Cloud-based collaborative non-intrusive identification method for energy consumption monitoring based on master station feature library technology

Abstract

The present invention relates to a cloud-based collaborative non-intrusive identification method for energy consumption monitoring based on the main station feature library technology, comprising the following steps: step 1: non-intrusive terminal performs load identification; step 2: upload the physical characteristics and time characteristics of the uncertain electrical appliance to the main station; step 3: construct a feature library of commonly used electrical appliances at the main station; step 4: cluster the physical characteristics and time characteristics uploaded by the non-intrusive terminal with the feature library samples; step 5: match the clustered unknown electrical appliance characteristics with the feature library constructed by the main station, if the match is successful, the unknown electrical appliance is identified as an electrical appliance corresponding to the main station feature library; if no suitable load is matched in the feature library, turn to step 6; step 6: determine the unknown electrical appliance through field research and home interviews. The present invention realizes cloud-based collaborative identification of uncertain electrical appliances, and more electrical appliances can be more accurately identified without entering the residents' homes, and the identification results help users understand the energy consumption of their homes.

Description

Technical Field

[0001] The present invention relates to a cloud-based collaborative non-intrusive identification method for energy consumption monitoring based on a master station feature library technology, and belongs to the technical field of intelligent management. Background Art

[0002] As an important part of the power load, the proportion of electricity consumption of residential load is gradually increasing. Therefore, it is an urgent need to establish an intelligent power consumption system that can realize the visualization of household electricity consumption. It helps users understand the power consumption of various electrical equipment at different times, formulate reasonable energy-saving plans, purchase energy-saving equipment in a targeted manner, and test the energy-saving effect, thereby reducing energy consumption and electricity bills. Electricity consumption visualization is also considered to be one of the effective means to smooth out peak electricity consumption. It can encourage users to choose to use electricity at low electricity price periods such as nighttime, achieve the goal of peak shaving and valley filling, and effectively improve the efficiency of power investment.

[0003] At present, residential power load monitoring and decomposition technologies are mainly divided into two categories: intrusive load monitoring and decomposition (ILMD) and non-intrusive load monitoring and decomposition (NILMD):

[0004] (1) Intrusive load monitoring decomposition technology (ILMD): Intrusive load monitoring installs sensors with digital communication functions at the interface between each appliance and the power grid, which can accurately monitor the operating status and power consumption of each load. However, the installation of a large number of monitoring sensors results in high construction and maintenance costs. Most importantly, intrusive load monitoring requires entering residents' homes for installation and debugging, which is likely to cause user resistance.

[0005] (2) Non-intrusive load monitoring and decomposition technology (NILMD): Only one sensor is installed at the user's entrance. By collecting and analyzing the total current and voltage at the entrance, the power consumption and working state of each or each type of electrical appliance in the house (for example, the air conditioner has different working states such as cooling, heating, and standby) are determined, thereby obtaining the residents' electricity consumption pattern. Compared with intrusive load decomposition, since only one monitoring sensor needs to be installed, the construction cost and subsequent maintenance difficulty of the non-intrusive load decomposition solution are greatly reduced; in addition, the sensor installation location can be selected at the user's meter box, and the construction will not invade the residents' homes at all. It can be considered that NILMD replaces the sensor network of the ILMD system with a decomposition algorithm, which has the advantages of simplicity, economy, reliability, data integrity, and easy and rapid promotion and application. It is expected to develop into a new generation of core technology in the advanced metering system (AMI) (after maturity, the NILMD algorithm can also be integrated into the chip of the smart meter), supporting advanced functions of smart electricity consumption such as demand-side management and customized electricity, and is also suitable for temporary load power consumption detail monitoring and investigation.

[0006] In summary, existing research and literature have studied the models and operating characteristics of common loads, but there are still many appliances that cannot be accurately identified. Therefore, this paper proposes a cloud-based collaborative non-intrusive identification method for energy monitoring based on the master station feature library technology, which can effectively identify uncertain appliances that cannot be identified by the terminal, which is beneficial for users to understand their own energy consumption composition and minimize energy consumption. It helps power companies to accurately understand the composition of user power loads and provide more accurate basic data for demand-side management, load forecasting, and system planning. Summary of the invention

[0007] In order to solve the above technical problems, the present invention provides a cloud-based collaborative non-intrusive identification method for energy consumption monitoring based on the master station feature library technology. The method can accurately, quickly and effectively identify the load that the terminal cannot identify, and can be used for a non-intrusive load identification and decomposition device. The specific technical solution is as follows:

[0008] A cloud-based collaborative non-intrusive identification method for energy consumption monitoring based on a master station feature library technology comprises the following steps:

[0009] Step 1: Use a non-intrusive terminal to collect voltage and current signals at the user's incoming line end to identify the load;

[0010] Step 2: If the non-intrusive terminal identifies the appliance, it directly outputs the appliance. If the non-intrusive terminal cannot identify the appliance, it uploads the physical characteristics and time characteristics of the uncertain appliance to the master station. The set of physical characteristics and time characteristics is called load characteristic quantity.

[0011] Step 3: Build a feature library of commonly used electrical appliances at the main site;

[0012] Step 4: Cluster the physical features and time features uploaded by the non-intrusive terminal with the feature library samples;

[0013] Step 5: Match the clustered unknown electrical appliance features with the feature library built by the master station. If the match is successful, the unknown electrical appliance is identified as the electrical appliance corresponding to the feature library of the master station; if no suitable load is matched in the feature library, go to step 6;

[0014] Step 6: Identify the unknown appliance through field research and home interviews.

[0015] Furthermore, the physical characteristics of the uncertain electrical appliance in step 2 include the active power change form, active power amplitude, reactive power amplitude and average change of current harmonics when the appliance is started, and the time characteristics of the uncertain electrical appliance include the operating time and operating time.

[0016] Furthermore, the average change of current harmonics is specifically: the average change of the second harmonic, the average change of the third harmonic, the average change of the fifth harmonic, the average change of the seventh harmonic, and the average change of the ninth harmonic. The second harmonic amplitude, the third harmonic amplitude, the fifth harmonic amplitude, the seventh harmonic amplitude, and the ninth harmonic amplitude are calculated in sequence through the average change of current harmonics.

[0017] Furthermore, the feature library of common electrical appliances constructed by the master station in step 3 includes refrigerators, washing machines, air conditioners, water heaters, electric kettles, rice cookers, microwave ovens, induction cookers, lighting, and televisions;

[0018] The feature library contains physical features and time features. The physical features include the active power change form, active amplitude, reactive amplitude and current harmonic amplitude when the appliance is started. The time features include the operating time and operating time.

[0019] Furthermore, the clustering in step 4 is specifically as follows: clustering is performed using the DBSCAN density clustering algorithm.

[0020] Input: The data set P consisting of the load characteristics uploaded by the non-intrusive terminal and the characteristic library samples, the radius parameter ε, the density threshold MinPts,

[0021] Output: clustering results and noise data.

[0022] Furthermore, the radius parameter ε is determined as follows:

[0023] Given a data set P = {p(i); i = 0, 1, ... n}, for any point p(i), calculate the distance between point p(i) and all points in the set S = {p(0), p(1), p(2), ..., p(i-1), p(i+1), ..., p(n)}, and sort the distances in ascending order. Let the sorted distance set be J = {j(1), j(2), ..., j(k-1), j(k), j(k+1), ..., j(n)}, then j(k) is called the k distance; the k distance is the distance between point p( i) is the kth closest distance to all non-p(i) points; for each point p(i) in the cluster set, the k distance is calculated to obtain the k distance set E = {e(1), e(2), ..., e(n)} of all points, where e(n) is the k distance of the nth point; after sorting the k distance set E of all points in ascending order, the sorted k-distance set E' is obtained, and a k distance change curve in the sorted E' set is fitted to draw a curve, and the k distance value corresponding to the position where the change is sharp is determined as the value of the radius ε;

[0024]

[0025] The value of k is the density threshold MinPts, where m is the total number of samples participating in the master station clustering.

[0026] Furthermore, the matching process in step five is: clustering the load characteristics of all users uploaded by non-intrusive terminals for at least 30 days with the main station feature library. If 80% of the samples of the uncertain appliance are clustered into the same cluster with the feature data set of a certain appliance in the main station, then the uncertain appliance can be identified as the appliance in the main station; if it is less than 80% or the samples of the uncertain appliance are clustered into a class alone, it means that this is a new appliance that does not exist in the main station feature library.

[0027] Furthermore, the step six is ​​specifically as follows: according to the characteristic quantity of the uncertain electrical appliance, the uncertain electrical appliance is finally identified by using expert knowledge through laboratory testing, field research and door-to-door interviews.

[0028] The beneficial effects of the present invention are:

[0029] The present invention identifies uncertain electrical appliances that the terminal has failed to identify successfully after clustering in the cloud based on the load characteristic quantities uploaded by the terminal. This method can more accurately identify more electrical appliances without entering the residents' homes, and has the advantages of being simple, economical, and easy to promote and apply. The identification results help users understand the energy consumption of their homes and can be applied to the demand-side management of residential users by power grid companies. BRIEF DESCRIPTION OF THE DRAWINGS

[0030] Figure 1 is the overall flow chart of the present invention,

[0031] Figure 2 is the active power variation form of the present invention,

[0032] Figure 3 It is the current harmonic amplitude variation of the present invention. DETAILED DESCRIPTION

[0033] Figure 1 This is the overall flow chart of the present invention. The present invention proposes a cloud-based collaborative non-intrusive identification method for energy consumption monitoring based on the master station feature library technology. The specific process steps are as follows:

[0034] Step 1: Use a non-intrusive terminal to collect voltage and current signals at the user's incoming line end for load identification.

[0035] Step 2: If the non-intrusive terminal identifies an appliance, it will directly output the appliance. If the non-intrusive terminal cannot identify the appliance, it will upload the physical characteristics and time characteristics of the uncertain appliance to the main station. The non-intrusive terminal will successfully identify some appliances, and will also fail to identify some appliances, that is, uncertain appliances. At this time, the non-intrusive terminal will upload the physical characteristics and time characteristics of the uncertain appliance. The physical characteristics are: active power change form (see Figure 2 ), active amplitude, reactive amplitude, second current harmonic amplitude, third current harmonic amplitude, fifth current harmonic amplitude, seventh current harmonic amplitude and ninth current harmonic amplitude (see Figure 3 ), the time characteristics include running time and running time.

[0036] See also Figure 2 , active power change form upload number 1-4, except Figure 2 Except for the active power variation form shown in FIG. 1 , the active power variation form is assigned a value of 4; the non-intrusive terminal divides 24 hours a day into 96 time periods, 15 minutes as one time period, so the operating time takes values ​​1-96.

[0037] In order to prevent the terminal from uploading too much redundant information and reduce the uploading pressure on the terminal, this embodiment stipulates that the same unknown electrical appliance shall not upload more than 5 pieces of characteristic data in one day.

[0038] Step 3: Build a feature library of common electrical appliances at the main station. Build a feature library of common electrical appliances at the main station, including refrigerators, washing machines, air conditioners, water heaters, electric kettles, rice cookers, microwave ovens, induction cookers, lighting, and televisions. The feature library contains physical features and time features.

[0039] Step 4: Perform DBSCAN clustering on all users' 30-day load characteristics uploaded by non-intrusive terminals together with the master station's characteristic database. The radius parameter ε and density threshold MinPts are calculated based on the k-distance variation curve and formula (1).

[0040] Step 5: The uncertain electrical appliance samples uploaded by the terminal are clustered into a single class, which means that this is a new electrical appliance that does not exist in the main station feature library.

[0041] Step 6: After field research and home interviews, use expert knowledge to finally identify the uncertain appliance.

[0042] The application of the present invention is described in detail below:

[0043] The load characteristics uploaded by non-intrusive terminals within 30 days are as follows:

[0044]

[0045]

[0046] The appliance was in operation for 25 days in 30 days, and appeared every morning between 8 and 9, with an active power of 500W, indicating that it was a small appliance used in the morning before going to work. The presence of active power and current harmonics indicated that the uncertain appliance was an appliance containing motors and power electronic components. Based on research and expert knowledge, it was finally determined that the appliance was a wall-breaking machine.

[0047] Based on the above ideal embodiments of the present invention, the relevant staff can make various changes and modifications without departing from the technical concept of the present invention through the above description. The technical scope of the present invention is not limited to the contents of the specification, and its technical scope must be determined according to the scope of the claims.

Claims

1. A cloud-based collaborative non-intrusive identification method for energy consumption monitoring based on master station feature library technology, characterized in that: The following steps are involved: Step 1: Use a non-intrusive terminal to collect voltage and current signals at the user's incoming line end to identify the load; Step 2: If the non-intrusive terminal identifies the appliance, it directly outputs the appliance. If the non-intrusive terminal cannot identify the appliance, it uploads the physical characteristics and time characteristics of the uncertain appliance to the master station. The set of physical characteristics and time characteristics is called load characteristic quantity. The physical characteristics of the uncertain electrical appliance in step 2 include the active power change form, active power amplitude, reactive power amplitude and current harmonic amplitude change when the electrical appliance is started, and the time characteristics of the uncertain electrical appliance include the operation time and operation time; Step 3: Build a feature library of commonly used electrical appliances at the main site; The feature library of common electrical appliances constructed by the master station in step 3, the common electrical appliances include refrigerators, washing machines, air conditioners, water heaters, electric kettles, rice cookers, microwave ovens, induction cookers, lighting, and televisions; The feature library contains physical features and time features. The physical features include the active power change form, active amplitude, reactive amplitude and current harmonic amplitude when the appliance is started. The time features include the running time and running time. Step 4: Cluster the physical features and time features uploaded by the non-intrusive terminal with the feature library samples; The clustering in step 4 is specifically as follows: clustering is performed using the DBSCAN density clustering algorithm, Input: The load characteristics uploaded by the non-intrusive terminal and the data set P consisting of the characteristic library samples, the radius parameter ε, the density threshold MinPts, Output: clustering results and noise data; The radius parameter ε is determined as follows: Given a data set P={p(i); i=0,1,…n}, for any point p(i), calculate the distance between point p(i) and all points in the set S={p(0),p(1),p(2),…,p(i-1),p(i+1),…,p(n)}, and sort the distances in ascending order. Let the sorted distance set be J={j(1),j(2),…,j(k-1),j(k),j(k+1),…,j(n)}, then j(k) is called the k-distance; the k-distance is the kth closest distance between point p(i) and all non-p(i) points; calculate the k-distance for each point p(i) in the cluster set, and get the k-distance set E={e(1), e(2),…, e(n)}, e(n) is the k-distance of the nth point; sort the k-distance set E of all points in ascending order to obtain the sorted k-distance set E', fit a curve of the change of k-distance in the sorted E' set, draw a curve, and determine the value of k-distance corresponding to the position where the change is sharp as the value of radius ε; (1) The value of k is the density threshold MinPts , where m is the total number of samples participating in the master station clustering; Step 5: Match the clustered uncertain electrical appliance features with the feature library built by the master station. If the match is successful, the uncertain electrical appliance is identified as the electrical appliance corresponding to the master station feature library; if no suitable load is matched in the feature library, go to step 6; The matching process in step 5 is as follows: clustering the load characteristics of all users uploaded by the non-intrusive terminal for at least 30 days with the main station feature library. If 80% of the samples of the uncertain appliance are clustered into the same cluster with the feature data set of a certain appliance in the main station, then the uncertain appliance can be identified as the appliance in the main station; if it is less than 80% or the samples of the uncertain appliance are clustered into a single class, it means that this is a new appliance that does not exist in the main station feature library; Step 6: After field research and home interviews, determine the uncertain appliance; The step six is ​​specifically as follows: based on the characteristic quantity of the uncertain electrical appliance, the uncertain electrical appliance is finally identified by using expert knowledge through laboratory testing, field research and door-to-door interviews.

Images