A non-intrusive electrical appliance identification method based on harmonic similarity algorithm

Abstract

The application discloses a non-invasive electrical appliance identification method based on a harmonic similarity algorithm. The method comprises the following steps: 1, first, learning process is carried out on the electrical appliance to be identified, in the learning process, the fundamental active power, the fundamental reactive power, the active power of each harmonic and the reactive power of each harmonic of each sample electrical appliance under the rated condition are measured; 2, the operation number N of the electrical appliance identification algorithm is initialized, N=0, and the N=N+1 operation is started; 3, the voltage and current time domain data of the unknown electrical appliance on the bus are collected; 4, the collected voltage and current time domain data are converted into frequency domain data through fast Fourier transform (FFT); 5, according to the current and voltage harmonics of different frequencies in the frequency domain data, the total fundamental active power, the active power of each harmonic and the reactive power of each harmonic under the current power consumption state are calculated. The non-invasive electrical appliance identification method based on the harmonic similarity provided by the application has a wider application occasion compared with the prior art.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of non-intrusive detection system and electric appliance identification, and particularly relates to a non-intrusive electric appliance identification method based on a harmonic similarity algorithm. BACKGROUND

[0002] Monitoring and statistics of power consumption behavior can not only provide effective safety power consumption reminders for users, but also provide reliable analysis data for governments and electric appliance manufacturers. Through research on power consumption behavior, an effective power consumption behavior data collection and load accurate identification scheme and algorithm are determined, which has practical and long-term significance for improving power consumption safety and promoting industrial reform.

[0003] At present, in the traditional electric appliance identification method, an invasive load monitoring method is mostly used. The traditional invasive load monitoring needs to install a detection device in the electric appliance inside the target to be detected to obtain the on-off state information of the electric appliance. However, the monitoring equipment itself has a certain cost, and related maintenance and maintenance are required in the process of use, which increases the cost, and the user's home needs to be entered in the process of installation and maintenance, which affects the normal life and is difficult to be applied on a large scale. SUMMARY

[0004] The present application provides a non-intrusive electric appliance identification method based on a harmonic similarity algorithm, which can identify multiple electric appliances working at the same time without entering the user's home by using a similarity index. Compared with the existing technical method, the application occasion is more extensive.

[0005] To achieve the above object, the present application provides the following scheme:

[0006] A single-phase electric appliance identification method based on harmonic power similarity, comprising the following steps:

[0007] Step 1: selecting sample electric appliances, first learning the electric appliances to be identified, and measuring the fundamental active power, fundamental reactive power and active power and reactive power of each harmonic of each sample electric appliance under standard conditions in the learning process;

[0008] Step 2: initializing the operation number N of the electric appliance identification algorithm to N=0, and starting the N=N+1th operation;

[0009] Step 3: collecting voltage and current time domain data of unknown electric appliances on the bus;

[0010] Step 4: converting the collected voltage and current time domain data into frequency domain data through fast Fourier transform (FFT);

[0011] Step 5: according to the different times of current and voltage harmonics in the frequency domain data, the total active power and reactive power of the fundamental wave and each harmonic are calculated under the current power consumption state;

[0012] Step 6: the similarity data of the recognition result is given by the constructed similarity function;

[0013] Step 7: whether the result similarity is greater than or equal to 0.9 is judged, if satisfied, the recognition result and the similarity are output; if not satisfied, whether the algorithm operation number N is greater than 10 times is judged, if satisfied, the optimal recognition result and the similarity are output; if not satisfied, the step 2 is returned to operate again.

[0014] Optionally, the step 1: selecting sample electrical appliances, firstly, the learning process of the electrical appliances to be identified is carried out, in the learning process, the fundamental wave active power, the fundamental wave reactive power and the active power and the reactive power of each harmonic of each sample electrical appliance under the standard condition are measured, and the highest number of the harmonics is seven.

[0015] Optionally, the step 5: the similarity data of the recognition result is given by the constructed similarity function, specifically including:

[0016] The similarity function is constructed;

[0017] The power values of the fundamental wave and each harmonic learned in the learning process and the power values of the fundamental wave and each harmonic of the unknown electrical appliance in the measurement process are substituted into the similarity function;

[0018] The power consumption situation of the monitoring target can be obtained by solving the similarity function.

[0019] Optionally, the similarity function is constructed, specifically including:

[0020] The state matrix A is established;

[0021] According to the state matrix A and the harmonic power of each sample electrical appliance learned in the learning process and the power of the 2k-1(k=1, 2, 3...)th harmonic on the bus in the actual situation obtained in the measurement process, the active power deviation vector ΔP of 2n combinations is calculated 2k-1 , the active power deviation vector ΔQ 2k-1 ;

[0022] According to the active power deviation vector ΔP 2k-1 , the reactive power deviation vector ΔQ 2k-1 , the relative deviation vector ΔP 2k-1 rel of the active power of each scheme and the actual situation is calculated respectively 2k-1 rel ;

[0023] a relative deviation vector of active power ΔP 2k-1 rel a relative deviation vector of reactive power ΔQ 2k-1 rel calculating a similarity vector S.

[0024] Optionally, the solving of the similarity function can obtain the power consumption of the current monitoring target, and specifically includes:

[0025] The appliance working state row vector in the state matrix A corresponding to the element closest to 1 in the similarity vector S is selected as the recognition result.

[0026] Compared with the prior art, the technology has the following beneficial effects:

[0027] The non-intrusive appliance recognition method based on the harmonic similarity algorithm provided by the application includes the following steps: (i) learning process; (ii) measurement process; (iii) calculation process; and (iv) recognition process. In the learning process, the fundamental wave active power, the fundamental wave reactive power, and the active power and the reactive power of some high-order harmonics of each sample appliance under standard conditions are measured. In the measurement process, the voltage and current time domain data of an unknown appliance on a bus are collected. In the calculation process, the active power and the reactive power of each harmonic under the current power consumption state are calculated according to the current and voltage harmonics of different orders in the frequency domain. In the recognition process, the similarity function is designed, the fundamental wave and harmonic power values learned in the learning process and the fundamental wave and harmonic power values of the unknown appliance measured in the measurement process are substituted into the similarity function, and the solving of the similarity function can obtain the power consumption of the current monitoring target. The active power and the reactive power of each harmonic of the appliance are matched with the data obtained in the learning process, and the type of the sample appliance is recognized. BRIEF DESCRIPTION OF DRAWINGS

[0028] In order to more clearly illustrate the technical solutions of the embodiments of the application or the prior art, the following will briefly introduce the drawings needed in the embodiments. Obviously, the drawings in the following description are only some embodiments of the application, and for those skilled in the art, other drawings can also be obtained from these drawings without creative labor.

[0029] Figure 1 The flow chart of the non-intrusive appliance recognition method based on the harmonic similarity algorithm of the embodiment of the application;

[0030] Figure 2 The distribution graph of the harmonic power of the No. 1 appliance of the embodiment of the application;

[0031] Figure 3 The distribution graph of each harmonic power of the No. 2 electric appliance in the embodiment of the present application is shown in the following table:

[0032] Figure 4 The distribution graph of each harmonic power of the No. 3 electric appliance in the embodiment of the present application is shown in the following table:

[0033] Figure 5 The distribution graph of each harmonic power of the No. 4 electric appliance in the embodiment of the present application is shown in the following table. DETAILED DESCRIPTION

[0034] The technical solutions in the embodiments of the present application will be described clearly and completely below with reference to the drawings in the embodiments of the present application. Obviously, the described embodiments are only a part of the embodiments of the present application, rather than all the embodiments of the present application. Based on the embodiments in the present application, all other embodiments obtained by those skilled in the art without creative work fall within the scope of the present application.

[0035] The present application provides a single-phase electric appliance identification method based on harmonic power similarity algorithm, which can identify multiple electric appliances working simultaneously without entering the user's home by using a similarity index. Compared with the prior art, the application has a wider application.

[0036] In order to make the above-mentioned objects, features and advantages of the present application more obvious and easy to understand, the present application will be further described in detail below with reference to the drawings and specific embodiments.

[0037] Figure 1 The flow chart of the non-intrusive electric appliance identification method based on harmonic similarity algorithm in the embodiment of the present application is shown in the following figure, which aims to identify the current power consumption of the monitoring object by collecting the relevant data of the monitoring object's outdoor bus, including the following steps: Figure 1

[0038] Step 1: The electric appliance to be identified is first subjected to a learning process. In the learning process, the fundamental wave active power, reactive power and some high-order harmonic active power and reactive power of each sample electric appliance under standard conditions are measured. The harmonic order to be measured of the learning object can reach the seventh harmonic.

[0039] Step 2: Collect the voltage and current time domain data of the unknown electric appliance on the bus;

[0040] Step 3: Convert the collected voltage and current time domain data into frequency domain data by fast Fourier transform (FFT);

[0041] Step 4: According to the different order of current and voltage harmonics in the frequency domain, the active and reactive data of each harmonic under the current power consumption state are calculated.​

[0042] Step 5: The recognition result is given a recognition similarity value by the constructed similarity function.

[0043] In the step (5), the matching process is: (i) establishing a similarity function; (ii) substituting the power values of the fundamental wave and each harmonic wave learned in the learning process and the power values of the fundamental wave and each harmonic wave of the unknown electrical appliance in the measurement process into the similarity function; (iii) solving the similarity function can obtain the power consumption situation of the current monitoring target.

[0044] Step 6: Matching the active and reactive power data of each harmonic wave of the electrical appliance obtained with the data obtained in the previous learning process to identify the type of the sample electrical appliance.

[0045] Suppose there are n electrical appliances participating in identification, and the specific identification process is:

[0046] First step: initialize the state to let the algorithm operation number N = 0. The electrical appliances to be identified are first subjected to a learning process, in which the active power and reactive power of the fundamental wave and the active power and reactive power of some high-order harmonics of each sample electrical appliance under standard conditions are measured; the harmonic order to be measured of the learning object can reach the seventh harmonic. n2k-1 P2k-1, n represents the 2k-1th harmonic active power of the nth sample electrical appliance learned in the learning process, k = 1, 2, 3, 4. n2k-1 Q2k-1, n represents the 2k-1th harmonic reactive power of the nth sample electrical appliance learned in the learning process, k = 1, 2, 3, 4.

[0047] Second step: let the algorithm operation number N = N + 1. The voltage and current time domain data of the unknown electrical appliance on the bus are collected, and the collected voltage and current time domain data are converted into frequency domain data by fast Fourier transform (FFT) and the 2k-1th harmonic active and reactive data values P 2k-1 mea , Q 2k-1 mea k = 1, 2, 3, 4.

[0048] Third step: establish a similarity function, first establish a state matrix A, which is in the following form:

[0049]

[0050] Each element in the A matrix is 0 or 1, each row represents a working state combination of an electrical appliance; since n electrical appliances are involved in the identification, each row has n elements, and 1 in each row represents that the corresponding electrical appliance is working, and 0 represents that the corresponding electrical appliance is not working, for example, in the 2n-1 row, the elements are (111...10), the first n-1 elements are 1, and the last element is 0, which indicates that the first n-1 electrical appliances are working, and the nth electrical appliance is not working; the rows in the A matrix are arranged from top to bottom according to the binary coding rule, and there are 2n rows in total, which represent all possible working conditions of n electrical appliances, and the dimension of the A matrix is 2n x n. n n n

[0051] The state matrix A and the harmonic power of each sample electrical appliance learned in the learning process and the actual power of the 2k-1 (k=1, 2, 3...) harmonic on the bus obtained in the measurement process are used to calculate 2n combinations of power deviation vectors ΔP and ΔQ. n 2k-1 2k-1 Wherein:

[0052]

[0053] ΔP 2k-1 represents the active power deviation vector of the 2k-1 harmonic, ΔP 2k-1(i) (i=1, 2...2 n ) represents the active power deviation value of the 2k-1 harmonic corresponding to the ith combination; P n2k-1 represents the active power of the 2k-1 harmonic of the nth sample electrical appliance learned in the learning process; P 2k-1 mea is the active power value of the 2k-1 harmonic on the bus obtained in the measurement process. ΔQ 2k-1 The calculation process is similar, as follows:

[0054]

[0055] ΔQ 2k-1 represents the reactive power deviation vector of the 2k-1 harmonic, ΔQ 2k-1(i) (i=1, 2...2 n ) represents the reactive power deviation value of the 2k-1 harmonic corresponding to the ith combination; Q n2k-1 represents the reactive power of the 2k-1 harmonic of the nth sample electrical appliance learned in the learning process; Q 2k-1 mea is the reactive power value of the 2k-1 harmonic on the bus obtained in the measurement process.

[0056] ​​​​​Calculate the relative deviation vector ΔP between each solution and the actual situation 2k-1 rel , ΔQ 2k-1 rel , the specific calculation method is as follows:

[0057]

[0058] Vector ΔP 2k-1 rel The i-th (i=1,2...,2 n ) element represents the relative deviation between the theoretically calculated 2k-1th harmonic active power and the 2k-1th harmonic active power measured on the actual bus in the i-th solution. 2k-1 rel The calculation method of is similar to that of , as follows:

[0059]

[0060] Vector ΔQ 2k-1 rel The i-th (i=1,2...,2 n ) element represents the relative deviation between the theoretically calculated 2k-1th harmonic reactive power and the 2k-1th harmonic reactive power measured on the bus in the i-th scenario. Each electrical appliance is connected to the bus, and the total current and voltage data on the bus are collected and analyzed during identification.

[0061] Establish a similarity function and calculate the similarity vector S, where the i-th element S i Represents the similarity percentage between the i-th solution and the actual situation.

[0062] Similarity S of the i-th solution i The calculation method is:

[0063] S i =1-(p1×ΔP 1i rel +p3×ΔP 3i rel +…p 2k-1 ×ΔP 2k-1i rel +q1×ΔQ 1i rel +q3×ΔQ 3i rel +…q 2k-1 ×ΔQ 2k-1i rel )

[0064] where p1, p3…p 2k-1 with q1,q3…q 2k-1are weight coefficients, and p1+p3+…+p 2k-1 =0.5, q1+q3+…+q 2k-1 =0.5. When the low-order harmonic wave power features of the to-be-identified electrical appliances are similar, the weight coefficients can be adjusted to prevent the high-order harmonic wave power features with small values from being submerged, and in normal cases, the weight coefficients can be defaulted to be equal. The similarity values of each scheme to the true situation can be calculated by the above formula, and the result is closer to 1, which indicates that the scheme is closer to the true situation, and if an element is less than 0, it is processed as equal to 0. n

[0065] Fourth step: selecting the electrical appliance working state row vector A in the A matrix corresponding to the element in the vector S closest to 1 i (here, the ith combination in the A matrix is the current optimal identification result), which is the identification result of this time.

[0066] Fifth step: judging whether the result similarity is greater than or equal to 0.9, if yes, outputting the identification result and the similarity; if no, judging whether the algorithm operation number N is greater than 10 times, if yes, outputting the current optimal identification result and the similarity; if no, returning to the second step to perform operation again.

[0067] The above algorithm is simulated and verified by using MATLAB software:

[0068] In the simulation process, there are four electrical appliances, and the seventh harmonic wave is collected for each electrical appliance, n = 4, and k = 4. The power conditions of each electrical appliance when normally running are shown in Table 1, and the harmonic wave power condition distribution diagram of each electrical appliance is shown in Figure 1. Figures 2 to 5

[0069] Table 1 Harmonic wave power of each electrical appliance in the learning process

[0070]

[0071] After the learning process is completed, the measurement and identification process can be started. First, a plurality of electrical appliances are arbitrarily run, and in this example, the scene is to run the No. 1 and No. 2 electrical appliances. The power conditions of each harmonic wave on the bus under the current state are measured and shown in Table 2.

[0072] Table 2 Power conditions in the measurement process

[0073]

[0074]

[0075] The data obtained by the above two processes are substituted into the algorithm, and the similarity data corresponding to each scheme are obtained by MATLAB simulation and shown in Table 3.

[0076] ​​Table 3 Similarity data table corresponding to all schemes

[0077]

[0078] According to the result, it can be seen that the algorithm calculates the most similar appliance working state row vector as (1100), and the similarity reaches 0.9581, which indicates that the situation that the first and second appliances work and the third and fourth appliances do not work is the closest to the actual situation, and the actual situation is that the first and second appliances work, which indicates that the verification result is completely consistent with the actual situation, directly verifying the effectiveness of the algorithm.

[0079] The non-intrusive appliance identification method based on the harmonic similarity algorithm provided in the application first performs a learning process on the appliances to be identified, and in the learning process, the fundamental active power, the fundamental reactive power and the active power and the reactive power of some high-order harmonics of each sample appliance under standard conditions are measured; the voltage and current time domain data of the unknown appliances on the bus are collected; the collected voltage and current time domain data are converted into frequency domain data through fast Fourier transform (FFT); according to the current and voltage harmonics of different orders in the frequency domain, the active and reactive power data of each order under the current power state are calculated; and the recognition similarity data are given to the recognition result through the designed similarity function. The specific identification process is as follows: (i) establishing a similarity function; (ii) substituting the fundamental and harmonic power values learned in the learning process and the fundamental and harmonic power values of the unknown appliances measured in the measurement process into the similarity function; (iii) solving the similarity function can obtain the power consumption situation of the monitoring target. The active and reactive power data of each order of the appliances obtained are matched with the data obtained in the previous learning process, and the type of the sample appliance is identified. The non-intrusive appliance identification method based on the harmonic similarity algorithm provided in the application can identify the appliances under the condition that multiple appliances work at the same time without entering the user's home by using the similarity index, and the application occasion of the method is more extensive compared with the existing technical methods.

[0080] The embodiments in the specification are described in a progressive manner, and each embodiment focuses on the difference from other embodiments, and the same or similar parts between the embodiments can be referred to each other.

[0081] The principles and implementation manners of the application are described by using specific examples in this paper, and the above embodiment description is only used to help understand the method of the application and its core idea; at the same time, for those skilled in the art, according to the idea of the application, the specific implementation manner and application range will be changed. In conclusion, the content of the specification should not be understood as a limitation of the application.

Claims

1. A non-intrusive electrical appliance identification method based on harmonic similarity algorithm, characterized in that: The following steps are involved: Step 1: Select sample appliances and first conduct a learning process on the appliances to be identified. During the learning process, measure the fundamental wave active power, fundamental wave reactive power, and active power and reactive power of each harmonic of each sample appliance under standard conditions; Step 2: Initialize the number of times N of the appliance identification algorithm is run to set N=0, and start the N=N+1th run; Step 3: Collect the voltage and current time domain data of the unknown electrical appliance on the bus; Step 4: Convert the collected time domain data of voltage and current into frequency domain data through fast Fourier transform (FFT); Step 5: Based on the current and voltage harmonics of different orders in the frequency domain data, calculate the active power and reactive power of the total fundamental wave and each harmonic under the current power consumption state; Step 6: Use the constructed similarity function to give the recognition similarity data for the recognition results; Step 7: Determine whether the similarity of the result is greater than or equal to 0.

9. If so, output the recognition result and similarity. If not, determine whether the number of algorithm operations N is greater than 10. If so, output the current optimal recognition result and similarity. If not, return to step 2 and perform the operation again. Step 6: Providing recognition similarity data for the recognition results using the constructed similarity function, specifically including: Construct similarity function; Substitute the power values ​​of the fundamental wave and each harmonic learned in the learning process and the power values ​​of the fundamental wave and each harmonic of the unknown electrical appliance in the measurement process into the similarity function; Solving the similarity function can obtain the current power consumption of the monitored target; The constructing of the similarity function specifically includes: Establish state matrix A; According to the state matrix A and the harmonic power of each sample electrical appliance learned in the learning process, the power of the 2k-1th (k=1,2,3...)th harmonic on the bus in the actual situation obtained during the measurement process is calculated to obtain 2 n Active power deviation vector ΔP 2k-1 , reactive power deviation vector ΔQ 2k-1 ; According to the active power deviation vector ΔP 2k-1 , reactive power deviation vector ΔQ 2k-1 Calculate the relative deviation vector ΔP of active power between each scheme and the actual situation respectively 2k-1 rel , the relative deviation vector of reactive power ΔQ 2k-1 rel ; According to the relative deviation vector ΔP of active power 2k-1 rel , the relative deviation vector of reactive power ΔQ 2k-1 rel Calculate the similarity vector S.

2. The non-invasive electrical appliance identification method based on harmonic similarity algorithm according to claim 1 is characterized in that: Step 1: Select sample electrical appliances and first perform a learning process on the electrical appliances to be identified. During the learning process, the fundamental wave active power, fundamental wave reactive power, and active power and reactive power of each harmonic of each sample electrical appliance under standard conditions are measured, where the highest order of each harmonic is seven.

3. The non-invasive electrical appliance identification method based on harmonic similarity algorithm according to claim 1, characterized in that: Solving the similarity function can obtain the current power consumption of the monitored target, specifically including: The row vector of the electrical appliance working state in the state matrix A corresponding to the element closest to 1 in the similarity vector S is selected as the recognition result.

Images