A non-household load identification method based on average load and total demand distortion rate

Abstract

The present invention discloses a method for identifying non-household loads based on average load and total demand distortion rate, relating to the technical field of load identification. The method comprises: collecting voltage and current signals at the household electricity inlet and performing filtering processing; calculating average conductance, average susceptance, and average admittance modulus and establishing a single-load offline feature library based on these three characteristic quantities; using a difference sequence of average conductance, average susceptance, average admittance modulus, and total demand distortion rate to detect load switching events and make steady-state judgments; and using average conductance, average susceptance, and average admittance modulus during steady-state operation to decompose and identify loads. The average conductance, average susceptance, average admittance, and total demand distortion rate characteristic quantities introduced in the present invention relatively comprehensively encompass the physical characteristics of the load, and the calculation process is simple when implementing load identification.

Description

Technical Field

[0001] The present invention relates to the field of residential household electricity load identification, and particularly to a non-in-home load identification method based on average load and total demand distortion rate. Background Art

[0002] With the construction and development of the global energy Internet, an interactive intelligent power grid service system has been initially built in China. The proportion of the total residential household electricity load in the power load is increasing, and the importance of the two-way interactive feedback between the power supply grid and residential users in the intelligent power grid is becoming more and more obvious. Identifying the residential household electricity load can not only provide a guiding basis for users' rational electricity consumption, but also be beneficial to the power grid for demand-side management.

[0003] Currently, load identification methods can be divided into two categories: in-home and non-in-home. The in-home load identification method requires installing monitoring devices on each household appliance to obtain load data. Although this method has the advantage of simple data processing, it has high hardware costs and no privacy, and the user experience is poor. The non-in-home load identification method only needs to install a monitoring device at the residential in-home end, and the purpose of load identification can be achieved by analyzing the voltage and current data of the residential in-home main line. Although this method has high requirements for the load identification algorithm, it greatly reduces the hardware cost and improves the user experience.

[0004] The non-in-home load identification method generally includes five steps: data acquisition and processing, event detection, feature extraction, and load decomposition and identification. Data acquisition is to sample the voltage and current waveforms, and usually the sampling data can be obtained through an electric energy meter; data processing mainly realizes the function of filtering and denoising, and common software algorithms include median filtering, mean filtering, and Gaussian filtering algorithms. Event detection is an important basis for load decomposition and identification. Event detection is to detect the load on / off events, and currently the effective current value and active power are commonly used as the load characteristic indicators for event detection. Load characteristics reflect the physical characteristics of a certain aspect of the load, and appropriate characteristics can directly affect the operation time and accuracy of load identification. Currently, load characteristics are mainly divided into two categories: transient characteristics and steady-state characteristics. Steady-state characteristics mainly include power, V-I trajectory, harmonic amplitude and phase angle, etc., and transient characteristics include transient duration, impulse current multiple, etc. Load decomposition and identification are carried out according to load characteristics and related algorithms, and genetic algorithms, particle swarm algorithms, and chicken swarm algorithms are often applied to load decomposition and identification.

[0005] The present invention proposes a non-household load identification method based on average load and total demand distortion rate. This method uses three characteristic quantities, namely average conductance, average susceptance, and average admittance modulus, which represent the average load, to establish a characteristic library, and uses the difference sequences of average conductance, average susceptance, average admittance modulus, and total demand distortion rate to detect load switching events and judge the steady state. At the same time, the average conductance, average susceptance, and average admittance modulus in the steady state are used for load decomposition and identification. Since the average conductance, average susceptance, average admittance modulus, and total demand distortion rate contain load physical characteristic information such as the amplitudes of each harmonic of the current, the fundamental voltage amplitude, and the fundamental power factor, the characteristic quantities used in the present invention comprehensively contain the physical characteristics of the load, and the calculation process is simple in the five steps of realizing load identification. Summary of the Invention

[0006] The purpose of the present invention is to provide a non-household load identification method based on average load and total demand distortion rate. This method selects four characteristic quantities, namely average conductance, average susceptance, average admittance modulus, and total demand distortion rate, which comprehensively contain load physical characteristic information, and can realize the identification of residential household electricity loads through a relatively simple calculation process. To achieve the above purpose, the technical solution adopted by the present invention is as follows:

[0007] A non-household load identification method based on average load and total demand distortion rate, comprising the following steps:

[0008] S0: Set the flag bit Flag = 0;

[0009] S1: Collect data of M complete cycles of the voltage and current waveforms at the household electricity inlet of a residential household;

[0010] S2: Filter the collected current data;

[0011] S3: Calculate the average conductance G, average susceptance B, average admittance modulus |Y|, and total demand distortion rate TDD for each cycle;

[0012] S4: Calculate the difference sequences ΔG, ΔB, Δ|Y|, and ΔTDD of the average conductance G, average susceptance B, average admittance modulus |Y|, and total demand distortion rate TDD for two adjacent cycles;

[0013] S5: Judge whether the flag bit Flag is 0. If Flag = 0, then perform S6; if Flag ≠ 0, then perform S7;

[0014] S6: Threshold judgment is performed based on the difference sequences ΔG, ΔB, and Δ|Y| of the average conductance G, average susceptance B, and average admittance modulus |Y|. If the difference of any of the above characteristic quantities exceeds its respective upper threshold, it is determined that a load switching event has occurred, and the switching time t1 is recorded, then S7 is performed; if the differences of all three characteristic quantities do not exceed their respective upper thresholds, it is determined that no load switching event has occurred and S1 is returned;

[0015] S7: Threshold judgment is performed on the difference sequences ΔG, ΔB, Δ|Y|, and ΔTDD of the average conductance G, average susceptance B, average admittance modulus |Y|, and total demand distortion rate TDD after switching occurs. If the continuous D differences of the above four characteristic quantities do not exceed their respective lower thresholds, it is determined that the switched load has entered a steady state, and the steady state entry time t2 is recorded. Then, the average conductance G, average susceptance B, and average admittance modulus |Y| in the first cycle after entering the steady state are extracted as the characteristic values of the steady state after load switching, and then S8 is performed; if the difference of any one of the characteristic quantities exceeds its respective lower threshold, it is determined that the switched load has not entered the steady state, the flag bit Flag = 1 is set, and S1 is returned;

[0016] S8: The genetic algorithm is used to calculate the characteristic quantities of the steady state of the newly connected or disconnected load eigenvalues obtained in S6 and the characteristic quantities in the off-line characteristic library, and load identification of the newly switched load is realized, and S0 is returned.

[0017] The following is a further principle explanation of the above steps.

[0018] 1. The expressions of the supply voltage u(t) and load current i(t) at the household electricity inlet of residential households are:

[0019]

[0020] where u(t) is the sinusoidal voltage provided by the power supply department, T is the power grid power frequency period, U is the voltage effective value, θ1 is the voltage initial phase angle, I h is the rms value of the hth current harmonic, is the phase angle value of the hth current harmonic, h is the harmonic order, h = 1, 2…∞;

[0021] By performing full-cycle acquisition on u(t) and i(t), the sampled data u(n) and i(n) are obtained as:

[0022]

[0023] where N is the number of single-cycle sampling points, n = 1,…, N.

[0024] 2. The calculations of the average conductance G, average susceptance B, average admittance modulus |Y|, and total demand distortion rate TDD are:

[0025]

[0026] From the above formula, it can be obtained that:

[0027]

[0028] In the formula, P is the active power, Q is the reactive power, S is the fundamental apparent power, and I0 is the rated current demand of residential household electricity consumption.

[0029] The average conductance G, average susceptance B, average admittance modulus |Y|, and total demand distortion rate TDD can be calculated using the sampled data u(n) and i(n):

[0030]

[0031]

[0032]

[0033]

[0034]

[0035] I1 = |Y|U(10)

[0036]

[0037] In the formula, v(n) is the voltage sampled data lagging behind u(n) by

[0038] 3. Before load identification, an offline feature library of residential household electricity loads is established in advance. According to steps S1, S2, and S3, data collection and processing are carried out for each single load under steady-state operating conditions, and the average conductance G, average susceptance B, average admittance modulus |Y|, and total demand distortion rate TDD are calculated to establish an offline feature library of single loads including all loads. The feature library includes Z G = [Z G1 , Z G2 … Z Gm , Z B = [Z B1 , Z B2 … Z Bm and Z |Y| = [Z |Y|1 , Z |Y|2 … Z |Y|m , where m is the total number of user loads.

[0039] Further, in the above S2, the median filtering method is used to filter the current sampling data. This method takes [i(q - 1), i(q), i(q + 1)] as the window, where q = 1, …, N, and the filtered current value i(q) is the median value of the sampling data [i(q - 1), i(q), i(q + 1)] before filtering.

[0040] Further, in the above S4, the calculation of the difference sequences of the average conductance G, average susceptance B, average admittance modulus |Y|, and total demand distortion rate TDD for two adjacent cycles is as follows:

[0041]

[0042] where p = 1, …, M - 1.

[0043] Further, in the above S6, the upper limit thresholds of the differences between the average conductance G, average susceptance B, and average admittance modulus |Y| are λ G+ , λ B+ and λ |Y|+ , λ G+ , λ B+ and λ |Y|+ are respectively 10% of the minimum values of G, B, and |Y| in the offline feature library.

[0044] Further, in the above S7, the lower limit thresholds of the differences between the average conductance G, average susceptance B, average admittance modulus |y|, and total demand distortion rate TDD are λ G- , λ B- , λ |Y|- and λ TDD- , λ G- , λ B- , λ |Y|- and λ TDD- are respectively 1%, 1%, 1%, and 2% of the minimum values of G, B, |Y|, and TDD in the offline feature library.

[0045] Further, in the above S8, the specific steps for load decomposition and identification based on the genetic algorithm are as follows:

[0046] First, configure the genetic algorithm parameters. The genetic algorithm parameters include the population size s, crossover rate Pn, mutation rate Mr, and fitness function F. The expression of the fitness function F is:

[0047] F min =(H G -Z G *X)+(H B -Z B *X)+(H |Y| -Z |Y| *X)(13)

[0048] where H G is the steady-state average conductance G value, H B is the steady-state average susceptance B value, H |Y| is the steady-state average admittance modulus |Y|, Z G =[Z G1 , Z G2 …Z Gm is the single-electrical-appliance average conductance G feature library, Z B =[Z B1 , Z B2 …Z Bm is the single-electrical-appliance average susceptance B feature library, Z |Y| =[Z |Y|1 , Z |Y|2 …Z |Y|m is the single-electrical-appliance average admittance modulus |Y| feature library, m is the number of user loads, X = [x1, x2…x m T , x j =0 indicates the off state of the electrical appliance, and x j =1 indicates the on state of the electrical appliance, j = 1, 2…m. Then, the minimum fitness value is obtained according to the fitness function, and it is judged whether the minimum fitness value meets the termination iteration condition. If it meets, the iteration is terminated; otherwise, the iteration continues until the maximum iteration number is reached. BRIEF DESCRIPTION OF THE DRAWINGS

[0049] The drawings of the present invention are described as follows.

[0050] Figure 1 is a schematic diagram of the working process of the non-household load identification method based on the average load and the total demand distortion rate of the present invention. SPECIFIC IMPLEMENTATION MODE

[0051] The embodiments of the present invention will be described in detail with reference to the drawings.

[0052] The present invention provides a non-household load identification method based on the average load and the total demand distortion rate. The method includes the following steps:

[0053] S0: Set the flag bit Flag = 0;

[0054] S1: N takes the value of 32, M takes the value of 20, and the voltage and current waveforms at the household electricity inlet of the residential household are collected for 20 full cycles of data at 32 points per cycle, and the sampling data of the voltage u(n) and i(n) are obtained, where n = 1, 2…20×32;

[0055] ​S2: Filter the current sampling data using the median filtering method. This method takes [i(q - 1), i(q), i(q + 1)] as the window, where q = 1, …, N, and the filtered current value i(q) is the median value of the sampling data [i(q - 1), i(q), i(q + 1)] before filtering;

[0056] S3: Calculate the average conductance G, average susceptance B, average admittance modulus |Y|, and total demand distortion rate TDD for each cycle:

[0057]

[0058]

[0059]

[0060]

[0061]

[0062] I1 = |Y|u (6)

[0063]

[0064] where v(n) is the voltage data lagging behind u(n), I0 = 20A is the rated current demand of residential household electricity, and the values of G, B, |Y|, and TDD for 20 cycles are obtained through the above calculations; Before load identification, according to steps S1, S2, and S3, data sampling and processing are carried out on each single load under the steady - state operation condition, and the average conductance G, average susceptance B, average admittance modulus |Y|, and total demand distortion rate TDD are calculated to pre - establish an offline feature library for single loads;

[0065] In addition, before load identification, according to steps S1, S2, and S3, data sampling and processing are carried out on each single load under the steady - state operation condition, and the average conductance G, average susceptance B, average admittance modulus |Y|, and total demand distortion rate TDD are calculated to pre - establish an offline feature library for single loads;

[0066] S4: Calculate the difference sequences ΔG, ΔB, Δ|Y|, and ΔTDD of the average conductance G, average susceptance B, average admittance modulus |Y|, and total demand distortion rate TDD between two adjacent cycles. The calculation formulas are as follows:

[0067]

[0068] where p = 1, …, 19;

[0069] S5: Judge whether the flag bit Flag is 0. If Flag = 0, then proceed to S6; if Flag ≠ 0, then proceed to S7;

[0070] S6: Threshold judgment is performed based on the difference sequences ΔG, ΔB, and Δ|Y| of the average conductance G, average susceptance B, and average admittance modulus |Y|. If any one of the values of ΔG(p), ΔG(p), and Δ|Y|(p) is greater than the corresponding upper threshold λ G+ , λ B+ , and λ |Y|+ , it is determined that a load switching event has occurred, and the switching time t1 is recorded, then S7 is performed; if the differences of the three characteristic quantities do not exceed their respective corresponding upper thresholds, it is determined that no load switching event has occurred and S1 is returned; λ G+ , λ B+ , and λ |Y|+ are respectively 10% of the minimum values of G, B, and |Y| in the off-line feature library;

[0071] S7: Threshold judgment is performed on the difference sequences ΔG, ΔB, Δ|Y|, and ΔTDD of the average conductance G, average susceptance B, average admittance modulus |Y|, and total demand distortion rate TDD after switching. If the continuous three differences of the above four characteristic quantities do not exceed their respective corresponding lower thresholds λ G- , λ B- , λ |Y|- , and λ TDD- , it is determined that the switched load has entered a steady state, and the steady-state entry time t2 is recorded. Then, the average conductance G, average susceptance B, and average admittance modulus |Y| in the first cycle after entering the steady state are extracted as the characteristic values of the steady state after load switching, and then S8 is performed; if the difference of any one characteristic quantity exceeds its respective corresponding lower threshold, it is determined that the switched load has not entered the steady state, the flag bit Flag = 1 is set and S1 is returned; λ G- , λ B- , λ |Y|- , and λ TDD- are respectively 1%, 1%, 1%, and 2% of the minimum values of G, B, and |Y| in the off-line feature library;

[0072] S8: The genetic algorithm is used to perform load decomposition and identification on the above steady-state characteristic quantities and the characteristic quantities in the off-line feature library, and S0 is returned. The specific steps of load decomposition and identification are as follows:

[0073] First, configure the genetic algorithm parameters. The genetic algorithm parameters include: population size s = 1000, crossover rate Pn = 0.5, mutation rate Mr = 0.1, fitness function F, and the expression of the fitness function F is:

[0074] F min =(H G -Z G *X)+(H B -Z B *X)+(H |Y| -Z |Y|*X) (9)

[0075] In the formula, H G is the steady-state average conductance G value, H B is the steady-state average susceptance B value, H |Y| is the steady-state average admittance modulus |Y|, Z G = [Z G1 , Z G2 … Z Gm is the single-electrical-appliance average conductance G feature library, Z B = [Z B1 , Z B2 … Z Bm is the single-electrical-appliance average susceptance B feature library, Z |Y| = [Z |Y|1 , Z |Y|2 … Z |Y|m is the single-electrical-appliance average admittance modulus |Y| feature library, m = 12 is the number of user loads, X = [x1, x2… x m T , x j = 0 represents the off state of the electrical appliance, and x j = 1 represents the on state of the electrical appliance, j = 1, 2… m. Then, the minimum fitness value is obtained according to the fitness function, and it is judged whether the minimum fitness value meets the termination iteration condition. If it meets, the iteration is terminated; otherwise, the iteration continues until the maximum iteration number is reached.

[0076] In summary, the non-in-home load identification method of this embodiment is based on the average load and the total demand distortion rate. Compared with the selection and calculation of feature quantities in existing load identification research, this method introduces the average conductance, average susceptance, average admittance modulus representing the average load, and the total demand distortion rate as feature quantities, which can more comprehensively contain the physical feature information of the load and simplify the calculation.

[0077] The above content is only an example to illustrate the process of the present invention and does not impose any limitations on the present invention. All technical solutions falling within the idea of the present invention belong to the protection scope of the present invention.​

Claims

1. A non-household load identification method based on average load and total demand distortion rate, characterized in that: including the following steps: S0: Set the flag bit Flag = 0; S1: Collect data of the voltage and current waveforms at the household power inlet of the residential household for M complete cycles; S2: Filter the collected current data; S3: Calculate the average conductance G, average susceptance B, average admittance modulus |Y|, and total demand distortion rate TDD for each cycle; S4: Calculate the difference sequences ΔG, ΔB, Δ|Y|, and ΔTDD of the average conductance G, average susceptance B, average admittance modulus |Y|, and total demand distortion rate TDD between two adjacent cycles; S5: Judge whether the flag bit Flag is 0. If Flag = 0, then perform S6; if Flag ≠ 0, then perform S7; S6: Perform threshold judgment according to the difference sequences ΔG, ΔB, and Δ|Y| of the average conductance G, average susceptance B, and average admittance modulus |Y|. If any one of the differences in the difference sequences ΔG, ΔB, and Δ|Y| exceeds its respective upper threshold, it is determined that a load switching event has occurred and record the switching time t1, then perform S7; if the difference sequences ΔG, ΔB, and Δ|Y| do not exceed their respective upper thresholds, it is determined that no load switching event has occurred and return to S1; S7: Perform threshold judgment on the difference sequences ΔG, ΔB, Δ|Y|, and ΔTDD of the average conductance G, average susceptance B, average admittance modulus |Y|, and total demand distortion rate TDD after switching. If the continuous D differences of the average conductance G, average susceptance B, average admittance modulus |Y|, and total demand distortion rate TDD do not exceed their respective lower thresholds, it is determined that the switched load has entered the steady state, and record the steady state entry time t2, then extract the average conductance G, average susceptance B, and average admittance modulus |Y| of the first cycle after entering the steady state as the characteristic values of the steady state after load switching, then perform S8; if any one of the differences in the difference sequences ΔG, ΔB, Δ|Y|, and ΔTDD exceeds its respective lower threshold, it is determined that the switched load has not entered the steady state, set the flag bit Flag = 1 and return to S1; S8: Use the genetic algorithm to perform load decomposition and identification on the steady state characteristic quantities obtained in S7 and the characteristic quantities in the off-line characteristic library, and return to S0.

2. The non-in-home load identification method based on average load and total demand distortion rate according to claim 1, characterized in that The calculation of the average conductance G, average susceptance B, average admittance modulus |Y|, and total demand distortion rate TDD in S4 is as follows: where T is the power grid power frequency period, u(t) is the supply voltage at the household electricity inlet of residential households, i(t) is the load current at the household electricity inlet of residential households, U is the effective voltage value, θ1 is the initial voltage phase angle, I h is the effective value of the h-th current harmonic, is the phase angle value of the h-th current harmonic, h is the harmonic order, and I0 is the rated current demand of residential household electricity.

3. The non-in-household load identification method based on average load and total demand distortion rate according to claim 1, characterized in that The upper limit thresholds λ of the difference sequences ΔG, ΔB, and Δ|Y| of the average conductance G, average susceptance B, and average admittance modulus |Y| for determining whether a load switching event occurs in S6 G+ , λ B+ and λ |Y|+ are respectively 10% of the minimum values of the average conductance G, average susceptance B, and average admittance modulus |Y| in the offline feature library.

4. The non-household load identification method based on average load and total demand distortion rate according to claim 1, characterized in that Lower threshold values λ of the difference sequences ΔG, ΔB, Δ|Y|, and ΔTDD of the average conductance G, average susceptance B, average admittance modulus |Y|, and total demand distortion rate TDD for determining whether to enter a steady state in S7 G- , λ B- , λ |Y|- and λ TDD- are 1%, 1%, 1%, and 2% of the minimum values of the average conductance G, average susceptance B, average admittance modulus |Y|, and total demand distortion rate TDD in the offline feature library, respectively.

5. The non-household load identification method based on average load and total demand distortion rate according to claim 1, wherein The parameters of the genetic algorithm in S8 include the population size s, crossover rate Pn, mutation rate Mr, and fitness function F. The expression of the fitness function F is: F min = (H G - Z G * X) + (H B - Z B * X) + (H |Y| - Z |Y| * X) (2) Where, H G is the steady-state average conductance G value, H B is the steady-state average susceptance B value, H |Y| is the steady-state average admittance modulus |Y|, Z G = [Z G1 , Z G2 … Z Gm is the single-electrical-appliance average conductance G feature library, Z B = [Z B1 , Z B2 … Z Bm is the single-electrical-appliance average susceptance B feature library, Z |Y| = [Z |Y|1 , Z |Y|2 … Z |Y|m is the single-electrical-appliance average admittance modulus |Y| feature library, m is the number of user loads, X = [x1, x2… x m T , x j = 0 indicates the off state of the electrical appliance, and x j = 1 indicates the on state of the electrical appliance, j = 1, 2… m.​

Images