An intelligent power load identification method

Abstract

The present invention discloses an intelligent power load identification method. The present invention relates to the technical field of power load identification, and solves the technical problem that power load identification cannot be effectively performed based on the extracted features, resulting in errors in identification. The present invention preprocesses the acquired power data to ensure the accuracy of subsequent data usage. Secondly, when extracting and identifying features, it calculates the corresponding effective values of voltage and current, calculates the corresponding crest factor according to the current effective value, and establishes a corresponding identification interval based on the crest factor. Subsequently, it calculates the crest factor according to the real-time features and matches it with the identification interval. During the matching process, it comprehensively analyzes the dynamic features to identify the power load, improves the overall identification accuracy, and avoids interference caused by similar features to the identification.

Description

Technical Field

[0001] The present invention relates to the technical field of power load identification, and particularly to an intelligent power load identification method. Background Art

[0002] Power load power consumption monitoring is the first step in carrying out the systematic project of energy conservation. Because only by clearly understanding how electric energy is utilized and consumed can effective energy conservation control measures and more efficient power consumption methods be found.

[0003] According to the patent with the publication number CN112968519A, an intelligent power load identification method is disclosed. It is not necessary to install sensors in front of each load, and lossless low-cost load identification can be achieved. This method realizes the change of the total circuit load by adding or removing electrical appliances in the circuit, records the change amount x(i) of the electrical parameters of the total incoming line of the circuit when the load changes, and at the same time records the current circuit load situation corresponding to x(i) as the graph node data g(i). Perform a fast Fourier transform on the graph node data g(i) to convert it into graph node frequency domain data h(i) with a fixed dimension. x(i) and h(i) form a training sample. A self-recurrent neural network is constructed. The input of the self-recurrent neural network is the change amount x(i) of the electrical parameters at the i-th moment and the graph node frequency domain data h(i - 1) at the (i - 1)-th moment, and the output is the graph node frequency domain data h(i) at the i-th moment; the self-recurrent neural network is cyclically trained with the training sample, and the trained self-recurrent neural network can be used for intelligent power load identification.

[0004] The above patent obtains the change amount of electrical parameters from the total incoming line of the circuit and obtains the current circuit load situation, and forms these data into a training sample to train the self-recurrent neural network, so that the self-recurrent neural network can reflect the mapping relationship between the change amount of electrical parameters and the circuit load situation. Through such a deep learning method, it is not necessary to install sensors in front of each load, and only the total electrical parameter data of the circuit needs to be obtained to realize the intelligent identification of the load.

[0005] However, when some existing power load identification systems are in use, on the one hand, due to the accuracy error in the data acquisition process, there is an error in the subsequent power load identification. On the other hand, the electrical characteristics of some loads may not be obvious or may be similar, resulting in difficulty in extracting effective distinguishing features. For example, some high-efficiency motors and traditional motors may be very close in terms of power factor, harmonic content, etc., and it is difficult to accurately distinguish them only through these common electrical characteristics. Summary of the Invention

[0006] Aiming at the deficiencies of the prior art, the present invention provides an intelligent power load identification method, which solves the problem that power load identification cannot be effectively carried out according to the extracted features, resulting in errors in identification.

[0007] To achieve the above object, the present invention is realized through the following technical solutions: An intelligent power load identification method, which specifically includes the following steps:

[0008] S101: Collect power data of the power system through a variety of sensors, and perform preprocessing operations on the power data to obtain preprocessed data;

[0009] S102: Perform corresponding feature extraction on the obtained preprocessed data, calculate the corresponding effective voltage and effective current based on time-domain features, and calculate the crest factor according to the effective current value;

[0010] S103: Analyze historical data, classify loads according to the crest factor to obtain the characteristics of the same type of load, and generate a peak matching interval according to the characteristics of the same type of load. Then, calculate the harmonic value corresponding to the peak matching interval;

[0011] S104: Calculate the corresponding real-time crest factor according to the obtained real-time features, match it with the peak matching interval to obtain a matching result, and determine the specific load reason according to the number of matching results, and generate load reason information;

[0012] S105: Perform dynamic feature extraction according to the load reason information, and perform verification analysis according to the current rise rate in the obtained dynamic features, and generate a verification result, where the verification result includes a correct verification result and an incorrect verification result. Then, perform secondary analysis on the incorrect verification result according to the dynamic features to generate specific reason information.

[0013] As a further solution of the present invention: The specific method for performing preprocessing operations on the power data to obtain preprocessed data in step S101 is:

[0014] Obtain power data, including voltage and current, and the voltage and current are their corresponding time series data. Then, perform filtering processing on the obtained power data to obtain preprocessed data.

[0015] As a further solution of the present invention: The specific method for calculating the crest factor according to the effective current value in step S102 is:

[0016] Obtain preprocessed data, define the time period T at the same time, and obtain the unit time voltage value and unit time current value corresponding to the time period T. Then, obtain the corresponding effective value by calculating the square root of the average value of the square of the current or voltage within the time period t, and obtain the peak value Fmax corresponding to the current signal within the time period T, and calculate the ratio of the peak value to the current effective value to obtain the crest factor.

[0017] As a further solution of the present invention: The specific method for generating the peak matching interval according to the characteristics of the same type of load in step S103 is as follows:

[0018] Obtain historical data, then extract the load characteristics corresponding to the load type in the historical data, and at the same time perform the same type classification processing on the load type to obtain the same type of load characteristics, and obtain the peak factor corresponding to the same type of load characteristics. Then generate the same type of peak matching interval of the same type of load characteristics according to the peak factor.

[0019] As a further solution of the present invention: The specific method for calculating the harmonic value corresponding to the peak matching interval in step S103 is as follows:

[0020] Judge whether there are multiple groups of the same peak factors in the same type of peak matching interval. If there are multiple groups of the same peak factors, mark the corresponding same type of peak matching interval and record it as the interval to be analyzed. If there are no multiple groups of the same peak factors, no processing is performed;

[0021] Obtain all the intervals to be analyzed, then calculate the amplitude and phase of the harmonics corresponding to the same peak factor in the intervals to be analyzed, and use the calculated amplitude and phase as the distinguishing characteristics of the corresponding load type.

[0022] As a further solution of the present invention: The specific method for obtaining the matching result in step S104 is as follows:

[0023] Obtain the real-time characteristics, calculate the corresponding real-time peak factor according to the real-time characteristics, and at the same time match the obtained real-time peak factor with the peak matching interval;

[0024] Match the real-time peak factor with multiple peak matching intervals, screen out the corresponding peak matching intervals that meet the requirements and record them as the marked intervals, and then match the real-time peak factor with the marked intervals, and screen out the load types with the same real-time peak factor and record them as the load types to be analyzed.

[0025] As a further solution of the present invention: The specific method for determining the specific load cause according to the number of matching results and generating the load cause information in step S104 is as follows:

[0026] When there is only one load type to be analyzed, record the corresponding load type to be analyzed as the standard load, and at the same time generate the load cause information;

[0027] When there are multiple load types to be analyzed, calculate the corresponding harmonic amplitude and phase according to the real-time characteristics, and match the calculated harmonic amplitude and phase with the amplitude and phase of the load types to be analyzed to obtain the specific load type, and at the same time generate the load cause information.

[0028] As a further solution of the present invention: The specific manner of generating the verification result in the step S105 is as follows:

[0029] Analyze the corresponding load type in the load cause information, calculate the current rise rate corresponding to the load type, and the calculation method is:

[0030] Use a current transformer and a data acquisition device to sample the current during the load startup process. Let the acquired current sequence be i(a), where a = 1, 2,..., b - 1, and a is the number of sampling points, and the corresponding time sequence is t(a). Then calculate the current difference between two adjacent sampling points Δi(a) = i(a + 1) - i(a), and the time interval between two adjacent sampling points Δt(a) = t(a + 1) - t(a). Substitute the calculated current difference and time interval into the formula Calculate the current rise rate r(a);

[0031] Then compare and judge the calculated current rise rate with the load cause information. If the current rise rate is the same as the current rise rate corresponding to the load cause information, it means that the load identification is correct, and a verification correct result is generated. On the contrary, if the current rise rate is different from the current rise rate corresponding to the load cause information, it means that the load identification is incorrect, and a verification error result is generated. Then, match the specific load based on the current rise rate as the standard, and at the same time generate specific cause information.

[0032] The present invention provides an intelligent power load identification method. Compared with the prior art, it has the following beneficial effects:

[0033] The present invention preprocesses the acquired power data to ensure the accuracy of subsequent data use. Secondly, when extracting and identifying features, by calculating the effective values corresponding to voltage and current, calculating the corresponding crest factor according to the current effective value, and establishing a corresponding identification interval according to the crest factor. Subsequently, calculate the crest factor according to the real-time features and match it with the identification interval. During the matching process, analyze the comprehensive dynamic features to identify the power load, improve the overall identification accuracy, and avoid interference caused by similar features to the identification. BRIEF DESCRIPTION OF THE DRAWINGS

[0034] Figure 1 It is a schematic diagram of the method steps of the present invention. DETAILED DESCRIPTION OF THE INVENTION

[0035] Next, the technical solutions in the embodiments of the present invention will be clearly and completely described in conjunction with the accompanying drawings in the embodiments of the present invention. Obviously, the described embodiments are only a part of the embodiments of the present invention, rather than all the embodiments. All other embodiments obtained by those of ordinary skill in the art based on the embodiments of the present invention without creative efforts shall fall within the protection scope of the present invention.

[0036] Embodiment 1. Please refer to Figure 1 , this application provides an intelligent power load identification method, which specifically includes the following steps.

[0037] Next, in step S101, power data of the power system is collected through multiple sensors, and the power data is preprocessed to obtain preprocessed data.

[0038] First, obtain the power data, where the power data includes the voltage and current corresponding to the power system, and the collected voltage and current here are their corresponding time series data. Usually, sensors (such as current transformers and voltage transformers) are used to collect at a relatively high sampling frequency (for example, hundreds to thousands of samples per second). At the same time, the current data reflects the change in the charge passing through the load, and the voltage data represents the potential difference across the circuit. Then, preprocess the obtained power data. The specific preprocessing operation is mainly to filter the power data. Common filters include low-pass filters, high-pass filters, and band-pass filters, etc. For example, a simple moving average filter or a digital filter (such as a Butterworth filter) is used to smooth the data, and further preprocessed data is obtained after filtering.

[0039] Next, in step S102, corresponding feature extraction is performed on the obtained preprocessed data. Calculate the effective voltage and effective current based on time-domain features, and calculate the crest factor according to the effective current value.

[0040] Obtain the preprocessed data, and at the same time define the time period T, and obtain the unit time voltage value and unit time current value corresponding within the time period T. Then, calculate the root mean square of the average value of the square of the current or voltage within the calculation time period t to obtain the corresponding effective value. For example, for the current signal i(t), its effective value The calculation method for the effective value of voltage is the same as that for the effective value of current. At the same time, obtain the peak value Fmax corresponding to the current signal within the time period T, and calculate the ratio of the peak value to the effective value of the current to obtain the crest factor. And the specific calculation formula

[0041] Next, in step S103, analyze the historical data, classify the loads according to the crest factor to obtain the same-type load characteristics, and generate a peak matching interval according to the same-type load characteristics. Then, calculate the harmonic values corresponding to the peak matching interval.

[0042] First, obtain the historical data, and here the historical data is the historical records corresponding to all load types. Then, extract the load characteristics corresponding to the load types in the historical data. The specific load characteristic represents the crest factor. At the same time, perform the same-type classification processing on the load types to obtain the same-type load characteristics, and obtain the crest factor corresponding to the same-type load characteristics. Then, generate the same-type peak matching interval of the same-type load characteristics according to the crest factor. Here, the peak matching interval is a peak interval composed of the minimum value and the maximum value corresponding to the same-type load characteristics. For incandescent lamps, the current and voltage are almost in phase, and it is assumed that the measured crest factor is between 1.0 - 1.1 under different operating conditions; the motor is a typical inductive load, its current lags the voltage, and the measured crest factor is around 1.3 - 1.5; the current of the capacitive load leads the voltage, and its crest factor is in the range of 1.2 - 1.4.

[0043] Next, analyze the obtained same-type peak matching interval to determine whether there are multiple groups of the same crest factors in the same-type peak matching interval. Here, the multiple groups mean two or more identical crest factors. If there are multiple groups of the same crest factors, mark the corresponding same-type peak matching interval and record it as the interval to be analyzed. If there are no multiple groups of the same crest factors, no processing is performed.

[0044] Obtain all the intervals to be analyzed, and then calculate the amplitude and phase of the harmonics corresponding to the same crest factor in the intervals to be analyzed. The specific calculation method is as follows:

[0045] First, use a data acquisition device (such as an oscilloscope, a power analyzer, etc.) to sample the voltage or current signal at a certain sampling frequency to obtain a discrete signal sample sequence X(n), where n = 0, 1, 2, …, n - 1, and n is the number of sampling points. Perform FFT calculation on the collected discrete signal sequence. Here, the FFT represents the Fourier transform method, and convert the time-domain signal into a frequency-domain signal to obtain the spectrum X(k) of the corresponding signal (voltage or current signal), where k = 0, 1, 2, …, n - 1;

[0046] Calculate the amplitude A of the harmonic based on the obtained spectrum X(k) k and the phase The specific calculation formulas are as follows:

[0047] When k = 0, and |X(k)| represents the modulus of the complex number X(k), where arg represents taking the argument of the complex number. Specifically, the argument represents the angle between the vector and the positive direction of the real axis, and the calculated amplitude and phase are used as the distinguishing features of the corresponding load type. Suppose a discrete signal sequence is transformed by Fourier transform to obtain X(1) = 1 + i, then and the value calculated here is the modulus value. Further, the amplitude is calculated phase and the calculated amplitude and phase are used as the distinguishing features.

[0048] Next, in step S104, the corresponding real-time crest factor is calculated according to the obtained real-time features, and it is matched with the crest matching interval to obtain a matching result. At the same time, the specific load cause is determined according to the number of matching results, and load cause information is generated.

[0049] The real-time features are obtained, and the corresponding real-time crest factor is calculated according to the real-time features. The calculation method here is the same as that in step S102. At the same time, the obtained real-time crest factor is matched with the crest matching interval. The specific matching method is as follows: First, the real-time crest factor is matched with multiple crest matching intervals, and the corresponding conforming crest matching interval is screened and denoted as the marked interval. Then, the real-time crest factor is matched with the marked interval, and the load type that is the same as the real-time crest factor is screened and denoted as the load type to be analyzed;

[0050] In a specific embodiment, (for example, the crest matching interval for resistive load is [1.0, 1.1], the crest matching interval for inductive load is [1.3, 1.5], and the crest matching interval for capacitive load is [1.2, 1.4]) are compared one by one, and the conforming crest matching interval is screened out and marked as the marked interval. Suppose the calculated real-time crest factor is 1.4. After comparing with the above three intervals, the conforming one is the crest matching interval [1.2, 1.4] for capacitive load. Then this interval is the marked interval. Then, further matching is performed in the marked interval. Since the real-time crest factor 1.4 is within the interval of capacitive load, the load type can be determined to be capacitive load.

[0051] Then, the number of load types to be analyzed is judged. When there is only one load type to be analyzed, the corresponding load type to be analyzed is denoted as the standard load, and load cause information is generated at the same time. When there are multiple load types to be analyzed, the corresponding harmonic amplitude and phase are calculated according to the real-time features, and the calculated harmonic amplitude and phase are matched with the amplitude and phase of the load type to be analyzed to obtain the specific load type, and load cause information is generated at the same time.

[0052] In a specific embodiment, for example, there are resistive loads (the harmonic amplitude may be 0.5 units at certain frequencies and the phase is 30 degrees), inductive loads (the harmonic amplitude is 0.8 units at the corresponding frequencies and the phase is 60 degrees), and capacitive loads (the harmonic amplitude is 0.6 units at this frequency and the phase is 45 degrees). At this time, the harmonic amplitude calculated according to the real-time characteristics is 0.8 units and the phase is 60 degrees. Then, the calculated harmonic amplitude and phase are matched with the amplitudes and phases of these load types to be analyzed. After comparison, it is found that they match the harmonic amplitude and phase of the inductive load, thus determining that the specific load type is an inductive load, and at the same time generating load cause information, such as "By matching the characteristics of the harmonic amplitude of 0.8 units and the phase of 60 degrees, the load type is determined to be an inductive load".

[0053] Next, in step S105, dynamic feature extraction is performed according to the load cause information. At the same time, verification analysis is carried out based on the current rise rate in the obtained dynamic features, and a verification result is generated. The verification result includes a correct verification result and an incorrect verification result. Then, a secondary analysis is performed on the incorrect verification result according to the dynamic features to generate specific cause information.

[0054] Analyze the corresponding load type in the load cause information. It is judged by calculating the current rise rate corresponding to the load type, and the calculation method of the current rise rate is as follows:

[0055] Use a current transformer and data acquisition equipment (such as an oscilloscope, a power data collector, etc.) to sample the current during the load startup process. Let the collected current sequence be i(a), where a = 1, 2,..., b - 1, and a is the number of sampling points, and b represents a natural number, which is the sorting label for a from small to large. The value is taken according to the specific value of a (the number of sampling points). For example, if the number of sampling points is 10, then a = 1, 2,..., 10, and b is 11. The corresponding time sequence is t(a). Then, calculate the current difference between adjacent two sampling points Δi(a) = i(a + 1) - i(a), and the time interval between adjacent two sampling points Δt(a) = t(a + 1) - t(a);

[0056] Substitute the calculated current difference and time interval into the formula Calculate the current rise rate r(a), and by analogy, calculate multiple consecutive current rise rate values. For example, take the average of the current rise rates from a = 0 to a = b - 1, that is

[0057] Next, compare and judge the calculated current rise rate with the load cause information. If the current rise rate is the same as the current rise rate corresponding to the load cause information, it means that the load is correctly identified, and a verification correct result is generated. On the contrary, if the current rise rate is different from the current rise rate corresponding to the load cause information, it means that the load is incorrectly identified, and a verification error result is generated. Then, match the specific load based on the current rise rate as the standard, and generate specific cause information at the same time. And here, screening is carried out based on the current rise rate as the standard.

[0058] Embodiment 2. As the second embodiment of the present invention, the difference from Embodiment 1 is that in step S105, the current rise rate is calculated by the continuous function approximation calculation method.

[0059] If the function expression i(t) of the current changing with time during the load startup process is known, the current rise rate r(t) is equal to the derivative of the current function i(t) with respect to time t, that is For example, for a simple exponential current growth model i(t) = I0(1 - e -kt ), where I0 is the steady-state current and k is the time constant, its current rise rate r(t) = kI0e -kt .

[0060] Embodiment 3. As the third embodiment of the present invention, the key lies in combining the implementation processes of Embodiment 1 and Embodiment 2.

[0061] Some of the data in the above formula are numerically calculated after removing the dimension, and the content not described in detail in this specification belongs to the prior art well-known to those skilled in the art.

[0062] The above embodiments are only used to illustrate the technical method of the present invention and not to limit it. Although the present invention has been described in detail with reference to the preferred embodiments, those of ordinary skill in the art should understand that the technical method of the present invention can be modified or equivalently replaced without departing from the spirit and scope of the technical method of the present invention.

Claims

1. An intelligent power load identification method, characterized in that: The method specifically comprises the following steps: S101: collecting power data of the power system through a variety of sensors, and performing preprocessing operations on the power data to obtain preprocessed data; S102: extracting corresponding features from the obtained pre-processed data, calculating the corresponding effective voltage and effective current based on the time domain features, and calculating the peak factor according to the effective current value; S103: Analyze historical data, classify loads according to peak factors to obtain characteristics of loads of the same type, generate peak matching intervals according to the characteristics of loads of the same type, and then calculate harmonic values ​​corresponding to the peak matching intervals; S104: Calculate the corresponding real-time peak factor according to the acquired real-time features, match it with the peak matching interval to obtain a matching result, and determine the specific load cause according to the number of matching results, and generate load cause information; S105: Extract dynamic features based on load cause information, and perform verification analysis based on the current rise rate in the obtained dynamic features, and generate verification results, where the verification results include correct verification results and incorrect verification results, and then perform secondary analysis on the incorrect verification results based on the dynamic features to generate specific cause information.

2. The intelligent power load identification method according to claim 1, characterized in that: The specific method of performing preprocessing operation on the power data in step S101 to obtain preprocessed data is: The power data including voltage and current are acquired, and the voltage and current are their corresponding time series data, and then the acquired power data is filtered to obtain pre-processed data.

3. The intelligent power load identification method according to claim 1, characterized in that: The specific method of calculating the peak factor according to the effective current value in step S102 is: Obtain preprocessed data, define a time period T, and obtain the corresponding unit time voltage value and unit time current value within the time period T. Then, obtain the corresponding effective value by calculating the square root of the average value of the square of the current or voltage within the time period t. At the same time, obtain the peak value Fmax corresponding to the current signal within the time period T, and calculate the ratio of the peak value to the current effective value to obtain the peak factor.

4. The intelligent power load identification method according to claim 1, characterized in that: The specific method of generating the peak matching interval according to the same type of load characteristics in step S103 is: The historical data is acquired, and then the load characteristics corresponding to the load types in the historical data are extracted. At the same time, the load types are classified into the same type to obtain the same type of load characteristics, and the peak factors corresponding to the same type of load characteristics are obtained. Then, the same type of peak matching intervals of the same type of load characteristics are generated according to the peak factors.

5. The intelligent power load identification method according to claim 1, characterized in that: The specific method of calculating the harmonic value corresponding to the peak matching interval in step S103 is: Determine whether there are multiple groups of the same peak factors in the same type of peak matching intervals. If there are multiple groups of the same peak factors, mark the corresponding same type of peak matching intervals and record them as the intervals to be analyzed. If there are no multiple groups of the same peak factors, no processing is performed. All the intervals to be analyzed are obtained, and then the amplitudes and phases of the harmonics of the corresponding load types with the same peak factor in the intervals to be analyzed are calculated, and the calculated amplitudes and phases are used as distinguishing features of the corresponding load types.

6. The intelligent power load identification method according to claim 1, characterized in that: The specific method of obtaining the matching result in step S104 is: Acquire real-time features, calculate corresponding real-time peak factors according to the real-time features, and match the acquired real-time peak factors with the peak matching interval; The real-time peak factor is matched with multiple peak matching intervals, and the corresponding peak matching intervals are screened and recorded as marked intervals. Then, the real-time peak factor is matched with the marked intervals, and the load type with the same real-time peak factor is screened and recorded as the load type to be analyzed.

7. The intelligent power load identification method according to claim 1, characterized in that: The specific method of determining the specific load reason and generating the load reason information according to the number of matching results in step S104 is: When there is only one load type to be analyzed, the corresponding load type to be analyzed is recorded as the standard load, and load reason information is generated at the same time; When there are multiple load types to be analyzed, the corresponding harmonic amplitude and phase are calculated according to the real-time characteristics, and the calculated harmonic amplitude and phase are matched with the amplitude and phase of the load type to be analyzed to obtain the specific load type, and the load cause information is generated at the same time.

8. The intelligent power load identification method according to claim 1, characterized in that: The specific method of generating the verification result in step S105 is: The load type corresponding to the load reason information is analyzed, and the current rise rate corresponding to the load type is calculated, and the calculation method is: Use current transformers and data acquisition equipment to sample the current during the load startup process. Suppose the collected current sequence is i(a), where a=1, 2, ..., b-1, and a is the number of sampling points. The corresponding time series is t(a). Then calculate the current difference between two adjacent sampling points Δi(a)=i(a+1)-i(a), and the time interval between two adjacent sampling points Δt(a)=t(a+1)-t(a). Substitute the calculated current difference and time interval into the formula The current rise rate r(a) is calculated; Then the calculated current rise rate is compared with the load cause information. If the current rise rate is the same as the current rise rate corresponding to the load cause information, it means that the load identification is correct, and a correct verification result is generated. Otherwise, if the current rise rate is different from the current rise rate corresponding to the load cause information, it means that the load identification is wrong, and a verification error result is generated. Then, the current rise rate is used as the standard to match the specific load and generate specific cause information.

Images