Event detection based non-intrusive power load component recognition method, system and media

Abstract

The application discloses a non-intrusive power load component identification method and system based on event detection, and a medium. The method comprises the following steps: acquiring active power and reactive power of a power load during normal operation; dividing power load operation data into data segments with the same length as sub-windows by using a sliding window; monitoring active power of the sub-windows in sequence to capture a power load operation event. The time of the event and the window where the event occurs are recorded, and the window where the event occurs is divided into meta-windows with the same length by using a smaller time scale sliding window. The maximum value, the minimum value, the average value and the range of active power and reactive power in the meta-windows are calculated in sequence to find the most stable time of the event and the time before and after the event. The active power variation after the load is put into operation, the impact coefficient and the reactive power variation are calculated. The load type is determined according to the impact coefficient and the reactive power variation after the load is put into operation, so that the load component identification of a motor, a static load and a power electronic interface load is realized.

Description

TECHNICAL FIELD

[0001] The present application belongs to the technical field of power system simulation analysis, especially the power system load model, and particularly relates to a non-intrusive power load component identification method, system and medium based on event detection. BACKGROUND

[0002] The power load is an important part of the power system. The power load model, referred to as the load model, generally refers to all power distribution networks and electrical equipment in a 220kV substation power supply area. The accuracy of power system simulation analysis is highly dependent on the accurate establishment of the load model, and accurately grasping the components of the power load is crucial to the load model.

[0003] In the load model commonly used in current power system simulation analysis, static load (including constant impedance, constant current, constant power components) and motor load are mainly used to represent electrical equipment. In recent years, more and more electrical equipment has been incorporated into the power grid through power electronic technology (switching power supply, variable frequency power supply, etc.), but the existing static load model and motor model cannot accurately describe the dynamic characteristics of the power electronic interface load, so it is believed that in the new power load model, static load (constant impedance load), power electronic interface load (including constant current and constant power components in the original static load), and motor load are the three most important components.

[0004] The composition of the power load has time-varying characteristics, which is a key factor affecting the accuracy of the load model. When establishing the load model, the typical model parameters of various load components can usually remain unchanged, and what needs to be changed is the power ratio of the load components. In order to adapt to the time-varying nature of the load components in the load modeling work, the power industry is currently building a load modeling system using an online statistical synthesis method, but this system can only obtain the power data of the user port meter and cannot obtain the load components inside the user. There are already many non-intrusive load identification methods, but their purpose is to identify the name of the equipment in operation, and their algorithms mostly rely on artificial neural networks, and high recognition rates are only limited to trained electrical equipment. Therefore, the existing non-intrusive load identification method is difficult to develop a low-cost identification device, and cannot meet the urgent need of power load modeling for load component identification. SUMMARY

[0005] In view of the problems in the prior art, according to the first aspect of the purpose of the present application, a non-intrusive power load component identification method based on event detection is proposed, which can identify the type of power load put into operation, i.e., which one of the three load components of motor, static load and power electronic interface load.

[0006] The event detection-based non-intrusive power load component identification method according to the above purpose of the present application comprises the following steps:

[0007] Step 1, obtaining data of the power load in normal operation, including active power P(t) and reactive power Q(t);

[0008] Step 2, setting a first sliding window with a first fixed length, and dividing the power load operation data into data segments with the same length according to the first sliding window, each data segment being a sub-window;

[0009] Step 3, using an event detection algorithm to monitor the mutation of the active power of the load for each sub-window in turn, thereby capturing the power load operation event;

[0010] Step 4, recording the time t at which the power load operation event occurs and the window in which the power load operation event occurs;

[0011] Step 5, setting a second sliding window with a second fixed length, and dividing the window in which the power load operation event occurs into meta-windows with the same length according to the second sliding window;

[0012] Step 6, calculating the active power parameters and the reactive power parameters in each meta-window in turn, the active power parameters including: the maximum value P max (i), the minimum value P min (i), the average value P mean (i) and the range P stab (i) of the active power; and the reactive power parameters including: the maximum value Q max (i), the minimum value Q min (i), the average value Q mean (i) and the range Q stab (i) of the reactive power;

[0013] Step 7, based on the active power parameters and the reactive power parameters in the meta-window, finding the time at which the load operation event is put into operation and the time at which the power is most stable before and after the time at which the load operation event is put into operation;

[0014] Step 8, calculating the active variation, the impact coefficient and the reactive variation after the load is put into operation; and

[0015] Step 9, determining the type of the load component put into operation according to the impact coefficient and the reactive variation after the load is put into operation.

[0016] As a preferred embodiment, in the step 9, determining the type of the load component put into operation according to the impact coefficient and the reactive variation after the load is put into operation, comprising:

[0017] if the impact coefficient is greater than 10, determining that the type of the load component put into operation is a motor;

[0018] If the impact factor is greater than 2 and less than 10, it is determined that the input load component type is a power electronic interface load;

[0019] If the impact factor is less than 2, further combined with the reactive power change amount to make further judgment, wherein:

[0020] If the reactive power change amount is less than -1, the input load component type is a power electronic interface load, otherwise it is determined to be a static load.

[0021] According to the second aspect of the object of the present application, a computer system is also proposed, comprising:

[0022] One or more processors;

[0023] Memory, storing instructions that can be operated;

[0024] Wherein, the instructions when executed by one or more processors make the aforementioned one or more processors perform operations, the operations comprising executing the process of the aforementioned event detection based non-intrusive power load component identification method.

[0025] According to the third aspect of the object of the present application, a computer readable storage medium is also proposed for storing one or more programs, the one or more programs comprising instructions or instruction sets that can be executed by one or more processors.

[0026] Wherein, the instructions or instruction sets when executed by one or more processors, execute the process of the aforementioned event detection based non-intrusive power load component identification method.

[0027] When different kinds of power loads are put into operation, there will be different starting characteristics: when the motor and power electronic interface load are put into operation, the active power will have an impact and then quickly recover to stability, the impact phenomenon of the former is more obvious than the latter, the impact factor can reach more than ten times, while the active impact of the static load when put into operation is very small, which can be approximately ignored. In addition, the input of power electronic interface load will make the reactive power decrease, while the input of static load has little effect on the reactive power, which can be approximately ignored. Therefore, the present application proposes an event detection based non-intrusive power load component identification method, which identifies which one of the three load components of motor, static load (constant impedance load) and power electronic interface load based on the difference between the impact of active power and the change of reactive power when different kinds of power loads are put into operation.

[0028] Compared with the prior art, the event detection based non-intrusive power load component identification method of the present application has the following obvious advantages:

[0029] (1) Different from the conventional non-intrusive load identification method for identifying specific electrical equipment, the non-intrusive power load component identification method based on event detection of the application only identifies the type and power of the input load equipment, only needs to obtain the active power and reactive power when the power load is stably running, has lower requirements for the collection equipment, and the load component identification process does not involve complex algorithms, is easy to realize at low cost and has high accuracy.

[0030] (2) The non-intrusive power load component identification method of the application can also determine which load component the device that exits operation belongs to, and the power of the exiting device. By using the method of the application at the time period when the user power is the lowest, the relatively accurate user load component composition in different time periods can be obtained through continuous operation and calculation. By accumulating the load component identification results of different users at the same time, the load component power ratio required for load modeling can be obtained, thereby providing support for establishing an accurate power load model. BRIEF DESCRIPTION OF DRAWINGS

[0031] Figure 1 is a flowchart of the non-intrusive power load component identification method based on event detection according to an embodiment of the application.

[0032] Figure 2 is an active power curve schematic diagram of an example of the non-intrusive power load component identification method based on event detection according to an embodiment of the application.

[0033] Figure 3 is a reactive power curve schematic diagram of an example of the non-intrusive power load component identification method based on event detection according to an embodiment of the application.

[0034] Figure 4 is an event window and alarm value schematic diagram of an example of the non-intrusive power load component identification method based on event detection according to an embodiment of the application. DETAILED DESCRIPTION

[0035] The technical solutions in the embodiments of the application will be described clearly and completely below with reference to the drawings in the examples of the application. Obviously, the described embodiments are only part of the embodiments of the application, rather than all the embodiments. Based on the embodiments in the application, all other embodiments obtained by those skilled in the art without creative labor fall within the scope of protection of the application.

[0036] Various aspects of the invention are described with reference to the accompanying drawings, which illustrate numerous illustrative embodiments. The disclosed embodiments are not necessarily intended to encompass all aspects of the invention. It should be understood that the various concepts and embodiments described above, as well as those described below in more detail, can be implemented in any of many ways, as the concepts and embodiments disclosed herein are not limited to any particular implementation. Furthermore, some aspects of the invention disclosed may be used alone or in any suitable combination with other aspects of the invention disclosed.

[0037] {Example 1}

[0038] The non-intrusive power load component identification method based on event detection proposed in this embodiment aims to identify the load component—whether it belongs to an electric motor, a static load (constant impedance load), or a power electronic interface load—by utilizing the differences between the active power surge and reactive power changes when different types of power loads are connected. For power load modeling needs, this embodiment only identifies the type and power of the connected load equipment. It only requires acquiring the active and reactive power during stable operation of the power load, has low requirements for the acquisition equipment, and the identification process does not involve complex algorithms, making it easy to implement at low cost with high accuracy.

[0039] like Figure 1 As shown in the example, the process of a non-intrusive power load identification method based on event detection includes the following steps:

[0040] Step 1: Obtain data on the power load during normal operation, including active power P(t) and reactive power Q(t);

[0041] Step 2: Set a first sliding window of a fixed length, and divide the power load operation data into data segments of equal duration according to the first sliding window, with each data segment being a sub-window;

[0042] Step 3: Use the event detection algorithm to monitor sudden changes in the active power of the load for each sub-window in turn, thereby capturing the power load commissioning event;

[0043] Step 4: Record the time t of the power load commissioning event and the window in which it occurs;

[0044] Step 5: Set a second fixed-length sliding window, and divide the window where the power load is put into operation into a meta-window of the same duration according to the second sliding window;

[0045] Step 6: Calculate the active power parameters and reactive power parameters of the load within each element window. The active power parameters include: the maximum value P of the active power.max (i), minimum value P min (i), average value P mean (i) and range P stab (i); the reactive power parameter comprises: a maximum value Q of the reactive power max (i), minimum value Q min (i), average value Q mean (i) and range Q stab (i);

[0046] Step 7, based on the active power parameter and the reactive power parameter of the load in the meta-window, finding the load input event input time and the time when the power is most stable before and after the input time;

[0047] Step 8, calculating the active power change, the impact coefficient and the reactive power change after the load is input; and

[0048] Step 9, according to the impact coefficient and the reactive power change after the load is input, determining the load composition type, that is, one of a static load, a power electronic interface load and a motor load, wherein the static load is a constant impedance load.

[0049] As an optional example, the second fixed length is less than the first fixed length. That is, in step 5, a sliding window of a smaller time scale is used to slide the window where the power load input operation event is located, and the window is divided into meta-windows of the same length.

[0050] As an optional example, in step 3, the CUSUM event detection algorithm is used to monitor the mutation of the active power of the load, so as to capture the power load input operation event, comprising:

[0051] For each sub-window range, the difference between each data and the average value is calculated in turn and accumulated, and when the difference between the cumulative sum at a certain time and the minimum cumulative sum exceeds the set threshold value, an alarm is triggered, indicating that the power load input is detected.

[0052] As an optional example, in step 6, during the calculation of the active power parameter and the reactive power parameter of the load, the range is determined based on the maximum value and the minimum value of the power, that is:

[0053] P stab (i) = P max (i) - P min (i);

[0054] Q stab (i) = Q max (i) - Q min (i).

[0055] As an optional example, in step 7, the active power extreme difference P stab (i) the meta-window corresponding to the maximum value, and the time point of the meta-window is the load input time point p0;

[0056] The active power extreme difference P stab (i) the meta-window corresponding to the minimum value, and the time point of the meta-window is the time point p1, p2 at which the active power is most stable;

[0057] The reactive power extreme difference Q stab (i) the meta-window corresponding to the minimum value, and the time point of the meta-window is the time point p3, p4 at which the reactive power is most stable.

[0058] As an optional example, in step 8, the active variation, the impact factor and the reactive variation after the load is input are calculated, including:

[0059] The active variation ΔP = P mean (p2) - P mean (p1);

[0060] The impact factor IMP = (P max (p0) - P mean (p1)) / ΔP;

[0061] The reactive variation ΔQ = Q mean (p4) - Q mean (p3).

[0062] As an optional example, in step 9, the load component type is determined according to the impact factor and the reactive variation after the load is input, including:

[0063] If the impact factor is greater than 10, it is determined that the load component type is a motor;

[0064] If the impact factor is greater than 2 and less than 10, it is determined that the load component type is a power electronic interface load;

[0065] If the impact factor is less than 2, further combined with the reactive variation, further determination is made, wherein:

[0066] If the reactive variation is less than -1, the load component type is a power electronic interface load, otherwise it is determined to be a static load.

[0067] {Example 2}

[0068] In this embodiment, the load input time point p0 is determined by combining the active power and the reactive power, and the load component type is determined according to the impact factor and the reactive variation after the load is input. Figures 2 to 4 As shown in FIG. 2, the load input time point p0 is determined more specifically by combining the active power and the reactive power, and the load component type is determined according to the impact factor and the reactive variation after the load is input. Figure 1The implementation process of the non-intrusive power load component recognition method based on event detection of the above embodiments is further described and explained.

[0069] Step 1, obtaining data of the power load in normal operation, including active power P(t) and reactive power Q(t).

[0070] In this embodiment, the experimental data come from a test system built in a simulation laboratory. The data of the power load in operation are collected by a fault recorder, and the sampling frequency is 1 kHz.

[0071] In this embodiment, one section is intercepted, with a duration of 1 min, containing the opening process of the energy-saving lamp. The curve of the active power P(t) is shown in Figure 2 , and the curve of the reactive power Q(t) is shown in Figure 3 .

[0072] Step 2, setting a sliding window with a fixed length to divide the power load operation data into data sections with the same length, and each data section is a sub-window.

[0073] In this embodiment, the window size is set to 10 s, and divided into 6 sub-windows.

[0074] Step 3, using the event detection algorithm (CUSUM) to monitor the mutation of the active power of the load for each sub-window in turn, so as to capture the power load operation event.

[0075] In this embodiment, CUSUM is used to calculate the difference between each data and the average value and accumulate it. When the difference between the cumulative sum at a certain time and the minimum cumulative sum exceeds the set threshold value, an alarm is triggered, which is the detection of the load input.

[0076] As shown in Figure 4 , the window in which the load input event is detected and the alarm value are exemplarily represented.

[0077] Step 4, recording the time t at which the power load operation event occurs and the window in which it occurs.

[0078] Combined with Figure 4 , according to the time at which the load input event occurs and the window in which it occurs, it is recorded that the load input is detected near t=26.895 s.

[0079] Step 5, setting a sliding window with a smaller time scale to divide the window in which the power load operation event occurs into meta-windows with the same length.

[0080] In this embodiment, the size of the second sliding window is set to 1 s, and divided into 10 meta-windows.

[0081] Step 6, the maximum value P of the active power of the load in each meta-window is calculated in turn max (i), the minimum value P min (i), the average value P mean (i) and the range P stab (i) and the maximum value Q of the reactive power max (i), the minimum value Q min (i), the average value Q mean (i) and the range Q stab (i).

[0082] Further, the range is equal to the maximum value minus the minimum value, i.e. P stab (i) = P max (i) - P min (i), Q stab (i) = Q max (i) - Q min (i).

[0083] P max (i) = [54.977, 57.146, 57.792, 73.590, 73.512, 73.887, 75.245, 76.471, 77.581, 78.845].

[0084] P mean (i) = [53.315, 55.083, 55.683, 58.676, 69.703, 70.484, 73.146, 72.647, 74.447, 76.189].

[0085] P min (i) = [51.699, 52.794, 53.778, 55.929, 65.909, 67.413, 69.292, 69.536, 71.435, 71.992].

[0086] P stab (i) = [3.277, 4.352, 4.014, 17.661, 7.602, 6.474, 5.953, 6.935, 6.146, 6.852].

[0087] Q max (i) = [2.090, 2.621, 1.759, 2.590, -0.381, 0.515, -0.999, 0.603, 0.725, 0.128].

[0088] Q mean(i) = [0.015, 0.559, 0.183, 0.596, -4.209, -2.387, -3.912, -3.167, -2.164, -3.412].

[0089] Q min (i) = [-2.201, -1.333, -1.549, -5.097, -6.657, -5.137, -6.719, -6.528, -6.198, -6.041].

[0090] Q stab (i) = [4.291, 3.954, 3.308, 7.687, 6.276, 5.651, 5.720, 7.130, 6.923, 6.169].

[0091] Step 7, find the load input event occurrence time and the power most stable time before and after it.

[0092] In this embodiment, the time when the active power range difference maximum element window is located is the load input time p0, and the active range minimum element window before and after it is found, and the time when it is located is the active power most stable time p1, p2. Similarly, the reactive power most stable time p3, p4 before and after the load input is obtained.

[0093] In this example, combined with the above calculation results, it is determined that p0=4, p1=1, p2=7, p3=3, and p4=6.

[0094] Step 8, calculate the active power change, impact coefficient and reactive power change after the load input.

[0095] In the embodiment, the calculation process is:

[0096] Active power change ΔP=P mean (p2)-P mean (p1);

[0097] Impact coefficient IMP=(P max (p0)-P mean (p1)) / △P;

[0098] Reactive power change ΔQ=Q mean (p4)-Q mean (p3).

[0099] The active power change, impact coefficient and reactive power change calculated thereby are as follows:

[0100] △P=19.831, IMP=1.022, and △Q=-2.570.

[0101] Step 9, according to the impact coefficient after the load input and the reactive power variation, determine the type of the input load.

[0102] In this embodiment, the determination of the input load component type is based on the following:

[0103] If the impact coefficient is greater than 10, it is determined that a motor is input, otherwise if the impact coefficient is greater than 2, it is determined that a power electronic interface load is input, and if the impact coefficient is less than 2, further determination is needed through the reactive power variation, and the reactive power variation less than -1 is also determined as a power electronic interface load, otherwise it is a static load.

[0104] In combination with the above active power variation, impact coefficient and reactive power variation calculation results, according to the above determination, it is identified that a power electronic interface load is input, and further calculation is performed to obtain the actual running power of 452.147W, thereby realizing the component identification of the power load.

[0105] {Example 3}

[0106] In combination with the implementation of the above embodiment of the event detection based non-intrusive power load component identification method, according to the present disclosure, a computer system is further provided, which comprises:

[0107] one or more processors;

[0108] a memory for storing executable instructions.

[0109] Wherein, the instructions, when executed by one or more processors, cause the aforementioned one or more processors to perform operations, and the operations comprise performing the processes of the event detection based non-intrusive power load component identification method of the aforementioned embodiments.

[0110] {Example 4}

[0111] In combination with the implementation of the above embodiment of the event detection based non-intrusive power load component identification method, according to the present disclosure, a computer readable storage medium is further provided for storing one or more programs.

[0112] Wherein, the one or more programs comprise instructions or instruction sets that can be executed by one or more processors, and the instructions or instruction sets, when executed by one or more processors, perform the processes of the event detection based non-intrusive power load component identification method of the aforementioned embodiments.

[0113] Although the present application has been disclosed with reference to the preferred embodiments above, it is not intended to limit the present application. Those skilled in the art, without departing from the spirit and scope of the present application, can make various modifications and improvements. Therefore, the protection scope of the present application shall be subject to the scope defined by the claims.

Claims

1. A non-intrusive power load component identification method based on event detection, characterized in that, Includes the following steps: Step 1: Obtain data on the power load during normal operation, including active power P(t) and reactive power Q(t); Step 2: Set a first sliding window of a fixed length, and divide the power load operation data into data segments of equal duration according to the first sliding window, with each data segment being a sub-window; Step 3: Use the event detection algorithm to monitor sudden changes in the active power of the load for each sub-window in turn, thereby capturing the power load commissioning event; Step 4: Record the time t of the power load commissioning event and the window in which it occurs; Step 5: Set a second fixed-length sliding window, and divide the window where the power load is put into operation into a meta-window of the same duration according to the second sliding window; Step 6: Calculate the active power parameters and reactive power parameters of the load within each element window. The active power parameters include: the maximum value P of the active power. max (i) Minimum value P min (i) Average value P mean (i) and range P stab (i); The reactive power parameters include: the maximum reactive power Q max (i) Minimum value Q min (i) Average value Q mean (i) and range Q stab (i); Step 7: Based on the active power parameters and reactive power parameters of the load within the meta-window, find the load connection time and the time when the power is most stable before and after the connection time. Step 8: Calculate the change in active power, the impact factor, and the change in reactive power after the load is applied; and Step 9: Determine the type of load component based on the impact coefficient and reactive power change after load input; In step 7, the active power range P is found. stab (i) The element window corresponding to the maximum value, and the time when the element window is located is the load input time p0; Find the active power range P before and after the load connection time p0. stab (i) The element window corresponding to the minimum value is the time when the active power is most stable, p1 and p2. Find the reactive power range Q before and after the load connection time p0. stab (i) The element window corresponding to the minimum value is the time when the reactive power is most stable, p3 and p4. In step 8, the changes in active power, impact factor, and reactive power after load input are calculated, including: Change in active power ΔP = P mean (p2)-P mean (p1); Impact coefficient IMP = (P max (p0)-P mean (p1)) / △P; Reactive power change ΔQ = Q mean (p4)-Q mean (p3); In step 9, the type of load component is determined based on the impact coefficient and reactive power change after load input, including: If the impact coefficient is greater than 10, the type of load component is determined to be an electric motor; If the impact factor is greater than 2 and less than 10, the type of input load component is determined to be a power electronic interface load. If the impact coefficient is less than 2, further judgment is made by combining the reactive power change, where: If the reactive power change is less than -1, the load component type is a power electronic interface load; otherwise, it is determined to be a static load.

2. The non-intrusive power load component identification method based on event detection according to claim 1, characterized in that, The second fixed length is less than the first fixed length.

3. The non-intrusive power load component identification method based on event detection according to claim 1, characterized in that, The load component types include one of static load, power electronic interface load and motor load, wherein the static load is a constant impedance load.

4. The non-intrusive power load component identification method based on event detection according to claim 1, characterized in that, In step 3, the CUSUM event detection algorithm is used to monitor sudden changes in load active power, thereby capturing power load activation events, including: For each data point within a sub-window, the difference between each data point and the average value is calculated sequentially and accumulated. When the difference between the accumulated sum at a certain moment and the minimum accumulated sum exceeds the set threshold, an alarm is triggered, indicating that power load has been detected.

5. The non-intrusive power load component identification method based on event detection according to claim 1, characterized in that, In step 6, during the calculation of the load's active power parameters and reactive power parameters, the range is determined based on the maximum and minimum power values, i.e.: P stab (i)=P max (i)-P min (i); Q stab (i)=Q max (i)-Q min (i)。 6. A computer system, characterized in that, include: One or more processors; Memory stores instructions that can be operated. When the instructions are executed by one or more processors, they cause the aforementioned one or more processors to perform operations, including the process of performing the event detection-based non-intrusive power load identification method according to any one of claims 1-5.

7. A computer-readable storage medium for storing one or more programs, characterized in that, The one or more programs include instructions or instruction sets that can be executed by one or more processors; When the instructions or instruction set are executed by one or more processors, they perform the process of the non-intrusive power load identification method based on event detection as described in any one of claims 1-5.

Images