阅读说明：本技术 一种基于电流空间轨迹的无监督非侵入式电视机辨识方法 (Unsupervised non-invasive television identification method based on current space trajectory ) 是由 栾文鹏 刘子帅 刘博� 余贻鑫 于 2021-08-26 设计创作，主要内容包括：本发明公开了一种基于电流空间轨迹的无监督非侵入式电视机辨识方法,本发明中按照电力消耗模式将电视机分为功率波动型和功率稳定型两种,按照电视机启动暂态过程中的功率变化波形将电视机分为阶跃启动型和长启动暂态型两种；并包括以下步骤：步骤一、基于功率波动区段的划分及波动特性计算结果进行功率波动型电视机预判；步骤二、采用动态变点检测的方法进行负荷事件检测；步骤三、电视机负荷印记提取和状态辨识,基于功率波动区段划分与负荷事件检测结果,利用所设计的组合功率特征和电流空间轨迹,并结合关于电视机运行特性的先验知识,完成电视机状态辨识；本发明具有较高的实用性和鲁棒性。(The invention discloses an unsupervised non-invasive television identification method based on current space trajectory, which divides a television into a power fluctuation type and a power stabilization type according to a power consumption mode, and divides the television into a step start type and a long start transient type according to a power change waveform in a transient start process of the television; and comprises the following steps: firstly, pre-judging a power fluctuation type television based on the division of a power fluctuation section and a fluctuation characteristic calculation result; step two, adopting a dynamic variable point detection method to detect the load event; thirdly, extracting a television load mark and identifying a state of the television, namely identifying the state of the television by utilizing the designed combined power characteristic and current space trajectory and combining the prior knowledge about the running characteristic of the television based on the power fluctuation section division and the load event detection result; the invention has higher practicability and robustness.)

1. An unsupervised non-invasive TV identification method based on current space track is characterized in that,

dividing the television into a power fluctuation type television and a power stabilization type television according to a power consumption mode, defining the television with irregular fluctuation of running power and fluctuation average amplitude larger than 10% of the running power per se or larger than 10W as the power fluctuation type television, and otherwise defining the television as the power stabilization type television;

dividing the television into a step start type television and a long start transient type television according to a power change waveform in the transient starting process of the television, defining the television with a power waveform in a step rising state in the starting process as the step start type television, and defining the television with a section of low power section before reaching a normal operation power level in the starting process as the long start transient type television;

and comprises the following steps:

firstly, according to the division of a power fluctuation section, carrying out power fluctuation type television prejudgment by utilizing the power fluctuation characteristic;

step two, adopting a dynamic variable point detection method to detect the load event;

thirdly, extracting a television load mark and identifying the state of the television, namely, determining the starting and stopping time of the television in a load total power time sequence based on the power fluctuation section division in the first step and the load event detection result in the second step and by combining the prior knowledge about the running characteristic of the television; for the television with stable power, the television state identification can be completed by searching the television load event; for the power fluctuation type television, sequentially carrying out a load event searching step and television state identification based on a current space track; and finally, judging whether the television runs in the power fluctuation section.

2. The unsupervised non-intrusive television identification method based on current space trajectory as claimed in claim 1, wherein step one comprises the steps of:

step 1-1) carrying out piecewise linearization on an input total load active power time sequence shown in a formula (1);

P＝[P₁,P₂,P₃,...P_i,...P_n] (1)

in the formula (1), P_iThe total load active power value at the ith second is represented, and n represents the total time length of the data; obtaining a segmentation point power sequence shown in a formula (2) through segmentation linearization;

in the formula (2), the reaction mixture is,the active power value at the ith segmentation point is represented, and m represents the number of all segmentation points in the data;

step 1-2) Power fluctuation segment division

Firstly, segment point power sequence P_sCarrying out difference and taking an absolute value to obtain a segment point power difference absolute value sequence delta P_sAs shown in formula (3);

in the formula (3), the reaction mixture is,then at Δ P_sThe interval between the starting points in the sequence is less than a certain threshold T_th、|ΔP_s| greater than a certain threshold Δ P_{s_th}The elements of (a) are drawn into the same fluctuation section;

step 1-3) Power Difference outlier detection

Δ P contained in each fluctuation section_sThe sequence carries out outlier detection and removes outliers to eliminate the interference of load events, and the power difference set of the residual sectional points in each fluctuation section is delta P_{s_re}；

Step 1-4) power fluctuation type television pre-judgment

Screening is carried out for each fluctuation section if the delta P of the fluctuation section_{s_re}If the number of elements in the set exceeds a certain threshold value C, starting fluctuation characteristic calculation and television existence prejudgment of the fluctuation section; if the formula (4) is established, judging that the fluctuation section is suspected to have the operation of the power fluctuation type television;

in the formula (4), μ (·), σ (·) and Median (·) respectively represent the operations of calculating the mean, standard deviation and Median; delta Q_{s_re}Is equal to Δ P_{s_re}A corresponding reactive delta; k₁、K₂And K₃Is a predefined threshold parameter.

3. The unsupervised non-intrusive television identification method based on current space trajectory as claimed in claim 2, wherein the specific content of step two is as follows:

performing point-by-point scanning on the total load active power time sequence, and when the absolute value of the power difference delta P between a scanning point and the next point is larger than a predefined threshold delta P_thThen the scanning point is regarded as a change point；

Dividing the input total load active power time sequence into a plurality of sections with a transient process section and a steady process section alternating in a variable point detection mode; wherein, there is no power change point in the steady state process section and the time length of the section is larger than the predefined threshold value delta t_thEach resulting load event record includes { t }_start,t_end，ΔP_evt,ΔQ_evt}; wherein, t_startIs the starting moment of the transient process of the load event, t_endIs the end time, Δ P, of the transient process of the load event_evt、ΔQ_evtRespectively is the active power variation and the reactive power variation of the transient process of the load event.

4. The unsupervised non-intrusive television identification method based on current space trajectory as claimed in claim 2, wherein the specific content of step three is as follows:

3-1) the step of searching the television load event is to search the start and stop load event corresponding to the television start-stop process in the load event detection result based on the television combined power characteristic; the method comprises the following steps:

step 3-1-1) defining the searching range of the load event

For the power fluctuation type television, the load event search range is expanded and searched in T time ranges before and after the starting time and the ending time of the fluctuation section, and the load event is associated with the fluctuation section; for a power-stable television, the load event search range is all detected load events;

step 3-1-2) searching for TV set loading event

For the power fluctuation type television, the load starting event which is closest to the starting point of the fluctuation section and satisfies the formula (5) is taken as a suspected television load starting event; for a power stable television, all the loading events satisfying equation (5) are taken as television loading events;

for a long-start transient television, two adjacent load starting events are detected in the starting process, and the two load events are searched according to a left formula in a formula (5); in the formula (5), t_start1And t_end1Respectively the starting time and the ending time of the transient process of the first load event in the two adjacent load events; t is t_start2And t_end2Respectively the starting time and the ending time of the transient process of the second load event in the two adjacent load events; (P)_{TV_min},P_{TV_max}) Operating the active power range for the television; q represents reactive power; t is t_{TV_ON}The obtained time length threshold value of the television starting process is researched;is the set minimum threshold value of the power factor angle tangent value when the television operates;

for the step start type television, only one load event is detected in the starting process, and the load event is searched according to the right formula in the formula (5);

step 3-1-3) screening load stop events corresponding to the television closing to match with the television load starting events

The judgment rule for screening the load stopping events corresponding to the closing of the television is shown as a formula (6), the matching with the television load starting events follows the nearest rule, namely, the load stopping event closest to the load starting event is selected from the load events meeting the judgment rule shown as the formula (6);

in the formula (6), Δ P_{TV_ON}For power variations during the start-up of a television set, for a long-start transient television set, Δ P_{TV_ON}The sum of the active power changes of two loading events in the starting process of the television set of the type found in the step 3-1-2); for step start-up television, Δ P_{TV_ON}For the type of electricity found in step 3-1-2)The active power change of a load starting event in the starting process of the video player is observed; delta P_{TV_th}Is a matching threshold for start and stop load events;

3-2) the television state identification based on the current space trajectory comprises:

step 3-2-1) obtaining current data of a power fluctuation section between the start and stop load events of the television through differential operation

Before the difference operation, uniformly converting the fundamental wave current and the third harmonic current of the total load in the whole time period corresponding to the time sequence of the active power of the total load input in the step 3-1) under the voltage phase, as shown in a formula (7);

in the formula (7), the reaction mixture is,is the angle of the power factor of the fundamental wave,the fundamental wave current in the form of phasor before and after conversion and the third harmonic current are in the same way; carrying out phasor subtraction operation on the converted total load fundamental current and third harmonic current sequence in the fluctuation section and the current in the stable process section before the television load starting event to obtain a fundamental difference current effective value sequence and a third harmonic current effective value sequence of the fluctuation section, wherein k is the sequence length as shown in a formula (8);

step 3-2-2) analyzing the third harmonic-fundamental wave current space trajectory to realize television state identification, comprising the following steps of:

firstly, the DBSCAN method is adopted to measure the Delta I₃-ΔI₁Clustering sample points in the space; then, the polymerization is carried outEach cluster in the class result is subjected to television judgment, namely: judging from two angles of the parameter range and the shape of the sample point cluster obtained by clustering, wherein the sample point cluster generated by independently operating the television in the third harmonic-fundamental wave current space meets the following requirements: a) the effective value of the fundamental current is (0, I)_1max) Inner, I_1maxThe maximum value threshold value of the television fundamental current is obtained through investigation; b) the ratio of the cluster-like center third harmonic current effective value to the fundamental current effective value is larger than a set threshold value r;

the method for judging the shape of the sample point cluster with continuous belt-shaped characteristics comprises the following steps:

firstly, respectively carrying out z-score standardization treatment on fundamental wave current and third harmonic current of each sample point cluster obtained by clustering; then, judging whether the formula (9) is established or not;

in formula (9), Δ I after z-score normalization₁-ΔI₃In-plane, Point₁、Point₂Respectively sample points with the minimum and maximum horizontal coordinates in the sample point cluster; point (Point)₃Point₄Respectively sample points with the minimum and maximum vertical coordinates near the horizontal center of the sample point cluster; dist (·,) is a euclidean distance computation notation, w is a predefined shape coefficient; if the formula (9) is true, the fluctuation section is determined as a television operation section; otherwise, if other load events which are screened as television starting still exist near the beginning of the fluctuation section, returning to the step 3-2) according to the other load events, and then performing current space trajectory extraction and television state identification; if the formula (9) does not hold, it is determined that the television operation does not exist in the surge section.

Technical Field

The invention belongs to the field of non-invasive power load monitoring, and particularly relates to an unsupervised non-invasive television identification method by using a current space trajectory.

Background

In recent years, with the continuous development and maturity of smart grid technology, it has become an urgent need to realize grid intellectualization to improve the participation of users in the power market, enhance the response of users on demand side, and the like. The monitoring of the details of the electricity consumption of the residential power load can provide important technical support for achieving the purposes, meanwhile, the interaction depth of the power company and residential users about the electricity consumption data can be improved, and planning, operation and management of the power company and energy efficiency upgrading of the residential users and the like are facilitated. And Non-intrusive load monitoring (NILM) does not need to intrude into the user to install a sensor for each device, and can acquire power consumption information of each power consumption device switched in the total load only based on detailed analysis of load power consumption total data measured at an inlet of a power supply. By virtue of the advantages of low cost, strong expansibility, convenience in maintenance, easiness in acceptance by users and the like, the non-invasive power load monitoring becomes a mainstream technical means for realizing the power consumption detail monitoring of the power load.

The accuracy of the non-invasive load identification result directly influences the scientificity of subsequent decisions and the trust degree of the user on the result, so that the positivity of the response of the user on the demand side is related. From the angle of large-scale application of the NILM, the identification accuracy of electrical equipment with high household ownership is guaranteed, and the effectiveness of the overall decision of the power company can be guaranteed; for the user, the attention is often paid to the identification result of the electrical equipment which is frequently used in the family. Therefore, it is important to improve the identification accuracy of electrical appliances such as televisions, air conditioners, water heaters, etc., which are in the front of the home ownership and the user usage frequency. Although many researchers have been exploring the NILM field, most of the existing research has poor or even no non-invasive identification of tv sets, and the difficulties and challenges are mainly:

(1) compared with electric appliances such as a water heater, an air conditioner and the like, the power of the television is smaller, and more electric appliances are arranged in the same power range, so that the change of the power caused by the starting and stopping processes of the television is not obviously differentiated. In addition, the running process of the high-power electrical appliance also fluctuates, so that the power change of the television is easy to be confused, and the identification difficulty is increased.

(2) The internal circuit structure of the television is complex, unlike the heating equipment with main components such as resistors or the refrigeration equipment with main components such as motors, the television comprises a large number of electronic devices, and meanwhile, the working process of the television is variable, such as channel switching, volume adjustment, video picture continuous change and the like, so that the power characteristic of the television in the running process is continuously changed, and it is a great challenge to find stable power characteristic quantity to represent the working state of the television.

(3) The television has a plurality of technical genres and often has different power consumption modes, so in order to achieve a large-scale commercial goal, the identification needs to take into account a plurality of types of televisions as much as possible, that is, the generalization capability of the non-invasive television identification method is ensured, which further increases the identification difficulty.

Therefore, in order to solve the above problems, a common identification strategy using a small amount of power characteristics cannot meet the requirements, and it is necessary to design a multidimensional load mark besides power variation, and to provide an adaptive non-intrusive television identification method.

Disclosure of Invention

Aiming at the prior art, the invention provides an unsupervised non-invasive television identification method based on current space trajectory, wherein the television is divided into a power fluctuation type television and a power stabilization type television according to a power consumption mode, the television with irregular fluctuation of running power and fluctuation average amplitude larger than 10% of the running power per se or larger than 10W is defined as the power fluctuation type television, and otherwise, the television is defined as the power stabilization type television; the method comprises the steps of dividing the television into a step start type television and a long start transient type television according to a power change waveform in the transient starting process of the television, defining the television with a power waveform in a step rising state in the starting process as the step start type television, and defining the television with a section of low power section before the power reaches a normal operation power level in the starting process as the long start transient type television. The method comprises the following steps:

firstly, according to the division of a power fluctuation section, carrying out power fluctuation type television prejudgment by utilizing the power fluctuation characteristic;

step two, adopting a dynamic variable point detection method to detect the load event;

thirdly, extracting a television load mark and identifying the state of the television, namely, determining the starting and stopping time of the television in a load total power time sequence based on the power fluctuation section division in the first step and the load event detection result in the second step and by combining the prior knowledge about the running characteristic of the television; for the television with stable power, the television state identification can be completed by searching the television load event; for the power fluctuation type television, sequentially carrying out a load event searching step and television state identification based on a current space track; and finally, judging whether the television runs in the power fluctuation section.

Further, the specific content of each step is as follows:

the first step comprises the following steps:

step 1-1) carrying out piecewise linearization on an input total load active power time sequence shown in a formula (1);

P＝[P₁,P₂,P₃,...P_i,...P_n] (1)

in the formula (1), P_iThe total load active power value at the ith second is represented, and n represents the total time length of the data; obtaining a segmentation point power sequence shown in a formula (2) through segmentation linearization;

in the formula (2), the reaction mixture is,represents the active power value at the ith segmentation point, and m represents all the segmentation points in the dataThe number of (2); step 1-2) Power fluctuation segment division

Firstly, segment point power sequence P_sCarrying out difference and taking an absolute value to obtain a segment point power difference absolute value sequence delta P_sAs shown in formula (3);

in the formula (3), the reaction mixture is,then at Δ P_sThe interval between the starting points in the sequence is less than a certain threshold T_th、|ΔP_s| greater than a certain threshold Δ P_{s_th}The elements of (a) are drawn into the same fluctuation section;

step 1-3) Power Difference outlier detection

Δ P contained in each fluctuation section_sThe sequence carries out outlier detection and removes outliers to eliminate the interference of load events, and the power difference set of the residual sectional points in each fluctuation section is delta P_{s_re}；

Step 1-4) power fluctuation type television pre-judgment

Screening is carried out for each fluctuation section if the delta P of the fluctuation section_{s_re}If the number of elements in the set exceeds a certain threshold value C, starting fluctuation characteristic calculation and television existence prejudgment of the fluctuation section; if the formula (4) is established, judging that the fluctuation section is suspected to have the operation of the power fluctuation type television;

in the formula (4), μ (·), σ (·) and Median (·) respectively represent the operations of calculating the mean, standard deviation and Median; delta Q_{s_re}Is equal to Δ P_{s_re}A corresponding reactive delta; k₁、K₂And K₃Is a predefined threshold parameter.

The specific content of the second step is as follows:

performing point-by-point scanning on the total load active power time sequence, and when the absolute value of the power difference delta P between a scanning point and the next point is larger than a predefined threshold delta P_thThen the scanning point is regarded as a variable point;

dividing the input total load active power time sequence into a plurality of sections with a transient process section and a steady process section alternating in a variable point detection mode; wherein, there is no power change point in the steady state process section and the time length of the section is larger than the predefined threshold value delta t_thEach resulting load event record includes { t }_start,t_end，ΔP_evt,ΔQ_evt}; wherein, t_startIs the starting moment of the transient process of the load event, t_endIs the end time, Δ P, of the transient process of the load event_evt、ΔQ_evtRespectively is the active power variation and the reactive power variation of the transient process of the load event.

The third step comprises the following specific contents:

3-1) the step of searching the television load event is to search the start and stop load event corresponding to the television start-stop process in the load event detection result based on the television combined power characteristic; the method comprises the following steps:

step 3-1-1) defining the searching range of the load event

For the power fluctuation type television, the load event search range is expanded and searched in T time ranges before and after the starting time and the ending time of the fluctuation section, and the load event is associated with the fluctuation section; for a power-stable television, the load event search range is all detected load events;

step 3-1-2) searching for TV set loading event

For the power fluctuation type television, the load starting event which is closest to the starting point of the fluctuation section and satisfies the formula (5) is taken as a suspected television load starting event; for a power stable television, all the loading events satisfying equation (5) are taken as television loading events;

for a long-start transient television, two adjacent load starting events are detected in the starting process, and the two load events are searched according to a left formula in a formula (5); in the formula (5), t_start1And t_end1Respectively the starting time and the ending time of the transient process of the first load event in the two adjacent load events; t is t_start2And t_end2Respectively the starting time and the ending time of the transient process of the second load event in the two adjacent load events; (P)_{TV_min},P_{TV_max}) Operating the active power range for the television; q represents reactive power; t is t_{TV_ON}The obtained time length threshold value of the television starting process is researched;is the set minimum threshold value of the power factor angle tangent value when the television operates;

for the step start type television, only one load event is detected in the starting process, and the load event is searched according to the right formula in the formula (5);

step 3-1-3) screening load stop events corresponding to the television closing to match with the television load starting events

The judgment rule for screening the load stopping events corresponding to the closing of the television is shown as a formula (6), the matching with the television load starting events follows the nearest rule, namely, the load stopping event closest to the load starting event is selected from the load events meeting the judgment rule shown as the formula (6);

in the formula (6), Δ P_{TV_ON}For power variations during the start-up of a television set, for a long-start transient television set, Δ P_{TV_ON}The sum of the active power changes of two loading events in the starting process of the television set of the type found in the step 3-1-2); for step start-up television, Δ P_{TV_ON}For those found in step 3-1-2)The active power of the start load event changes in the starting process of the television set; delta P_{TV_th}Is a matching threshold for start and stop load events;

3-2) the television state identification based on the current space trajectory comprises:

step 3-2-1) obtaining current data of a power fluctuation section between the start and stop load events of the television through differential operation

Before the difference operation, uniformly converting the fundamental wave current and the third harmonic current of the total load in the whole time period corresponding to the time sequence of the active power of the total load input in the step 3-1) under the voltage phase, as shown in a formula (7);

in the formula (7), the reaction mixture is,is the angle of the power factor of the fundamental wave,the fundamental wave current in the form of phasor before and after conversion and the third harmonic current are in the same way; carrying out phasor subtraction operation on the converted total load fundamental current and third harmonic current sequence in the fluctuation section and the current in the stable process section before the television load starting event to obtain a fundamental difference current effective value sequence and a third harmonic current effective value sequence of the fluctuation section, wherein k is the sequence length as shown in a formula (8);

step 3-2-2) analyzing the third harmonic-fundamental wave current space trajectory to realize television state identification, comprising the following steps of:

firstly, the DBSCAN method is adopted to measure the Delta I₃-ΔI₁Clustering sample points in the space; then, each cluster in the clustering result is electrically conductedAnd (4) machine-viewing judgment, namely: judging from two angles of the parameter range and the shape of the sample point cluster obtained by clustering, wherein the sample point cluster generated by independently operating the television in the third harmonic-fundamental wave current space meets the following requirements: a) the effective value of the fundamental current is (0, I)_1max) Inner, I_1maxThe maximum value threshold value of the television fundamental current is obtained through investigation; b) the ratio of the cluster-like center third harmonic current effective value to the fundamental current effective value is larger than a set threshold value r;

the method for judging the shape of the sample point cluster with continuous belt-shaped characteristics comprises the following steps:

firstly, respectively carrying out z-score standardization treatment on fundamental wave current and third harmonic current of each sample point cluster obtained by clustering; then, judging whether the formula (9) is established or not;

in formula (9), Δ I after z-score normalization₁-ΔI₃In-plane, Point₁、Point₂Respectively sample points with the minimum and maximum horizontal coordinates in the sample point cluster; point (Point)₃ Point₄Respectively sample points with the minimum and maximum vertical coordinates near the horizontal center of the sample point cluster; dist (·,) is a euclidean distance computation notation, w is a predefined shape coefficient; if the formula (9) is true, the fluctuation section is determined as a television operation section; otherwise, if other load events which are screened as television starting still exist near the beginning of the fluctuation section, returning to the step 3-2) according to the other load events, and then performing current space trajectory extraction and television state identification; if the formula (9) does not hold, it is determined that the television operation does not exist in the surge section.

Compared with the prior art, the invention has the beneficial effects that:

aiming at the non-intrusive television identification of the television, the invention designs two novel load marks of a combined power characteristic and a harmonic current space track during starting, and establishes an unsupervised non-intrusive television identification frame and a specific implementation method thereof. The running state of the television can be identified from the total load data completely without supervision only based on some empirical prior knowledge, the running state of the television can be effectively distinguished from other electrical appliances in the same power range, non-intrusive identification of the television with different power consumption modes under various technical genres can be considered, and the method has high practicability and robustness.

Drawings

FIG. 1 is a system framework for unsupervised non-intrusive television identification based on current space trajectory in accordance with the present invention;

FIG. 2 is a flow chart of power fluctuation type TV pre-determination in the present invention;

FIG. 3 total load power data for a length of time;

FIG. 4 illustrates an exemplary detection of a total load active power load event for a time period;

FIG. 5 is a diagram illustrating a screen for TV startup load events;

FIG. 6 shows the screening results of the start and stop events of the TV set;

FIG. 7 third harmonic-fundamental current delta cluster analysis results;

fig. 8 shows a cluster of points corresponding to independent operation of a tv set.

Detailed Description

The invention will be further described with reference to the following figures and specific examples, which are not intended to limit the invention in any way.

Through experimental test and analysis of a plurality of televisions, the television is divided into two types according to the power consumption mode: in the invention, the television with the irregular fluctuation of the running power and the fluctuation average amplitude which is more than 10 percent of the running power per se or more than 10W is defined as a power fluctuation type television; another type of television set, which operates relatively smoothly, is referred to herein as a power-stabilized television set. The method comprises the steps of dividing the television into a step start type television and a long start transient type television according to a power change waveform in the transient starting process of the television, defining the television with a power waveform in a step rising state in the starting process as the step start type television, and defining the television with a section of low power section before the power reaches a normal operation power level in the starting process as the long start transient type television.

Through the combination of experimental analysis and mechanism analysis, the invention designs two novel load marks aiming at television identification: the power signature and harmonic current spatial trajectory are combined at start-up. The combined power characteristic during starting is characterized by combining the power change in the transient process of starting the television from three angles: 1) a "low power section" exhibited when a part of the television is turned on; 2) the power of the television after being started is kept in a fixed range; 3) the TV circuit has reactive characteristic caused by capacitance. The harmonic current space track means that although the current, the power and the like are randomly changed in the operation process of the power fluctuation type television, the correlation relationship between the fundamental current and the third harmonic current is relatively stable, and a continuous strip-shaped track is formed in the fundamental current-third harmonic current effective value space.

In order to take identification of multiple types of televisions as much as possible into consideration, the invention designs a system framework for realizing the unsupervised non-intrusive television identification method based on the current space track, and can realize accurate identification of power fluctuation type televisions and televisions with obvious transient processes in strange scenes only by depending on some empirical prior knowledge. As shown in fig. 1, the framework includes 4 functional modules: the system comprises a total load data acquisition module, a power fluctuation type television pre-judging module, a load event detection module and a television load mark extraction and state identification module.

The total load data acquisition module is responsible for providing target original data to be subjected to non-invasive identification, and is total load data with a certain time length (generally one day), including voltage and current waveform digital signals, active power data and reactive power data;

the power fluctuation type television pre-judging module is responsible for judging whether a load mark extraction step of a subsequent harmonic current space track needs to be started and executed or not and determining an approximate running interval of the power fluctuation type television. Because the power of the power fluctuation type television during operation fluctuates in a large value and irregularly, the active power of the input section can be judged, and the section with certain fluctuation range is suspected to be the television operation section.

The load event detection module is responsible for dividing a section of total load data into a transient section (namely, a load event occurs) and a steady section (namely, a steady section) which alternately appear and are used for starting and stopping an electric appliance or converting the state of the electric appliance, and the steady section is used for generating a no-load event. The load event detection object is total load active power, and the detection is realized according to the change of the total load active power within a certain range when the state of the electric appliance is converted.

The television load imprint extraction and state identification module extracts combined power characteristics and harmonic current space track load imprints during starting on the basis of the load event detection result on the total load data, and then realizes identification of the television state by combining with some priori knowledge about television operation, so that some operation records about the television, including the start-stop time of the television and the average power level during operation, are output. The identification of the power-stable television with obvious transient process depends on the combination of the power characteristics during starting, and the identification of the power fluctuation television depends on the combination of the power characteristics and the harmonic current space trajectory during starting.

The unsupervised non-invasive television identification method based on the current space trajectory, which is disclosed by the invention, provides a specific implementation method for the main modules in the frame, and mainly comprises the following steps:

firstly, according to the division of a power fluctuation section, carrying out power fluctuation type television prejudgment by utilizing the power fluctuation characteristic;

step two, adopting a dynamic variable point detection method to detect the load event;

thirdly, extracting a television load mark and identifying the state of the television, namely, determining the starting and stopping time of the television in a load total power time sequence based on the power fluctuation section division in the first step and the load event detection result in the second step and by combining the prior knowledge about the running characteristic of the television; for the television with stable power, the television state identification can be completed by searching the television load event; for the power fluctuation type television, sequentially carrying out a load event searching step and television state identification based on a current space track; and finally, judging whether the television runs in the power fluctuation section.

The above steps are described in detail as follows:

firstly, according to the division of a power fluctuation section, carrying out power fluctuation type television prejudgment by utilizing the power fluctuation characteristic; the implementation flow is shown in fig. 2. According to the flow chart of the method,

firstly, carrying out piecewise linearization on the input total load active power time sequence shown in the formula (1).

P＝[P₁,P₂,P₃,...P_i,...P_n] (1)

Wherein, P_iThe total load active power value at the ith second is shown, and n is the total time length of the data. The piecewise power sequence shown in the formula (2) is obtained through piecewise linearization.

Wherein the content of the first and second substances,the active power value at the ith segmentation point is shown, and m represents the number of all segmentation points in the data.

And secondly, dividing a power fluctuation section. First to P_sCarrying out difference and taking an absolute value to obtain a segment point power difference absolute value sequence delta P_sAs shown in formula (3).

Wherein the content of the first and second substances,then at Δ P_sThe interval between the starting points in the sequence is less than a certain threshold T_th、|ΔP_s| greater than a certain threshold Δ P_{s_th}The elements of (a) are drawn into the same undulating section.

And thirdly, detecting the power difference outlier. Δ P contained in each fluctuation section_sThe sequence performs outlier detection and removes outliers to eliminate interference from load events. The power difference set of the remaining subsection points of each fluctuation section is delta P_{s_re}。

Fourthly, the power fluctuation type television is judged in advance. Screening is carried out for each fluctuation section if the delta P of the fluctuation section_{s_re}And (4) when the number of the elements in the set exceeds a certain threshold value C, starting fluctuation characteristic calculation and television existence prejudgment of the fluctuation section. If the formula (4) is satisfied, it is determined that the fluctuation section is suspected to exist in the operation of the television set

Wherein, mu (·), sigma (·) and Median (·) respectively represent the operation of calculating the mean, standard deviation and Median; delta Q_{s_re}Is equal to Δ P_{s_re}A corresponding reactive delta; k₁、K₂And K₃Is a corresponding preset threshold.

The invention uses a section of input active power data (as shown in figure 3) to carry out power fluctuation type television pre-judgment, and compares T with T_thTaken as 25 seconds,. DELTA.P_{s_th}Taking 10W, C60, K₁、K₂、K₃The pseudo power fluctuation type tv operation sections are obtained by setting 1/3, 1/3, and 80%, respectively, as shown by the dashed line boxes in fig. 3.

Step two, adopting a dynamic variable point detection method to detect the load event

The invention carries out load event detection on active power, and adopts a dynamic load event detection method: performing point-by-point scanning on the power second data, and when the absolute value of the power difference delta P between a scanning point and a subsequent point is larger than a predefined threshold delta P_thThe scanning point is regarded as a "change point". Dividing an input power sequence into a transient process section and a steady-state process section by detecting a change pointAlternating segments. Wherein the requirement of the steady state process section is that the process has no power change point and its time length is greater than a predefined threshold value Δ t_th. Each resulting load event record includes { t }_start,t_end，ΔP_evt,ΔQ_evt}. Wherein, t_startIs the starting moment of the transient process of the load event, t_endIs the end time, Δ P, of the transient process of the load event_evt、ΔQ_evtThe active and reactive power variation of the transient process of the load event.

The invention converts Delta P_thSelecting 30W, delta t_thSelecting 3s, and performing load event detection on a section of input total load active power to obtain a result shown in fig. 4, wherein the circle coverage mark is the detected load starting event, namely Δ P_evtDiamond-marked > 0 for detected load shedding events, i.e. Δ P_evt＜0。

And step three, extracting the load mark of the television and identifying the state.

Based on the power fluctuation section division and the load event detection result, the state identification of the television can be completed by combining with some prior knowledge about the television. For a power stable television, the state identification can be completed only by a load event searching step; the power fluctuation type television needs to be further identified by combining the current space trajectory. The method comprises the following specific steps:

1) TV set load event searching.

And searching the start and stop load events corresponding to the start and stop process of the television in the load event detection result based on the combined power characteristics of the television. First, the load event search range is defined: for the power fluctuation type television, in view of the power fluctuation characteristic existing in the operation process, searching is carried out in T time ranges before and after the starting time and the ending time of a fluctuation section so as to narrow the searching range, and a load event is associated with the fluctuation section; for a power stable television, the load event search range is all load events detected. Then, searching for a television set loading event: for the power fluctuation type television, the load starting event which is closest to the starting point of the fluctuation section and satisfies the formula (5) is taken as a suspected television load starting event; for a power stable television, all loading events that satisfy equation (5) are considered television loading events. And searching for the loading event which is closest to the starting point of the fluctuation section and satisfies the formula (5) as the suspected television loading event.

Wherein, the starting process of the long-start transient television can be represented as two load events, and the searching is performed according to the left formula in the formula (5), and the variable diagram is shown in fig. 5; for step start type TV, right type search can be used. (P)_{TV_min},P_{TV_max}) Is the television operating power range;is the set minimum threshold value of the power factor tangent value when the television is operated.

And finally, screening the load stopping event corresponding to the closing of the television to match with the load starting event of the television. The discrimination rule is shown in formula (6). The matching with the television set load starting events follows the nearest rule, namely, the load stopping event closest to the load starting event is selected from the load events meeting the judgment rule.

Wherein, Δ P_{TV_ON}For power variations during the start-up of a television set, for a long-start transient television set, Δ P_{TV_ON}The sum of the active power changes of two loading events in the starting process of the television is found; for step start-up television, Δ P_{TV_ON}Active power changes for load events during startup of such found televisions; delta P_{TV_th}Is the matching threshold for start-stop load events.

Invention t_{TV_ON}Is set to 15, (P)_{TV_min},P_{TV_max}) Set to (45,300) is carried out,set to 0.1, Δ P_thThe power load signature is used for carrying out non-intrusive television identification on the total load power data of a day in which a family including a television operates according to the scheme, wherein the total load power data is set to be 40W. The screening of the load event shown in fig. 4 can obtain the screening result of the television start-stop load event shown in fig. 6.

2) Television state identification based on current spatial trajectories.

The current space trajectory provided by the invention is utilized to further analyze the power fluctuation section so as to ensure the correctness of the television existence judgment and the screening result of the start-stop load event, and the implementation is divided into the following two steps.

Firstly, obtaining current data of a fluctuation section between start and stop load events of the television through differential operation. Before the difference operation, the fundamental current and the third harmonic current are uniformly converted under the voltage phase, as shown in formula (7).

Wherein the content of the first and second substances,is the angle of the power factor of the fundamental wave,the fundamental wave current and the third harmonic wave current in the form of phasor before and after conversion are treated in the same way. Then, the fundamental current and the third harmonic current sequence of the converted fluctuation section and the current of the previous steady-state section (the section between the transient sections of the adjacent load events) of the loading event of the television are subjected to phasor subtraction operation to eliminate background noise and the interference of other previous electrical appliances. The effective value sequence of the fundamental wave differential current and the effective value sequence of the third harmonic current of the fluctuation section can be obtained, as shown in formula (8), k is the sequence length.

And secondly, analyzing the third harmonic-fundamental wave current space track to realize television state identification. Firstly, the invention adopts a DBSCAN method delta I₁-ΔI₃Sample points in space are clustered. Then, television judgment is carried out on each cluster in the clustering result. Specifically, the judgment is carried out from two aspects of the parameter range and the shape of the point cluster. The point clusters generated by the independent operation of the television set satisfy the following conditions: a. the effective value of the fundamental current is (0, I)_1max) Internal; b. the ratio of the effective value of the third harmonic current in the cluster-like center to the effective value of the fundamental current should be greater than a certain threshold value r.

The shape of the cluster of decision points is then characterized as having a "continuous band". First, z-score normalization processing is performed on the fundamental current and the third harmonic current of the point cluster, respectively. Then, it is determined whether or not the formula (9) is established.

Wherein after normalization,. DELTA.I₁-ΔI₃In-plane, Point₁、Point₂Respectively sample points with the minimum and maximum horizontal coordinates in the point cluster; point (Point)₃ Point₄The vertical coordinate of the sample point with the minimum and the maximum near the horizontal center of the point cluster. dist (·,) is the Euclidean distance between two points, w is a predefined shape coefficient. If the formula (9) is established, the fluctuation section is determined as the television operation section. Otherwise, if other load events screened as the television starting still exist near the beginning of the fluctuation section, current space trajectory extraction and television state identification are carried out according to the other load events. Otherwise, judging that the television does not run in the fluctuation section.

The invention sets the following parameters: i is_1maxSet to 2A, r is 0.15 and w is 2. DBSCAN clustering by using fundamental current difference-third harmonic current difference result of power fluctuation section in FIG. 3The results shown in FIG. 7 were obtained. Further, each effective point cluster is determined, and the calculation result of the second type of point cluster by the equation (9) is 3.2872, as shown in fig. 8, it can be determined that the television operation exists in the fluctuation operation section.

While the present invention has been described with reference to the accompanying drawings, the present invention is not limited to the above-described embodiments, which are illustrative only and not restrictive, and various modifications which do not depart from the spirit of the present invention and which are intended to be covered by the claims of the present invention may be made by those skilled in the art.