Non-intrusive online rapid detection method for charging load of electric bicycles

Abstract

The present invention discloses a non-intrusive online rapid detection method for charging load of electric bicycles, including: performing state conversion removal, downsampling and SG filtering on the power data of 1Hz electric bicycles charged separately, calculating the differential signals of active / reactive power respectively, and constructing the charging load template of electric bicycles using the fitting curve of the differential signal in the constant voltage stage. Collect the active / reactive power data at the user's home, take the time window as 6 hours and the step length as 10 minutes, perform state conversion removal, downsampling and SG filtering in the window in turn, calculate the differential signals of active / reactive power respectively, match the continuous negative subsequences in the active power differential signal with the continuous positive subsequences in the reactive power differential signal, calculate the distance between the successfully matched subsequence and the template, and if it is less than the distance threshold, it is judged that there is an electric bicycle charging load at that place. The present invention has broad application prospects in the fields of electric bicycle illegal charging inspection and so on.

Description

Technical Field

[0001] The invention relates to the field of electric bicycle charging load monitoring, and in particular to a non-intrusive electric bicycle charging load online rapid detection method based on local features. Background Art

[0002] In the field of transportation, my country's new energy vehicle and electric bicycle industries have also entered a period of rapid development. Among them, the number of electric bicycles in society has exceeded 300 million. However, due to the lack of planning and management of electric bicycle charging sites and the weak safety awareness of residents, related fire accidents occur frequently, often causing huge casualties and property losses. For this reason, power, property and other departments often need to manually detect users' illegal charging behaviors on site, but there are problems such as low efficiency and low user cooperation. Non-intrusive load monitoring technology (NILM) does not need to invade the user's internal, but only needs to process and analyze the total load power consumption data to obtain detailed power consumption information of each user's electrical appliance, and can also analyze the user's power consumption behavior based on this. Therefore, the application of NILM technology to the efficient detection of illegal charging of electric bicycles has great practical feasibility. This efficient and convenient monitoring technology will also have broad application prospects in the fields of electric bicycle health status assessment, charging power query, energy efficiency analysis, etc.

[0003] At present, the application of non-intrusive load monitoring technology among residential users mainly focuses on some common household appliances, and there are few studies on the detection methods of electric bicycle charging load. And although the most advanced NILM method can decompose the loads of most household appliances, the charging load of electric bicycles is a continuously variable load, and the identification and decomposition of this type of load is still a difficult task. Existing methods for identifying continuously variable loads either require large training data sets or high sampling frequency data for transient feature extraction, which are not met by billing smart meters and also limit the large-scale promotion and application of these methods. On the other hand, the operation time of electric bicycle charging loads is relatively long, and it may overlap with other electrical loads, which brings great challenges to identification and decomposition. In addition, although there are a small number of unsupervised non-intrusive electric bicycle charging load detection methods, online detection cannot be achieved. Summary of the invention

[0004] Considering the shortcomings of the existing technology, in order to further realize the rapid discovery of the charging load of electric bicycles, the present invention combines non-intrusive load monitoring technology and proposes a non-intrusive online rapid detection method of the charging load of electric bicycles based on local features, aiming to meet the needs of timely and rapid discovery and positioning of electric bicycle charging in actual power usage scenarios. This invention can accurately and quickly realize online detection of electric bicycles, and has broad application prospects in the fields of electric bicycle illegal charging inspection and other fields.

[0005] In order to solve the above technical problems, the present invention proposes a non-intrusive online rapid detection method for electric bicycle charging load, which mainly includes: constructing an electric bicycle charging load template; judging whether there is a suspected electric bicycle charging load in the power load of the detected user; and finally detecting whether there is an electric bicycle charging load in the power load of the detected user. The specific steps are as follows:

[0006] Step 1: Build an electric bicycle charging load template. The steps are as follows:

[0007] 1-1) Collect active power and reactive power data of several electric bicycles charged individually at a sampling frequency of 1 Hz;

[0008] 1-2) Data preprocessing: Take a time window for the active power and reactive power data collected in step 1-1), and use the state transition removal algorithm to remove the load events in the total active power and total reactive power in the window; reduce the sampling frequency of the active power and reactive power data to 1 / 30 Hz, and use Savitzky-Golay (SG) filtering to reduce the noise in the power signal after frequency reduction;

[0009] 1-3) calculating the difference of the active power and reactive power data preprocessed in step 1-2) to obtain an active power differential signal and a reactive power differential signal;

[0010] 1-4) respectively fitting the active power differential signal and the reactive power differential signal in the constant voltage charging stage into a line segment, using the maximum point and slope of the fitted line segment as characteristic vector parameters, thereby establishing an electric bicycle charging load template including an active power differential signal template and a reactive power differential signal template;

[0011] Step 2: Determine whether there is a suspected electric bicycle charging load in the power load of the detected user:

[0012] 2-1) collecting active power data and reactive power data at the entrance of the detected user, and calculating the active power differential signal and reactive power differential signal at the entrance of the user according to the above steps 1-2) and 1-3);

[0013] 2-2) matching the consecutive negative subsequences in the active power differential signal at the user's entrance obtained in step 2-1) with the consecutive positive subsequences in the reactive power differential signal. If the match is successful, executing step 3, otherwise, repeating step 2;

[0014] Step 3: Calculate the distance L between the continuous subsequence successfully matched in step 2 and the electric bicycle separate charging load template constructed in step 1. If the distance L ≥ the preset distance threshold, return to step 2; otherwise, it is detected that the user's power load contains an electric bicycle charging load.

[0015] Furthermore, the non-intrusive online rapid detection method for charging load of an electric bicycle of the present invention comprises:

[0016] For step 1-1), the number of electric bicycles is 10.

[0017] For steps 1-2), the length of the time window is set to 6 hours and the step size is set to 10 minutes; the state transition removal algorithm is used to remove load events in the total active power and total reactive power in the window, which means that in the power data of the electric bicycle charged alone, the start-up event of the electric bicycle is removed; in the power data at the user entrance, the start-up, operation, and shutdown events of electrical loads other than electric bicycles are removed.

[0018] For step 2-2), the continuous positive subsequence is a differential signal sequence in which the reactive power differential signal is greater than 0.06 and more than 8 consecutive sampling points are in the same state; the continuous negative subsequence is a differential signal sequence in which the active power differential signal is less than -0.1 and more than 8 consecutive sampling points are in the same state; the successful match means that the continuous negative active power differential signal subsequence and the continuous positive reactive power differential signal subsequence overlap in time.

[0019] For step 3, the distance L refers to the average value of the sum of squares of the differences between the active power differential signal template in the electric bicycle charging load template and the successfully matched active power differential continuous subsequence and the sum of squares of the differences between the reactive power differential signal template in the electric bicycle charging load template and the successfully matched reactive power differential continuous subsequence.

[0020] Compared with the prior art, the present invention has the following beneficial effects:

[0021] The present invention applies non-intrusive load monitoring technology to the detection of electric bicycle charging load, extracts the local characteristics of electric bicycle charging load, and establishes a non-intrusive electric bicycle charging load online rapid detection method based on local characteristics, which can accurately and quickly realize the online detection of electric bicycles without intruding the user's internal. This method can meet the needs of timely and rapid detection and positioning of electric bicycle charging in actual power consumption scenarios, and has broad application prospects in the fields of electric bicycle illegal charging inspection and other fields. BRIEF DESCRIPTION OF THE DRAWINGS

[0022] Figure 1 It is a flow chart of the online rapid detection method of the present invention;

[0023] Figure 2(a) is a schematic diagram of active power filtering of the electric bicycle charging stage model;

[0024] Figure 2(b) is a schematic diagram of reactive power filtering of the electric bicycle charging stage model;

[0025] Figure 2(c) is a schematic diagram of the active power differential signal after model filtering during the charging phase of an electric bicycle;

[0026] Figure 2(d) is a schematic diagram of the reactive power differential signal after model filtering during the charging phase of an electric bicycle;

[0027] FIG3( a ) is a schematic diagram of the detected load event detection result and SG filtering;

[0028] FIG3( b ) is a schematic diagram of detected state transition removal and SG filtering;

[0029] Figure 4(a) is a graph of the charging load slope-active power difference of 10 electric bicycles;

[0030] Figure 4(b) is a graph of the slope-reactive power difference of the charging load of 10 electric bicycles;

[0031] FIG5( a ) is a schematic diagram of the active power difference signal and the continuous negative subsequence after SG filtering;

[0032] FIG5( b ) is a schematic diagram of the reactive power differential signal and the continuous positive subsequence after SG filtering;

[0033] Figure 6(a) shows the real value of the user power and EBCL power of the online detection of the electric bicycle charging load of user No. 1;

[0034] FIG6( b ) is a schematic diagram of the active power difference signal and the continuous negative subsequence after SG filtering of an online detection of the charging load of an electric bicycle of user No. 1;

[0035] FIG6( c ) is a schematic diagram of the reactive power differential signal and the continuous positive subsequence after SG filtering during online detection of an electric bicycle charging load of user No. 1. DETAILED DESCRIPTION

[0036] The design idea of ​​the non-invasive online rapid detection method for electric bicycle charging load proposed in the present invention is that, during the research process, through the collected active power and reactive power data of electric bicycles charged alone, it was found that electric bicycles have a constant voltage charging stage that is different from other electrical loads. This stage exhibits a load characteristic of a gentle slope of active power decrease and a gentle slope of reactive power increase. Therefore, the present invention adopts the local characteristics of the constant voltage stage, that is, the load characteristic of a gradually increasing difference in the amplitude of active power and reactive power to detect the charging load of the electric bicycle.

[0037] The present invention will be further described below in conjunction with the accompanying drawings and specific embodiments, but the following embodiments are by no means intended to limit the present invention in any way.

[0038] The method for realizing the non-invasive online rapid detection of electric bicycle charging load based on local features of the present invention mainly includes constructing an electric bicycle charging load template using the local features of the constant voltage stage, and then collecting the active power data and reactive power data of the detected user's home, and judging whether the detected user's power load is suspected to contain an electric bicycle charging load based on the calculated active power differential signal and reactive power differential signal at the user's home; finally, by comparing and calculating with the electric bicycle charging load template, it is detected whether the detected user's power load contains an electric bicycle charging load. Figure 1 As shown, the specific steps are as follows:

[0039] Step 1: Build an electric bicycle charging load template. The steps are as follows:

[0040] 1-1) With a sampling frequency of 1 Hz, the active power and reactive power data of 10 electric bicycles are collected simultaneously;

[0041] 1-2) Data preprocessing: Take a time window for the data of the electric bicycle charging collected in step 1-1), and use a state transition removal algorithm to remove load events in the total active power and total reactive power in the window.

[0042] This method adopts an adaptive two-stage event detection method proposed by [Luan W, Liu Z, Liu B, et al. An Adaptive Two-stage Load Event Detection Method for Nonintrusive Load Monitoring.], which adaptively adjusts the event detection threshold according to the different degrees of electrical load fluctuation in the total power consumption data, and adopts an improved edge detection method and a window-based detection method combining moving average and sliding T test for quasi-step events and long transient events with different waveform characteristics, respectively, to achieve accurate detection and positioning of electrical load events in the total power consumption data. Then, the state transition removal algorithm is used to remove the detected events from the total power consumption data.

[0043] The load events in reactive power are determined based on the events in active power, and then the state transition removal algorithm is used to remove the detected events. The purpose of state transition removal is to restore the gentle slope trend characteristics of the electric bicycle charging load (EBCL) that is separated due to the state transition of other electrical loads. When a load event is detected in the power signal, it is removed from the total data using formula (1):

[0044]

[0045] Where τ represents the time index of the analyzed signal, Z represents the active power and reactive power signals, ΔZ represents the active power and reactive power values ​​of the detected event, and L t Indicates the active power signal and reactive power signal after the state transition is removed.

[0046] Downsampling and filtering are performed to reduce the sampling frequency of the electric bicycle individual charging data to 1 / 30 Hz, and Savitzky-Golay (SG) filtering is used to reduce the noise in the power signal after the frequency reduction.

[0047] In the present invention, in order to retain the gentle slope characteristics of the electric bicycle and reduce the interference of other electrical loads at the same time, the mean is calculated once every 30 sampling points as a new sampling point, that is, the sampling frequency is reduced to 1 / 30Hz. Since the signal after the state transition is removed still contains fluctuations in the operation of the electrical load, in order to further reduce this fluctuation and smooth the gentle slope characteristics corresponding to the constant voltage charging stage of the electric bicycle, the present invention uses Savitzky-Golay filtering (SG filtering) to smooth the signal. SG filtering is a filtering method based on local least squares polynomial fitting using a sliding window. This filtering method retains the peak value and width of the original signal while eliminating noise of different frequencies, and is widely used in signal denoising with non-Gaussian noise. Compared with filtering methods such as mean filtering and Kalman filtering, SG filtering has better signal shape retention and denoising performance without losing resolution.

[0048] Given a local symmetric data window n=[l -m ,l -m+1 ,...,l0,...,l m-1 ,l m ],l i Represents the filtered signal L t The active power and reactive power data of a sampling point in , the signal after SG filtering is:

[0049]

[0050] Among them, p<2m, represents the order of the least squares polynomial, a k represents the coefficients of the polynomial, l' n It is the active power and reactive power signal corresponding to the data window n after SG filtering.

[0051] In the present invention, the length of the time window is set to 6 hours and the step size is set to 10 minutes. The state transition removal algorithm is used to remove the load events in the total active power and total reactive power in the window. For the power data of the electric bicycle charged separately, the start event of the electric bicycle is removed.

[0052] 1-3) Calculate the power data difference, calculate the difference of the electric bicycle individual charging data preprocessed in step 1-2) to obtain the active power difference signal and reactive power difference signal of the electric bicycle individual charging.

[0053] 1-4) Use the pre-processed electric bicycle individual charging power data to construct an electric bicycle charging load template.

[0054] SG filtering makes the transition process from the constant current charging stage to the constant voltage charging stage smoother, and the small fluctuations in the constant voltage charging stage are also smoothed. The differential signal is calculated for the processed active power data and reactive power data, as shown in Figure 2(a), Figure 2(b), Figure 2(c) and Figure 2(d). The power differential signal in the constant voltage charging stage can be fitted into a line segment, and the maximum point and slope of the fitted line segment are used as the characteristic vector parameters of the template. The constant voltage segments of the active power and reactive power data are fitted respectively to establish the active power differential signal template of the electric bicycle charging load. and reactive power differential signal template Figure 2(a) shows a schematic diagram of active power filtering of the electric bicycle charging stage model; Figure 2(b) shows a schematic diagram of reactive power filtering of the electric bicycle charging stage model; Figure 2(c) shows the active power differential signal after filtering of the electric bicycle charging stage model; Figure 2(d) shows the reactive power differential signal after filtering of the electric bicycle charging stage model.

[0055] Step 2: Determine whether there is a suspected electric bicycle charging load in the power load of the detected user.

[0056] 2-1) Collect active power data and reactive power data at the entrance of the detected user, and according to the above steps 1-2) and 1-3), the data substituted in the process is the power data collected from the user's entrance. After downsampling and filtering, the power data at the user's entrance, the opening, running, and closing events of electrical loads other than electric bicycles are removed. The active power differential signal and reactive power differential signal at the user's entrance are obtained by calculating the power data differential.

[0057] 2-2) Calculating power data difference: Matching the continuous negative subsequences in the active power difference signal at the user's entrance obtained by step 2-1) with the continuous positive subsequences in the reactive power difference signal.

[0058] The present invention sets the active power differential less than -0.1 as negative, the reactive power differential greater than 0.06 as positive, and stipulates that more than 8 consecutive sampling points in the same state can be divided into continuous subsequences. The active power ramp down and the reactive power ramp up in the constant voltage charging stage of the electric bicycle are synchronized in time. Therefore, it is necessary to match the continuous negative active power differential signal subsequence and the continuous positive reactive power differential signal subsequence, that is, the continuous negative active power differential signal subsequence and the continuous positive reactive power differential signal subsequence overlap in time and are considered to be matched. If the match is successful, step 3 is executed, otherwise, step 2 is repeated.

[0059] Step 3: Matching with the electric bicycle charging load template, by calculating the distance between the matched continuous subsequence and the electric bicycle charging load template, if the distance is less than a preset distance threshold, it is considered to match the template.

[0060] For the subsequences of active power differential signals that are continuously negative and reactive power differential signals that are synchronized in time (i.e., successfully matched), the distances from the electric bicycle charging load template are calculated respectively. The distances refer to the average of the sum of squares of the differences between the active power differential signal template in the electric bicycle charging load template and the successfully matched active power differential continuous subsequences and the average of the sum of squares of the differences between the reactive power differential signal template in the electric bicycle charging load template and the successfully matched reactive power differential continuous subsequences, as shown in formula (3):

[0061]

[0062] Among them, x is the template signal of the electric bicycle charging load, y is the successfully matched power difference continuous subsequence, x1 is the maximum point of the template, y1 is the maximum point of the continuous subsequence, n is the number of data points from the maximum point in the continuous subsequence of the differential signal to the end point of the subsequence. It should be noted that in order to avoid the situation where the maximum point in the continuous subsequence is at the last bit and it is close to the highest point of the template, it is necessary to set n>10, ΔS represents the active signal subsequence distance and the power signal subsequence distance. The average of the active signal subsequence distance and the reactive signal subsequence distance is taken as the distance between the matching subsequence and the template. If it is less than the distance threshold set by the heuristic method, it is considered that the subsequence matches the template and there is an electric bicycle constant voltage charging stage, that is, the electric bicycle charging load is detected in the user's power load; if it is greater than or equal to the threshold, it is considered that the subsequence does not match the template and there is no electric bicycle constant voltage charging stage.

[0063] Research material example 1:

[0064] Figure 3(a) and Figure 3(b) respectively show the load event detection results and the results after state transition removal for the user entrance data in a certain period of time, and also show the comparison images before and after SG filtering. It can be seen that the load event detection algorithm can accurately detect events in the total electricity consumption data.

[0065] The active power data and reactive power data of 10 electric bicycles charged separately are used to construct an electric bicycle charging load template. The 10 electric bicycle charging load model parameters shown in the table above are obtained. The model diagrams of the 10 electric bicycle charging loads are drawn according to the data in Table 1. Figure 4(a) shows the slope-active power difference diagram of the 10 electric bicycle charging loads, and Figure 4(b) shows the slope-reactive power difference diagram of the 10 electric bicycle charging loads. It can be seen that the model of the No. 5 lead-acid battery deviates significantly from the other 9 models, so it is eliminated, and the average of the remaining 9 models is taken to obtain the EBCL template, that is, Low: -1.2836; Slope: 0.0100; Highest point: 0.5548; Slope: -0.0054.

[0066] Table 1 Parameters of 10 electric bicycle charging load models

[0067]

[0068] The differential signal is calculated for the active power and reactive power data at the entrance, and the continuous negative subsequences in the active power differential signal and the continuous positive subsequences in the reactive power differential signal are matched. As shown in Figure 5(a) and Figure 5(b), it can be seen that the two subsequences near 00:00 are matched. Finally, the distance threshold is set to 0.25. Figures 6(a), 6(b) and 6(c) show the online detection results of the electric bicycle charging load of user No. 1, where Figure 6(a) is the user power and the true value of the EBCL power of the online detection of the electric bicycle charging load of user No. 1; Figure 6(b) is the active power differential signal and the continuous negative subsequence after SG filtering of the online detection of the electric bicycle charging load of user No. 1; Figure 6(c) is the reactive power differential signal and the continuous positive subsequence after SG filtering of the online detection of the electric bicycle charging load of user No. 1.

[0069] Research material example 2:

[0070] At the same time, the active power and reactive power data at the entrance of 7 users for a total of 411 days were used to conduct online detection of the charging load of electric bicycles. Table 2 shows the experimental results within the 7 users. It can be seen that there is almost no missed detection of charging load. At the same time, online rapid detection is basically achieved 20 minutes to 45 minutes after the start of constant voltage charging.

[0071] Table 2 Online rapid detection results of electric bicycle charging load

[0072]

[0073] According to the above implementation, the present invention can quickly detect the charging load of electric bicycles without intruding into the user's internal environment according to the summarized charging load template of electric bicycles, with high accuracy. This method has great practical feasibility in the application of efficient detection of illegal charging of electric bicycles, and can meet the needs of rapid discovery and positioning of illegal charging behavior of electric bicycles in actual power usage scenarios, and has broad application prospects in the fields of illegal charging inspection of electric bicycles.

[0074] Although the present invention has been described above in conjunction with the accompanying drawings, the present invention is not limited to the above-mentioned specific embodiments, which are merely illustrative rather than restrictive. Under the guidance of the present invention, ordinary technicians in this field can make many modifications without departing from the purpose of the present invention, all of which are within the protection of the present invention.

Claims

1. A non-intrusive online rapid detection method for charging load of an electric bicycle, characterized in that: The following steps are involved: Step 1: Build an electric bicycle charging load template. The steps are as follows: 1-1) Collect active power and reactive power data of several electric bicycles charged individually at a sampling frequency of 1 Hz; 1-2) Data preprocessing: Take a time window for the active power and reactive power data collected in step 1-1), and use the state transition removal algorithm to remove the load events in the total active power and total reactive power in the window; reduce the sampling frequency of the active power and reactive power data to 1 / 30 Hz, and use Savitzky-Golay (SG) filtering to reduce the noise in the power signal after frequency reduction; 1-3) calculating the difference of the active power and reactive power data preprocessed in step 1-2) to obtain an active power differential signal and a reactive power differential signal; 1-4) respectively fitting the active power differential signal and the reactive power differential signal in the constant voltage charging stage into a line segment, using the maximum point and slope of the fitted line segment as characteristic vector parameters, thereby establishing an electric bicycle charging load template including an active power differential signal template and a reactive power differential signal template; Step 2: Determine whether there is a suspected electric bicycle charging load in the power load of the detected user: 2-1) collecting active power data and reactive power data at the entrance of the detected user, and calculating the active power differential signal and reactive power differential signal at the entrance of the user according to the above steps 1-2) and 1-3); 2-2) matching the consecutive negative subsequences in the active power differential signal at the user's entrance obtained in step 2-1) with the consecutive positive subsequences in the reactive power differential signal. If the match is successful, executing step 3, otherwise, repeating step 2; Step 3: Calculate the distance L between the continuous subsequence successfully matched in step 2 and the electric bicycle separate charging load template constructed in step 1. If the distance L ≥ the preset distance threshold, return to step 2; otherwise, it is detected that the user's power load contains an electric bicycle charging load.

2. The non-intrusive online rapid detection method for charging load of an electric bicycle according to claim 1 is characterized in that: For step 1-1), the number of electric bicycles is 10.

3. The non-intrusive online rapid detection method for charging load of an electric bicycle according to claim 1 is characterized in that: For step 1-2), the length of the time window is set to 6 hours and the step length is set to 10 minutes; The state transition removal algorithm is used to remove load events from the total active power and total reactive power in the window, which means that in the power data of the electric bicycle charged alone, the start-up event of the electric bicycle is removed; in the power data at the user entrance, the start-up, operation, and shutdown events of electrical loads other than the electric bicycle are removed.

4. The non-intrusive online rapid detection method for charging load of an electric bicycle according to claim 1 is characterized in that: For step 2-2), the continuous positive subsequence is a differential signal sequence in which the reactive power differential signal is greater than 0.06 and more than 8 consecutive sampling points are in the same state; the continuous negative subsequence is a differential signal sequence in which the active power differential signal is less than -0.1 and more than 8 consecutive sampling points are in the same state; the successful match means that the continuous negative active power differential signal subsequence and the continuous positive reactive power differential signal subsequence overlap in time.

5. The non-intrusive online rapid detection method for charging load of an electric bicycle according to claim 1 is characterized in that: For step 3, the distance L refers to the average value of the sum of squares of the differences between the active power differential signal template in the electric bicycle charging load template and the successfully matched active power differential continuous subsequence and the sum of squares of the differences between the reactive power differential signal template in the electric bicycle charging load template and the successfully matched reactive power differential continuous subsequence.

Images