阅读说明：本技术 一种基于v-i轨迹的非侵入式负荷识别方法 (Non-invasive load identification method based on V-I track ) 是由 金皓 谢岳 宋名扬 于 2020-05-14 设计创作，主要内容包括：本发明公开了一种基于V-I轨迹的非侵入式负荷识别方法,所述方法包括：对采集到的入户的电压电流数据标准化处理；通过监测入户的有功功率变化来判断有无负荷投切事件以及负荷运行是否进入稳态；计算负荷的Z值；对稳态电压电流数据进行标幺化并构建V-I轨迹；对确定的计算单元求取平均轨迹作为该单元特征轨迹并获取各计算单元的轨迹误差累积和以及相应的特征值；将待测的负荷特征序列与预设的负荷特征库结合特征权重进行相似度计算得出识别结果。与现有技术相比,本发明能够提高提取的负荷特征信息的可靠性、能够区分功率相近的负荷,进而提高负荷识别的准确度。(The invention discloses a non-invasive load identification method based on a V-I track, which comprises the following steps: standardizing the collected voltage and current data of the house; judging whether a load switching event exists or not and whether the load operation enters a stable state or not by monitoring the active power change of the house; calculating the Z value of the load; performing per unit on the steady-state voltage and current data and constructing a V-I track; calculating an average track of the determined calculation units as the characteristic track of the unit, and acquiring track error cumulative sum and corresponding characteristic values of each calculation unit; and performing similarity calculation on the load characteristic sequence to be detected and a preset load characteristic library in combination with the characteristic weight to obtain an identification result. Compared with the prior art, the method and the device can improve the reliability of the extracted load characteristic information, can distinguish loads with similar power, and further improve the accuracy of load identification.)

1. A non-invasive load identification method based on a V-I track comprises the following steps:

s1, collecting the voltage and current data of the user terminal, and preprocessing the obtained voltage and current data;

s2, calculating active power according to the preprocessed voltage and current data, and monitoring and judging the switching of the load through the active power (if no switching event occurs, continuously executing S2);

and S3, detecting whether the load operation enters a steady state. After the load is detected to enter a steady-state stage, the steady-state voltage and current data of the load are stored, and a Z value is calculated;

s4, performing per unit on the steady-state voltage and current data obtained in S3, and then constructing a per unit V-I track by taking the voltage as an abscissa and the current as an ordinate;

and S5, taking the N tracks of the V-I track obtained in the S4 as a calculation unit, and extracting the average track of each calculation unit as the characteristic track of the unit. Taking each V-I track in the calculating unit and the calculating error of the characteristic track corresponding to the calculating unit as the V-I track error accumulation sum of the calculating unit, and extracting corresponding V-I track characteristic values from the obtained characteristic tracks;

and S6, calculating a weight value corresponding to each load characteristic according to a preset load characteristic library, matching the preset load characteristic library with the acquired V-I track characteristics by combining the calculated characteristic weight values, and obtaining a recognition result by a maximum membership principle.

2. The method of claim 1, wherein the step S1 is implemented by high-frequency sampling of voltage and current data at the end of the service and normalizing the voltage and current data, and the calculation method is as follows:

in the formula: v (k), i (k) 1,2,3 … are raw voltage and current data collected, v (k) is a set of data obtained by a computer, and v (k) is a set of data obtained by a computer_rmsIs the effective value of the voltage, v_sta(k)、i_sta(k) Normalized voltage current data.

3. The non-intrusive load identification method based on the V-I track as claimed in claim 1, wherein the method for judging the load switching event in S2 is as follows:

(1) calculating the active power difference value of the load in the adjacent period at a certain moment;

(2) if the difference is larger than the set threshold value, the switching event is detected to occur.

4. The non-intrusive load identification method based on V-I track as claimed in claim 1, wherein the method for judging that the load enters the steady state in S3 is as follows:

and when a load switching event is detected at a certain moment, starting to judge the load steady state, and if N continuous periods meet the condition that the active power change of the adjacent period is smaller than a set threshold value, indicating that the load operation enters the steady state.

5. The non-invasive load identification method based on the V-I track according to claim 1, wherein the Z value calculation method in S3 is as follows:

in the formula: i.e. i_rms、v_rmsIs a current i_mEffective value of (d) and voltage v_mAnd (4) effective value.

6. The method for nonintrusive load identification based on V-I trajectory according to claim 1, wherein the method in S5 is:

(1) dividing the track into an upper part and a lower part through a connecting line of a voltage maximum value point and a voltage minimum value point of the track obtained in the step S4, and then respectively fitting the upper part and the lower part of the track;

(2) storing the N tracks as a computing unit, and calculating the average track of the N tracks as the characteristic track of the corresponding computing unit;

(3) in a certain calculation unit, the accumulated error is calculated by using the characteristic track of the unit and the stored track as the accumulated sum of the errors of the corresponding calculation unit.

7. The non-invasive load identification method based on V-I track according to claim 1, wherein the extracting corresponding characteristic values of the V-I track in S6 includes:

(1) the value of Z; (2) the area of the closed track; (3) the trajectory symmetry; (4) the slope of the middle part of the track; (5) area of the middle part of the track; (6) the number of self-cross points of the track; the extraction steps are as follows:

(a) acquiring a Z value;

(b) constructing a per-unit V-I characteristic track;

(c) acquiring the closed area of the per-unit V-I characteristic track;

(d) acquiring the symmetry of the per-unit V-I characteristic track;

(e) defining a closed part surrounded by the salient points of the track obtained according to the change of the tracking track as a middle part of the V-I characteristic track;

(f) acquiring the slope of the middle part of the per-unit V-I characteristic track;

(g) acquiring the area of the middle part of the per-unit V-I characteristic track;

(h) and obtaining the number of selfing points of the per-unit V-I characteristic track.

8. The method of claim 1 wherein the load recognition decision is given by the following equation:

in the formula: h is_ijThe characteristic j of the load i, f is the characteristic sequence of the load to be identified, n is the total number of the characteristics, the type of the load to be identified can be determined by the maximum similarity coefficient when the SIM card is used_iAnd when the maximum value is reached, the identification result is the load i.

Technical Field

The invention relates to a non-invasive load identification method, in particular to a non-invasive load identification method based on a V-I track.

Background

Load identification methods are mainly divided into two major categories, invasive load identification and non-invasive load identification. The intrusive load identification method needs to install a monitoring device on each electric device to record the operation condition of the device, has more accurate load identification result, but has high cost and is not easy to implement large-scale deployment; the non-invasive load identification method can monitor the load working condition in the house through the load identification algorithm by only installing one monitoring device at the home terminal, has low cost and strong practicability, and becomes the development trend of the future load monitoring direction.

The non-invasive load identification method is generally divided into five steps of data acquisition, data processing, event detection, feature extraction and load identification. The load feature extraction is the basis of load identification, and the feature values of various loads are extracted under different running states, so that the calculation amount of load identification can be reduced, and the accuracy of load identification can be improved. The existing load characteristics can be divided into two categories, namely transient characteristics and steady-state characteristics, wherein the steady-state characteristics mainly comprise active power, reactive power, a V-I track, fundamental components and harmonic components of voltage and current, and the transient characteristics mainly comprise instantaneous current, instantaneous power, voltage noise and the like. Transient characteristics are easily interfered by noise and are not easy to extract, and steady-state characteristics are relatively stable to extract and have strong anti-interference performance. The load characteristics are extracted according to the relation between the voltage and the current when the load operates by using the V-I track characteristics, and the stable V-I track characteristics reflect various electrical characteristics of the load.

In the existing research, characteristics such as track closed area, track symmetry, track self-intersection number and the like are extracted for load identification aiming at a load V-I track, and a better identification rate is obtained. However, the load with similar power and large V-I track fluctuation in operation cannot be accurately identified, and the invention provides a method for obtaining a stable V-I characteristic track by analyzing the change of the V-I track of the load in the steady-state operation process and identifying the load by a characteristic extraction technology.

Disclosure of Invention

The invention provides a non-invasive load identification method based on a V-I track, which adopts the following technical scheme:

a non-invasive load identification method based on a V-I track comprises the following steps:

s1, collecting the voltage and current data of the user terminal, and preprocessing the obtained voltage and current data;

s2, calculating active power according to the preprocessed voltage and current data, and monitoring and judging the switching of the load through the active power (if no switching event occurs, continuously executing S2);

and S3, detecting whether the load operation enters a steady state. After the load is detected to enter a steady-state stage, the steady-state voltage and current data of the load are stored, and a Z value is calculated;

s4, performing per unit on the steady-state voltage and current data obtained in S3, and then constructing a per unit V-I track by taking the voltage as an abscissa and the current as an ordinate;

and S5, taking the N tracks of the V-I track obtained in the S4 as a calculation unit, and extracting the average track of each calculation unit as the characteristic track of the unit. Taking each V-I track in the calculating unit and the calculating error of the characteristic track corresponding to the calculating unit as the V-I track error accumulated sum of the calculating unit, and extracting corresponding V-I track characteristic values from the obtained characteristic tracks;

and S6, calculating a weight value corresponding to each load characteristic according to a preset load characteristic library, matching the preset load characteristic library with the acquired V-I track characteristics by combining the calculated characteristic weight values, and obtaining a recognition result by a maximum membership principle.

Further, the voltage and current data in S1 are high-frequency sampled voltage data and high-frequency sampled current data;

further, the method for determining the load switching event in S2 includes:

(1) calculating the active power change value of the load in the adjacent period at a certain moment;

(2) if the value is larger than the set threshold value, the switching event is detected to occur.

Further, the method for determining that the load enters the steady state in S3 includes: and when a load switching event is detected at a certain moment, starting to judge the load steady state, and if N continuous periods meet the condition that the active power change of the adjacent period is smaller than a set threshold value, indicating that the load operation enters the steady state. The proposed Z value calculation method is:in the formula: i.e. i_rms、v_rmsIs a current i_mEffective value of (d) and voltage v_mAnd (4) effective value.

Further, the method in S5 above is:

(1) dividing the track into an upper part and a lower part through a connecting line of a voltage maximum value point and a voltage minimum value point of the track obtained in the step S4, and then respectively fitting the upper part and the lower part of the track;

(2) the load is fluctuated in the V-I track in the running process, so that the N tracks are stored as a computing unit, and the average track of the N tracks is obtained to be used as the characteristic track of the corresponding computing unit;

(3) in a certain calculation unit, the accumulated error is calculated by using the characteristic track of the unit and the stored track as the accumulated sum of the errors of the corresponding calculation unit.

Further, the extracting of the corresponding V-I trajectory feature value in S6 includes (1) a trajectory closed area; (2) the trajectory symmetry; (3) the slope of the middle part of the track; (4) area of the middle part of the track; (5) number of self-cross points of the track.

The extraction steps are as follows:

(a) acquiring the closed area of the per-unit V-I characteristic track;

(b) acquiring the symmetry of the per-unit V-I characteristic track;

(c) defining a closed part surrounded by the salient points of the track obtained according to the change of the tracking track as a middle part of the V-I characteristic track;

(d) acquiring the slope of the middle part of the per-unit V-I characteristic track;

(e) acquiring the area of the middle part of the per-unit V-I characteristic track;

(f) acquiring the number of selfing points of the per-unit V-I characteristic track;

the preset load characteristic library comprises the following information: (1) the value of Z; (2) the area of the closed track; (3) the trajectory symmetry; (4) slope of the middle part of the track; (5) area of the middle part of the track; (6) the number of self-cross points of the track; matching the preset load characteristic library with the acquired V-I track characteristics by combining the weights corresponding to the load characteristics obtained by solving, and obtaining a recognition result by a maximum membership principle.

Drawings

FIG. 1 is a flow chart of a method for non-intrusive load identification based on V-I trajectory;

FIG. 2 is a schematic view of a V-I trajectory according to an embodiment of the present invention;

FIG. 3 is a schematic diagram of V-I trajectory segmentation according to an embodiment of the present invention;

FIG. 4 is a schematic view of a V-I trajectory in operation according to an embodiment of the present invention;

FIG. 5 is a schematic diagram of a middle portion of a V-I trace according to an embodiment of the present invention;

Detailed Description

The invention is explained by combining the drawings and the embodiment, and the concrete implementation steps are as follows:

the invention provides a non-invasive load identification method based on a V-I track, which comprises the following implementation steps:

s1: the collected raw voltage and current data are normalized, and the normalization formula in the embodiment of the invention is as follows:

in the formula: v (k), i (k) 1,2,3 … are raw voltage and current data collected, v (k) is a set of data obtained by a computer, and v (k) is a set of data obtained by a computer_rmsIs the effective value of the voltage, v_sta(k)、i_sta(k) Normalized voltage current data.

S2: the current data collected were:

wherein m is the total number of loads, a_nCoefficient (a) for current load opening_nIs 0 or 1), i_n(k) The term (n ═ 1, 2.. times.m) represents the current at which each load operates, and the term (e) (k) represents noise.

Let the active power at time t be P_tIf the active power variation at time t is Δ P ═ P_t-P_t-1Therefore, whether the electric load is switched or not can be judged according to the formula delta P & gt ξ, and ξ is taken1/20 for the last instant of active power.

S3: when Δ P is satisfied at a certain time_tWhen the voltage is more than ξ, the load steady state judgment is started, if N continuous periods can meet the condition that delta P is less than ξ, the load operation enters the steady state, and the steady state voltage v is stored_mSteady state current i_mAnd calculating the Z value:in the formula: i.e. i_rms、v_rmsAre respectively current i_mEffective value of (d) and voltage v_mAnd (4) effective value.

S4: the voltage and current data (v) obtained at S3_m(k)、i_m(k) Per unit according to formula (4) and formula (5):

in the formula: v. of_g(k) Is the voltage data per unit i_g(k) Is the per unit current data, v_maxIs the maximum value of voltage, i_maxIs the maximum value of the current, v_m(k)、i_m(k) Voltage and current data is obtained for S3. The load V-I locus is obtained by taking the per-unit voltage as the abscissa and the per-unit current as the ordinate as shown in FIG. 2.

S5: by obtaining the voltage maximum point v of the track_gmaxAnd voltage minimum point v_gminDividing the formed VI locus into an upper part A, B and a lower part A, B shown in figure 3, and obtaining a fitting form i of the current of the locus A, B part relative to the voltage by least square fitting_A(u)、i_B(u) and u is a voltage.

In order to ensure that a stable load characteristic value is obtained, the V-I track obtained by S4 takes N tracks as a calculation unit, and an average track of each calculation unit is extracted

And as the characteristic track of the unit, calculating errors of each V-I track in each calculating unit and the characteristic track corresponding to the calculating unit according to the formula (6) to obtain the accumulated sum of the V-I track errors of the calculating unit.

In the formula: i.e. i_An(u)、i_Bn(u) (N is 1,2, … N) is a formula of the locus in the calculation unit obtained by fitting,and N is the number of the V-I tracks of the calculation unit, and E is the accumulated sum of errors.

Analyzing the V-I track, and calculating a track characteristic value:

(1) calculating to obtain the enclosed image area S enclosed by the V-I track, and calculating as follows:

(2) defining the similarity of the V-I track and a graph obtained after the V-I track is rotated by 180 degrees as a symmetry characteristic, and calculating the symmetry of the V-I track according to the following formula:

in the formula: HD is the symmetry index of the V-I track.

(3) According to the formula

(in the formulaAs a characteristic locus, u_nIs a voltage value, λ is a set threshold value

) Tracking the slope change of the A track to obtain the voltage value u of the corresponding two points₁、u₂If (as in fig. 5) the slope of the middle of the V-I trace is calculated, then the slope of the middle of trace A, B is calculated as follows:

in the formula: k_A、K_BSlope of the middle portion of the trace A, B, u₁、u₂Is the voltage value on the trace.

(4) The area of the middle portion of the V-I trace is calculated as follows:

in the formula: s_midIs the area of the middle portion of the V-I trace, u₁、u₂Is the voltage value.

(5) Judging whether the two parts of the V-I track A, B have an intersection or not:

in the formula:

for the fitting expression of the trajectory, the number of solutions thereof is taken as the number of self-intersection points SC of the trajectory by solving equation (13).

S6: in an embodiment of the present invention, m loads are provided, and a preset feature library is defined as follows:

the load characteristic information stored in the characteristic library includes: z value of V-I track, area S enclosed by V-I track, symmetry index HD of V-I track, and slope K of middle part of V-I track_AAnd K_BMiddle part area S of V-I track_midThe number of selfing points SC of the V-I track, and the error accumulation sum E of the V-I track.

Calculating the weight omega of each feature according to the established load feature library, wherein the calculation process is as follows:

(1) normalization of feature values in a feature library D

In the formula: h is_ijIs the characteristic value j, h of the load i in D_iminIs the minimum value of the characteristic sequence of the load i, h_imaxIs the maximum value of the load i signature sequence.

(2) Calculating the specific gravity of the features:

in the formula: n is the number of feature classes.

(3) And (3) calculating the information entropy of each characteristic value:

in the formula: k ═ ln (m)^-1And m is the total number of loads.

(4) Calculating the feature weight of each feature:

similarity calculation is carried out on the obtained V-I track characteristics of the load to be identified and a load characteristic library, and the characteristic sequence of the load to be identified is defined as f ═ Z S HD K_AK_BSmid SC E]^TD is defined as negative to be recognizedThe similarity matrix of the load and the feature library is as follows:

in the formula: d_ijThe characteristic j of the load i in the load characteristic library and the characteristic f of the load to be identified_jThe calculation formula of the similarity is as follows:

in the formula: h is_ijIs the characteristic j, f of the load i_jThe characteristic j and n of the load to be identified are characteristic numbers.

The similarity coefficient between the load to be measured and the load characteristic library is

The load type to be identified can be determined by the maximum similarity coefficient when the SIM card is used_iAnd when the maximum value is reached, the identification result is the load i.