency drift trend has occurred in the current power grid when their algebraic signs (positive and negative signs) are consistent.

[0163] Once the trend is confirmed, calculate this continuous Arithmetic mean of the offsets By averaging in the time domain, residual random interference can be effectively filtered out, providing a robust input estimate for frequency compensation.

[0164] Step S803: Based on the average value, the current fundamental frequency is updated gradually to obtain the updated current fundamental frequency.

[0165] To ensure the stability of the system's phase tracking and avoid algorithm lock loss or step noise caused by sudden changes in frequency parameters, this step employs a gradual update strategy to adjust the current fundamental frequency. The updated formula is as follows:

[0166] in, This is the smoothing factor (step size coefficient), and its value is usually much smaller than... (For example This is used to control the updated damping characteristics and ensure smooth convergence of frequency tracking.

[0167] Meanwhile, to ensure system security, a hard constraint mechanism is set up: the updated current fundamental frequency It must be limited to a preset safety range centered on the nominal frequency (e.g.) If the calculation result exceeds this range, it will be clamped to the range boundary value.

[0168] In the aforementioned embodiments, steps S401-S404 achieve robust phase difference estimation using a multi-cycle confidence-weighted time-delay correlation function method. However, this method has relatively high computational complexity, involving multiple FFT / IFFT and weighted fusion operations. On resource-constrained embedded platforms for electricity meters, providing a more computationally efficient and hardware-simple alternative would offer flexible options for products with varying cost and performance requirements. Orthogonal decomposition methods based on the digital phase-locked loop (DPLL) concept, due to their clear structure and low computational cost, have become an ideal candidate solution, such as... Figure 10 As shown, the specific implementation process includes steps S901 to S906: Step S901: Generate a digital local oscillator signal that is synchronized with the current fundamental frequency.

[0169] The digital local oscillator signal includes in-phase and quadrature components. The digital local oscillator signal is a pair of frequencies that are correlated with the current fundamental frequency of the power grid. A strictly synchronized discrete sine and cosine sequence. Its in-phase components are: Orthogonal components are ,in The sampling interval is... This is the sampling point index. This pair of signals constitutes a complex exponential signal. The real and imaginary parts.

[0170] Generating a synchronous local oscillator signal is fundamental to orthogonal decomposition. It acts as a "frequency probe," coherently extracting the fundamental component with the same frequency as the local oscillator from a complex mixed signal. When the local oscillator frequency... When the frequency is exactly the same as the true fundamental frequency of the signal, the extracted component, after filtering, becomes a pure DC signal; if the frequencies are inconsistent, low-frequency beat fluctuations will occur. In this embodiment, an internally maintained numerically controlled oscillator (NCO) is used, which adjusts according to the current fundamental frequency. (This frequency can be obtained in real time by zero-crossing detection and recursive filtering in step S10) The phase accumulation step size is dynamically adjusted to generate a high-precision synchronization sequence.

[0171] Step S902: Multiply the load current signal by the in-phase component and the quadrature component of the digital local oscillator signal respectively to obtain the fundamental in-phase component and the fundamental quadrature component of the load current signal.

[0172] This step performs the mixing operation in coherent demodulation. It converts the load current signal... respectively with and Multiplying them yields two product sequences:

[0173] According to the principle of product-sum-difference, these two product signals contain the "sum frequency" and "difference frequency" components of each frequency component in the original current signal and the local oscillator frequency. Among them, the difference frequency term generated by the fundamental current component and the local oscillator signal is located near zero frequency (DC), while other higher harmonics and the sum frequency term are located in the high-frequency range.

[0174] Step S903: Multiply the grid voltage signal by the in-phase component and the quadrature component of the digital local oscillator signal respectively to obtain the fundamental in-phase component and the fundamental quadrature component of the grid voltage signal.

[0175] Similarly, for the grid voltage signal Perform the same coherent demodulation operation:

[0176] At this point, the voltage and current signals are projected onto the complex plane in-phase (I) and quadrature (Q) coordinate axes based on the local oscillator frequency.

[0177] Step S904: Determine the fundamental phase of the grid voltage based on the in-phase component and the quadrature component of the fundamental voltage signal.

[0178] After mixing, a low-pass filter (LPF) is used to suppress high-frequency components (such as second harmonics and higher-order terms) in the product signal, retaining only the difference frequency DC component. After filtering, the sequence... and It will converge to a stable DC value and Together, they constitute the complex representation of the fundamental voltage of the power grid. .

[0179] grid voltage fundamental phase The argument of this complex number can be calculated using the arctangent function:

[0180] in It is the arctangent function in the fourth quadrant, which can be determined according to... and The sign of the angle determines the exact quadrant to which it belongs. arrive ).

[0181] Step S905: Determine the fundamental phase of the load current based on the in-phase component and the quadrature component of the fundamental current signal.

[0182] Similarly, for the product sequence of currents and The same low-pass filtering process is performed to obtain a stable DC value. and The complex representation of the fundamental wave of the current The corresponding fundamental phase of the load current. for:

[0183] Step S906: Calculate the difference between the fundamental phase of the grid voltage and the fundamental phase of the load current to obtain the fundamental phase difference.

[0184] Finally, the fundamental phase difference Defined as the difference between the initial phase of the voltage and the initial phase of the current:

[0185] The core advantage of this method lies in the fact that once the local oscillator signal is frequency-synchronized with the grid fundamental frequency, all operations can be performed in the time domain through multiplication and addition, greatly avoiding the high computational overhead and spectral leakage problems associated with the sliding window Fast Fourier Transform (FFT). Its estimation accuracy primarily depends on the local oscillator frequency. This approach improves real-time tracking accuracy and designs the cutoff frequency of the low-pass filter. It provides an efficient and robust phase difference estimation path.

[0186] In the aforementioned embodiments (such as steps S404, S803, and S901), it is repeatedly emphasized that a high-precision "current fundamental frequency" is needed as a key parameter. This frequency is not only the benchmark for phase difference calculation but also the source for digital local oscillator signal generation and frequency adaptive updates. Although the nominal frequency of the power grid is 50Hz, in actual operation, its frequency will dynamically fluctuate within a small range (such as ±0.5Hz). If the nominal value is used directly, it will introduce a non-negligible cumulative error. Therefore, a dedicated module capable of tracking the instantaneous frequency of the power grid in real time, smoothly, and accurately must be deployed. Zero-crossing detection combined with recursive filtering is a classic solution that balances accuracy and efficiency, such as... Figure 11 As shown, the specific implementation process includes steps S1001 to S1003: Step S1001: Perform zero-crossing detection on the power grid voltage signal to obtain a rough estimate of the power frequency cycle.

[0187] Zero-crossing detection refers to the real-time monitoring of voltage signals in a digital power grid. The instantaneous moment when a negative value crosses into a positive value (i.e., a positive zero crossing). Each positive zero crossing event marks the beginning of a complete fundamental frequency cycle of the power grid. This is achieved by recording the time interval between two consecutive positive zero crossing events (in terms of the number of sampling points). If calculated, then This will give you a rough observation value of the power frequency cycle. .

[0188] Zero-crossing detection is the most intuitive and computationally inexpensive method for obtaining the power grid frequency, and it is easy to implement in embedded microcontrollers. However, this method is extremely sensitive to broadband noise. Any high-frequency glitches or nonlinear harmonic interference on the voltage waveform can lead to false zero-crossing determinations, thus causing... The violent shaking. Therefore, this embodiment employs comparison logic with hysteresis: only when... From below The level rises to higher than Only after the level reaches a certain value is it considered a valid positive zero-crossing event (where...). (This is a preset noise immunity threshold). This mechanism effectively filters out tiny random oscillations near the zero point, ensuring the stability of the observations.

[0189] Step S1002: Take the rough power frequency period estimate as the observation input, perform state estimation through a recursive filtering algorithm, and output the smoothed instantaneous frequency estimate.

[0190] To overcome the inherent noise of zero-crossing observations, this step introduces a recursive filtering algorithm for the sequence. Real-time smoothing is performed. This algorithm treats the grid frequency as a slowly changing state variable, while... The input is considered to contain observation noise. By establishing a state-space model, the filter can fuse historical trends with current observations, outputting a high signal-to-noise ratio instantaneous frequency estimate. .

[0191] The advantage of recursive filtering lies in achieving a balance between dynamic tracking speed and steady-state smoothness. In an efficient implementation, a first-order low-pass recursive filter is used: in, For example, take the smoothing coefficient (e.g., take...) This is used to balance response time and noise suppression effectiveness. Then, the instantaneous frequency is calculated. For scenarios requiring higher accuracy, a Kalman filter can be used to model the period as a state equation. The observation equation is modeled as Kalman filters can detect process noise. With observation noise The statistical characteristics are automatically optimized to improve the filter gain, thereby providing the best linear unbiased frequency estimation in complex power grid environments.

[0192] Step S1003: Estimate the instantaneous frequency and use it as the current fundamental frequency.

[0193] After depth smoothing by recursive filter As a stable, low-noise, and real-time parameter, it was adopted as the system's "current fundamental frequency." This parameter serves as a globally shared variable, benefiting all frequency-dependent calculation modules within the system. In step S404: the time delay calculated by cross-correlation is accurately converted into electrical angular phase difference to eliminate the transformation error caused by frequency drift.

[0194] In step S803: the reference frequency is used as a reference value, and correction feedback based on the spectral centroid offset is received to realize dual frequency closed-loop tracking.

[0195] In step S901: This is used to drive a numerically controlled oscillator (NCO) in real time to generate a quadrature digital local oscillator signal with continuous phase and synchronized frequency. and .

[0196] Based on the same inventive concept, this application also provides a load monitoring device for an electricity meter to implement the load monitoring of an electricity meter as described above. The solution provided by this device is similar to the solution described in the above method. Therefore, the specific limitations of one or more embodiments of a load monitoring device for an electricity meter provided below can be found in the limitations of the load monitoring method for an electricity meter described above, and will not be repeated here.

[0197] Each module in the load monitoring device of the aforementioned electricity meter can be implemented entirely or partially through software, hardware, or a combination thereof. These modules can be embedded in the processor of a computer device in hardware form or independent of it, or stored in the memory of a computer device in software form, so that the processor can call and execute the corresponding operations of each module.

[0198] In one exemplary embodiment, a smart energy meter is provided, including a memory and a processor. The memory stores a computer program, and the processor executes the computer program to implement the load monitoring method of the energy meter described above.

[0199] In one embodiment, a computer-readable storage medium is provided having a computer program stored thereon, which determines that when the computer program is executed by a processor, it implements the above-described load monitoring method for an energy meter.

[0200] In one embodiment, a computer program product is provided, including a computer program that, when executed by a processor, implements the above-described load monitoring method for an electricity meter.

[0201] Those skilled in the art will understand that all or part of the processes in the methods of the above embodiments can be implemented by a computer program instructing related hardware. The computer program can be stored in a non-volatile deterministic machine-readable storage medium. When executed, the computer program can include the processes of the embodiments of the above methods. Any references to memory, databases, or other media used in the embodiments provided in this application can include at least one of non-volatile memory and volatile memory. Non-volatile memory can include read-only memory (ROM), magnetic tape, floppy disk, flash memory, optical memory, high-density embedded non-volatile memory, resistive random access memory (ReRAM), magnetic random access memory (MRAM), ferroelectric random access memory (FRAM), phase change memory (PCM), graphene memory, etc. Volatile memory can include random access memory (RAM) or external cache memory, etc. By way of illustration and not limitation, RAM can take many forms, such as Static Random Access Memory (SRAM) or Dynamic Random Access Memory (DRAM). The databases involved in the embodiments provided in this application may include at least one type of relational database and non-relational database. Non-relational databases may include, but are not limited to, blockchain-based distributed databases. The processors involved in the embodiments provided in this application may be general-purpose processors, central processing units, graphics processing units, digital signal processors, programmable logic devices, quantum computing-based data processing logic devices, artificial intelligence (AI) processors, etc., and are not limited to these.

[0202] The technical features of the above embodiments can be combined in any way. For the sake of brevity, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this application.

[0203] The embodiments described above are merely illustrative of several implementation methods of this application, and while the descriptions are specific and detailed, they should not be construed as limiting the scope of this patent application. It should be noted that those skilled in the art can make various modifications and improvements without departing from the concept of this application, and these all fall within the protection scope of this application. Therefore, the protection scope of this application should be determined by the appended claims.

Claims

1. A method for load monitoring of an electricity meter, characterized in that, The method includes: Simultaneously acquire grid voltage and load current signals; The phase difference between the grid voltage signal and the load current signal is estimated to obtain the fundamental phase difference. The power factor is calculated based on the fundamental phase difference, the effective value of the fundamental current of the load current signal, and the effective value of the total current. Calculate the active power based on the grid voltage signal and the load current signal; Based on the active power and the power factor, the operating status of the target monitoring device is determined; wherein, the operating status includes standby status or device shutdown event.

2. The method as described in claim 1, characterized in that, Determining the fundamental current RMS value and the total current RMS value of the load current signal specifically includes: Extract current data corresponding to multiple consecutive fundamental cycles of the power grid from the load current signal; For each current data point, frequency domain decomposition is performed to obtain the amplitude of the fundamental component and the amplitude of each harmonic component of the load current signal. The effective value of the fundamental current is calculated based on the amplitude of the fundamental component of the load current signal; The effective value of the total current is calculated based on the amplitude of the fundamental component and the amplitudes of each harmonic component.

3. The method as described in claim 2, characterized in that, The determination of the operating status of the target monitoring equipment based on the active power and the power factor specifically includes: For each harmonic component, the rate of change of the harmonic component amplitude between adjacent fundamental periods of the power grid is calculated to obtain the rate of change sequence. Based on each rate of change sequence, dynamic harmonic distribution characteristics reflecting the stability of the harmonic components of the load current signal are generated; Based on the active power, the power factor, and the dynamic harmonic distribution characteristics, it is determined whether the target monitoring equipment has experienced a shutdown event.

4. The method as described in claim 2, characterized in that, The step of estimating the phase difference between the grid voltage signal and the load current signal to obtain the fundamental phase difference specifically includes: For the grid voltage signal and load current signal in multiple consecutive grid fundamental cycles, phase difference estimation processing is performed to obtain multiple original time delay correlation functions and the preliminary phase difference corresponding to each original time delay correlation function; For each initial phase difference, the confidence weight of the initial phase difference is determined based on the degree of deviation between the initial phase difference and the initial phase difference of the adjacent period; where the smaller the degree of deviation, the higher the confidence weight. The multiple original time delay correlation functions are weighted and superimposed according to their respective confidence weights to obtain a weighted time delay correlation function. The time delay corresponding to the main peak value is extracted from the weighted time delay correlation function, and the time delay value is converted into the fundamental frequency phase difference based on the current fundamental frequency; wherein the current fundamental frequency is used as the frequency reference for the phase difference calculation.

5. The method as described in claim 4, characterized in that, For the grid voltage signal and load current signal within each grid fundamental cycle, phase difference estimation is performed to obtain the original time delay correlation function for that grid fundamental cycle, specifically including: Using the nominal fundamental frequency of the power grid as the initial center, a continuous weighting function for focusing the fundamental frequency band is constructed; Calculate the cross power spectrum of the grid voltage signal and the load current signal within the fundamental period of the grid; The cross-power spectrum is weighted using the continuous weighting function to obtain a weighted cross-power spectrum; Perform an inverse Fourier transform on the weighted cross-power spectrum to obtain the original time delay correlation function of the fundamental period of the power grid.

6. The method as described in claim 3, characterized in that, The determination of the operating status of the target monitoring equipment based on the active power and the power factor specifically includes: Based on the amplitude of the fundamental component and the amplitude of each harmonic component of the load current signal, the complex component of the fundamental component and the complex component of the preset harmonic are extracted. The fundamental complex component and harmonic complex component are matched with a pre-established typical electrical equipment harmonic feature template library, and the similarity score of each equipment template is calculated. If the similarity score of a certain device template exceeds the preset similarity threshold, and the deviation between the theoretical fundamental phase corresponding to the device template and the fundamental phase difference is less than the preset phase tolerance, then the device is marked as a candidate operating device. Based on the information entropy value and the dynamic harmonic distribution characteristics, the candidate operating devices are cross-validated. Only when the cross-validation passes, the operating status of the candidate operating device is determined based on the active power and the power factor, and the operating status of the candidate operating device is used as the operating status of the target monitoring device.

7. The method as described in claim 6, characterized in that, The step of determining the operating status of the candidate operating equipment based on the active power and the power factor specifically includes: When the active power is lower than a first power threshold and the power factor is lower than a second power factor threshold, calculate the regularity entropy value used to characterize the waveform of the weighted time delay correlation function: If the information entropy value is lower than the third entropy threshold, then the candidate operating device is determined to be in standby mode. If the information entropy value is higher than the fourth entropy threshold, it is determined that the candidate operating device has experienced a device shutdown event; wherein the third entropy threshold is less than the fourth entropy threshold.

8. The method as described in claim 2, characterized in that, After generating dynamic harmonic distribution characteristics that reflect the stability of the harmonic components of the load current signal, the method further includes a frequency update step, comprising: For each fundamental period of the power grid in the series of consecutive fundamental periods, the spectral centroid offset of each fundamental period is calculated based on the dynamic harmonic distribution characteristics to obtain a spectral centroid offset sequence. If multiple consecutive spectral centroid offsets in the spectral centroid offset sequence exceed a preset offset threshold and have the same offset direction, then calculate the average value of the multiple consecutive spectral centroid offsets. Based on the average value, the current fundamental frequency is updated gradually to obtain the updated current fundamental frequency; wherein the updated current fundamental frequency does not exceed a preset safety range centered on the nominal fundamental frequency of the power grid.

9. The method as described in claim 1, characterized in that, The step of estimating the phase difference between the grid voltage signal and the load current signal to obtain the fundamental phase difference specifically includes: Generate a digital local oscillator signal synchronized with the current fundamental frequency; wherein the digital local oscillator signal includes in-phase components and quadrature components; The load current signal is multiplied by the in-phase component and the quadrature component of the digital local oscillator signal, respectively, to obtain the fundamental in-phase component and the fundamental quadrature component of the load current signal. The grid voltage signal is multiplied by the in-phase component and the quadrature component of the digital local oscillator signal, respectively, to obtain the fundamental in-phase component and the fundamental quadrature component of the grid voltage signal. The fundamental phase of the grid voltage is determined based on the in-phase and quadrature components of the fundamental voltage signal. The fundamental phase of the load current is determined based on the in-phase component and the quadrature component of the fundamental current signal. The fundamental phase difference is obtained by calculating the difference between the fundamental phase of the grid voltage and the fundamental phase of the load current.

10. The method as described in claim 4 or 9, characterized in that, Obtain the current fundamental frequency, specifically including: Zero-crossing detection is performed on the power grid voltage signal to obtain a rough estimate of the power frequency period; The rough power frequency period estimate is used as the observation input, and the state is estimated by a recursive filtering algorithm to output a smoothed instantaneous frequency estimate. The instantaneous frequency estimate is used as the current fundamental frequency.

Images