CN115758188A - Non-invasive load identification method, device, equipment and medium - Google Patents
Non-invasive load identification method, device, equipment and medium Download PDFInfo
- Publication number
- CN115758188A CN115758188A CN202211182000.9A CN202211182000A CN115758188A CN 115758188 A CN115758188 A CN 115758188A CN 202211182000 A CN202211182000 A CN 202211182000A CN 115758188 A CN115758188 A CN 115758188A
- Authority
- CN
- China
- Prior art keywords
- initial
- load
- power information
- clustering
- sampling number
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Pending
Links
Images
Landscapes
- Management, Administration, Business Operations System, And Electronic Commerce (AREA)
Abstract
The invention discloses a non-invasive load identification method, a non-invasive load identification device, non-invasive load identification equipment and a non-invasive load identification medium, wherein the method comprises the steps of receiving a power information sequence of a power system to be identified in real time and dividing the power information sequence into a plurality of power information subsequences according to a preset length; sequentially detecting each power information subsequence by adopting a preset sliding window, and judging whether the power information subsequence has a load change event or not; if yes, extracting a plurality of multi-dimensional load characteristics from the power information subsequence to which the load change event belongs; performing dimension reduction operation on each multi-dimensional load characteristic, and respectively constructing initial characteristic object points; and clustering the initial feature object points, and identifying the load type of each initial feature object point. Therefore, the technical problem of low load identification accuracy caused by the European distance or the artificial parameter setting clustering scheme in the prior art is solved, and the load identification accuracy is effectively improved.
Description
Technical Field
The present invention relates to the field of load identification technologies, and in particular, to a non-intrusive load identification method, apparatus, device, and medium.
Background
With the increasing shortage of fossil energy and the increasing pressure of environmental protection, renewable energy is inevitably the main supply source of human energy. However, the characteristics of renewable energy sources (wind power generation, photovoltaic power generation, etc.) such as volatility, intermittency and regional imbalance severely restrict the large-scale use of renewable energy sources. In order to solve the problems caused by the fact that renewable energy is connected to a traditional power grid in a large scale, related concepts or technologies such as an energy internet, a smart power grid and a smart power internet of things are provided.
In an electric power system, power plants, substations and consumers are connected by transmission lines to form an electric power network. The power network is one of the complex systems with complex topological structure and huge system element number. In recent years, the development of new energy power generation is restricted by the problems of randomness, fluctuation and the like when the new energy power generation is incorporated into a traditional power grid. The effective demand side management can not only help the low-voltage distribution network side to enhance the operation efficiency of the power grid, but also relieve the energy pressure and improve the energy utilization efficiency. With the advance of the management work of the demand side, the common residential user domain is an important factor for realizing the intelligent management of the demand side. The actual energy consumption levels of various loads in the user can be known through load monitoring, and scientific collection and management of energy efficiency data are achieved.
In the prior art, load identification is usually performed by acquiring steady-state information, calculation is performed in a Euclidean distance mode on single-type load identification, the identification precision is easily reduced due to the influence of similar phases, and meanwhile, in the process of identifying multiple types of loads, clustering results are limited by artificial parameter setting, so that the accuracy of load identification is low.
Disclosure of Invention
The invention provides a non-invasive load identification method, a non-invasive load identification device, non-invasive load identification equipment and a non-invasive load identification medium, and solves the technical problem of low load identification accuracy caused by a Euclidean distance or a clustering scheme of manually set parameters in the prior art.
The invention provides a non-intrusive load identification method in a first aspect, which comprises the following steps:
receiving a power information sequence of a power system to be identified in real time, and dividing the power information sequence into a plurality of power information subsequences according to a preset length;
sequentially detecting each power information subsequence by adopting a preset sliding window, and judging whether a load change event occurs in the power information subsequences;
if yes, extracting a plurality of multi-dimensional load characteristics from the power information subsequence to which the load change event belongs;
performing dimension reduction operation on each multi-dimensional load characteristic, and respectively constructing initial characteristic object points;
and clustering the initial characteristic object points, and identifying the load type of each initial characteristic object point.
Optionally, the sliding window includes an average value calculation window and a variable point detection window which are connected in sequence; the step of sequentially detecting each power information subsequence by adopting a preset sliding window and judging whether the power information subsequence has a load change event or not comprises the following steps:
calculating the sequence average value of the power information subsequence in the average value calculation window, and accumulating the load switching quantity in real time;
if the load switching number is not accumulated to a preset switching threshold value in the average value calculation window, judging that the power information subsequence does not generate a load change event;
calculating a first superposed sampling number of the initial sampling number of the average value calculation window and the detectable sampling number of the variable point detection window;
taking the first superposed sampling number as the initial sampling number of the average value calculation window at the next moment;
if the load switching number is accumulated to the switching threshold value in the average value calculation window, judging that a load change event occurs in the power information subsequence;
calculating the initial sampling number of the average value calculation window at the current moment and a second superposed sampling number of the detectable sampling numbers;
and determining the difference value between the first superposed sampling number and the maximum time delay value as the initial sampling number of the average calculation window at the next moment.
Optionally, the clustering the initial feature object points and identifying the load type to which each of the initial feature object points belongs includes:
selecting any initial characteristic object point from the plurality of initial characteristic object points as a point to be judged;
determining a cluster growth radius corresponding to the point to be judged according to a preset target neighborhood parameter and a preset mapping relation;
performing DBSCAN clustering according to the cluster growth radius to obtain a target cluster corresponding to the point to be judged;
deleting initial feature object points belonging to the target cluster from the plurality of initial feature object points;
skipping to execute the step of selecting any initial characteristic object point from the plurality of initial characteristic object points as a point to be judged until no initial characteristic object point exists, and obtaining a plurality of target cluster clusters;
and determining the load type of each target cluster according to the comparison result of the multi-dimensional load characteristics corresponding to the points to be judged in each target cluster and the preset load characteristic categories.
Optionally, the method further comprises:
acquiring a plurality of initial samples and initial neighborhood parameters;
performing DBSCAN clustering on the plurality of initial samples according to the neighborhood parameters to obtain an initial clustering result;
judging whether the initial clustering result meets a preset clustering condition or not;
if not, selecting new initial neighborhood parameters to execute cross operation and mutation operation, and judging whether the current iteration times reach the maximum iteration times;
if the maximum iteration times are not reached, skipping to execute the step of carrying out DBSCAN clustering on the plurality of initial samples according to the neighborhood parameters to obtain an initial clustering result;
if the maximum iteration times are reached, determining the initial neighborhood parameters of the current moment as the initial values of the ant colony algorithm pheromones, and initializing the ant colony algorithm;
and calculating the state transition probability, updating the local pheromone until the optimal local pheromone is obtained, and determining the target neighborhood parameters by the optimal local pheromone.
A second aspect of the present invention provides a non-intrusive load identification apparatus, comprising:
the sequence dividing module is used for receiving the power information sequence of the power system to be identified in real time and dividing the power information sequence into a plurality of power information subsequences according to a preset length;
the event judgment module is used for sequentially detecting each power information subsequence by adopting a preset sliding window and judging whether the power information subsequence has a load change event or not;
a load feature extraction module, configured to extract a plurality of multidimensional load features from the power information subsequence to which the load change event belongs if the load change event belongs;
the characteristic dimension reduction module is used for executing dimension reduction operation on each multi-dimensional load characteristic and respectively constructing initial characteristic object points;
and the load type identification module is used for clustering the initial characteristic object points and identifying the load type of each initial characteristic object point.
Optionally, the sliding window includes an average value calculation window and a variable point detection window which are connected in sequence; the event judgment module is specifically configured to:
calculating the sequence average value of the power information subsequence in the average value calculation window, and accumulating the load switching quantity in real time;
if the load switching number is not accumulated to a preset switching threshold value in the average value calculation window, judging that the power information subsequence has no load change event;
calculating a first superposed sampling number of the initial sampling number of the average value calculation window and the detectable sampling number of the variable point detection window;
taking the first superposed sampling number as the initial sampling number of the average value calculation window at the next moment;
if the load switching quantity is accumulated to the switching threshold value in the average value calculation window, judging that a load change event occurs in the power information subsequence;
calculating the initial sampling number of the average value calculation window at the current moment and a second superposed sampling number of the detectable sampling numbers;
and determining the difference value between the first superposed sampling number and the maximum time delay value as the initial sampling number of the average calculation window at the next moment.
Optionally, the load type identifying module is specifically configured to:
selecting any initial characteristic object point from the plurality of initial characteristic object points as a point to be judged;
determining a cluster growth radius corresponding to the point to be judged according to a preset target neighborhood parameter and a preset mapping relation;
performing DBSCAN clustering according to the cluster growth radius to obtain a target cluster corresponding to the point to be judged;
deleting initial feature object points belonging to the target cluster from the plurality of initial feature object points;
skipping to execute the step of selecting any initial characteristic object point from the plurality of initial characteristic object points as a point to be judged until no initial characteristic object point exists, and obtaining a plurality of target cluster clusters;
and determining the load type of each target cluster according to the comparison result of the multi-dimensional load characteristics corresponding to the points to be judged in each target cluster and the preset load characteristic categories.
Optionally, the apparatus further comprises:
the initial data acquisition module is used for acquiring a plurality of initial samples and initial neighborhood parameters;
the clustering module is used for carrying out DBSCAN clustering on the initial samples according to the neighborhood parameters to obtain an initial clustering result;
the clustering condition judging module is used for judging whether the initial clustering result meets a preset clustering condition;
the judging module is used for selecting new initial neighborhood parameters to execute cross operation and mutation operation if the current iteration times reach the maximum iteration times;
the cluster skipping module is used for skipping to execute the step of carrying out DBSCAN clustering on the initial samples according to the neighborhood parameters to obtain an initial clustering result if the maximum iteration times are not reached;
the ant colony optimization module is used for determining the initial neighborhood parameters at the current moment as the initial values of the ant colony algorithm pheromone and initializing the ant colony algorithm if the maximum iteration times are reached;
and the target neighborhood parameter determining module is used for calculating the state transition probability, updating the local pheromone until the optimal local pheromone is obtained, and determining the target neighborhood parameter by the optimal local pheromone.
A third aspect of the present invention provides an electronic device comprising a memory and a processor, the memory having stored therein a computer program, which, when executed by the processor, causes the processor to perform the steps of the non-intrusive load identification method according to any one of the first aspects of the present invention.
A fourth aspect of the invention provides a computer readable storage medium having stored thereon a computer program which, when executed, implements a non-intrusive load identification method as defined in any one of the first aspects of the invention.
According to the technical scheme, the invention has the following advantages:
the method comprises the steps that a power information sequence of a power system to be identified is received in real time and is divided into a plurality of power information subsequences according to a preset length; sequentially detecting each power information subsequence by adopting a preset sliding window, and judging whether the power information subsequence has a load change event or not; if yes, extracting a plurality of multi-dimensional load characteristics from the power information subsequence to which the load change event belongs; performing dimension reduction operation on each multi-dimensional load characteristic, and respectively constructing initial characteristic object points; and clustering the initial characteristic object points, and identifying the load type of each initial characteristic object point. Therefore, the technical problem of low load identification accuracy caused by the European distance or the artificial parameter setting clustering scheme in the prior art is solved, and the load identification accuracy is effectively improved.
Drawings
In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings used in the description of the embodiments or the prior art will be briefly described below, and it is obvious that the drawings in the following description are only some embodiments of the present invention, and for those skilled in the art, other drawings can be obtained according to these drawings without inventive exercise.
Fig. 1 is a flowchart illustrating steps of a non-intrusive load identification method according to an embodiment of the present invention;
FIG. 2 is a schematic diagram of a bilateral CUSUM algorithm with sliding window according to the present invention;
FIG. 3 is a schematic diagram of transient feature extraction according to an embodiment of the present invention;
fig. 4 is a block diagram of a non-intrusive load identification apparatus according to an embodiment of the present invention.
Detailed Description
The intrusive load monitoring technology has relatively complex hardware and relatively simple software, and a monitoring device is required to be installed on each load to monitor real-time information of load electricity consumption. The method has reliable monitoring data, but has high cost, complex process and low practicability. The non-invasive load monitoring method has simpler hardware and more complex software, and only needs to install a monitoring device at a power supply inlet, identify various loads through a load identification algorithm and calculate the power consumption. For the power consumer, the non-intrusive load identification enables the power consumer to know the power consumption of the electric equipment in more detail, and leads the user to actively participate in peak clipping and valley filling of the power grid under the mechanism of power price stimulation. Meanwhile, the system can help a user to make a reasonable energy-saving plan, adjust the use of electric equipment, purchase energy-saving equipment in a targeted manner, and check the effects of the energy-saving plan and node equipment. Therefore, the electric energy consumption and the electricity charge expenditure are reduced on the premise that the normal production and life of the power consumer are not influenced.
For the power grid, the load composition of the power system can be known more truly, the load power utilization is standardized, the power grid utilization efficiency is improved, the power system investment is reduced, the running grid loss of the system is reduced, and the power failure time of power consumers is shortened. The non-intrusive load identification technology has very important significance for energy conservation and emission reduction and power system planning and operation of the whole society (including power consumers and power companies). Thus, non-intrusive load monitoring techniques have many advantages over intrusive load monitoring techniques.
The embodiment of the invention provides a non-invasive load identification method, a non-invasive load identification device, non-invasive load identification equipment and a non-invasive load identification medium, which are used for solving the technical problem of low load identification accuracy caused by a Euclidean distance or a clustering scheme of manually set parameters in the prior art.
In order to make the objects, features and advantages of the present invention more obvious and understandable, the technical solutions in the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings in the embodiments of the present invention, and it is obvious that the embodiments described below are only a part of the embodiments of the present invention, and not all of the embodiments. All other embodiments, which can be derived by a person skilled in the art from the embodiments given herein without making any creative effort, shall fall within the protection scope of the present invention.
Referring to fig. 1, fig. 1 is a flowchart illustrating a non-intrusive load identification method according to an embodiment of the present invention.
The invention provides a non-invasive load identification method, which comprises the following steps:
in the embodiment of the present invention, a power information sequence including, but not limited to, a time sequence of power, current or voltage, etc. may be received from the power system to be identified. After the power information sequence is received, the power information sequence is divided according to a preset length, and a plurality of power information subsequences are obtained and serve as data bases for subsequent load identification.
optionally, the sliding window includes an average value calculation window and a variable point detection window which are connected in sequence; step 102 may include the following sub-steps:
calculating the sequence average value of the power information subsequence in an average value calculation window, and accumulating the load switching quantity in real time;
if the load switching number is not accumulated to a preset switching threshold value in the average value calculation window, judging that no load change event occurs in the power information subsequence;
calculating the initial sampling number of an average value calculation window and a first superposed sampling number of detectable sampling numbers of a variable point detection window;
taking the first superposed sampling number as an average value to calculate the initial sampling number of the window at the next moment;
if the load switching number is accumulated to a switching threshold value in the average value calculation window, judging that a load change event occurs in the power information subsequence;
calculating the initial sampling number of the average value calculation window at the current moment and a second superposed sampling number capable of detecting the sampling number;
and determining the difference value between the first superposed sampling number and the maximum time delay value as the initial sampling number of the average calculation window at the next moment.
In the embodiment of the invention, the sequence average value of the power information subsequence in the average value calculation window is calculated, the load switching number is accumulated in real time, if the load switching number is not accumulated to a preset switching threshold value in the average value calculation window, the power information subsequence is judged not to have a load change event, the initial sampling number of the average value calculation window and the first superposed sampling number of the detectable sampling number of the variable point detection window are calculated, the first superposed sampling number is used as the initial sampling number of the average value calculation window at the next moment, if the load switching number is accumulated to the switching threshold value in the average value calculation window, the power information subsequence is judged to have the load change event, the initial sampling number of the average value calculation window at the current moment and the second superposed sampling number of the detectable sampling number are calculated, and the difference value of the first superposed sampling number and the maximum time delay value is determined as the initial sampling number of the average value calculation window at the next moment.
Referring to fig. 2, fig. 2 shows a schematic diagram of a bilateral CUSUM algorithm with a sliding Window according to the present invention, wherein the sliding Window in the diagram includes two windows, namely a Window of Mean (WM) and a Window of Detection (WD).
In the WM window, the algorithm calculates the average value mu of the relative stable period, and the number of sampling points in the WM window is N m . In a WD window, an algorithm detects whether a time sequence has a change point, and the number of sampling points in the WD window is N d 。N m And N d Is suitably chosen as N m Too small, the calculated mean is not accurate, if N is too small m If the size is too large, the variable points may be covered, so that detection is missed; if N is present d If it is too small, it may not cover a complete change point detection process, resulting in missed detection, if N is d Too large, multiple change points may occur in one window. Thus, N m And N d The specific value rule of (a) can be summarized as: n is a radical of m The length of the variable point is required to meet the precision requirement of calculating the mean value and cannot be larger than the distance between two variable points; n is a radical of hydrogen d The length of (a) is required to cover at least one complete change point detection process, and cannot cover two change points simultaneously. During the detection, the WM and WD windows continue to slide to the right for a new period. When no change point occurs, before opening the next sliding window, the sliding window is opened
And
and (4) returning to zero. At this time, to prevent missing detection, it should be ensured that the WD window is connected end to end, i.e. the sliding window moves by a distance N d (ii) a When the change point occurs at the end of the WD window,
or
Will accumulate from the time the change point occurs until the end of the WD window, at which time
Or
Greater than zero. But if
And
there is insufficient time to accumulate to the threshold h, the change point is not detected in the WD window. To prevent missing detection, the task of detecting the change point should be handed over to the next WD window, and the head end of the next WD window should be moved to the moment of the change point, i.e. the distance of moving the sliding window is
Combining the above two situations, if no change point is detected in the current sliding window, the distance moved by the next set of sliding windows is
If a change point is detected in the current sliding window, the next set of sliding windows should be slid to the next time the algorithm detects the change point.
It should be noted that the change point refers to the embodiment of the present inventionThe event of a change in the load is,
for maximum time delay, the load switching quantity comprises load input data
And number of load shedding
after the event detection task is completed, load features need to be extracted. The load characteristics are mainly classified into steady-state characteristics and transient characteristics. Wherein, the steady-state feature extraction is simpler. However, many loads have relatively similar steady-state characteristics under certain conditions, and therefore, it is difficult to achieve accurate load identification using only the steady-state characteristics. While most loads are unique in transient processes at the start of the device. This enables the transient characteristics to provide more valuable information for load identification, and therefore, a plurality of multidimensional load characteristics can be extracted from the power information subsequence to which the load change event belongs, wherein the multidimensional load characteristics can include a transient characteristic and a steady-state characteristic, wherein, during the transient process of the load start, the peak value of the current is defined as a peak current, the steady-state value of the current is defined as a steady-state current, the time from the current start to rise to reach the peak value is defined as a peak time, and the time from the current start to rise to reach the steady-state is defined as a steady-state time. The four characteristics of different loads are generally different, and the four characteristics under the condition of different operation times of the same load are approximately the same, which shows that the four characteristics are unique to each load and can be used as a basis for distinguishing different loads, so that the transient characteristics selected and extracted in the text are the four characteristics of peak current, peak time, steady-state current and steady-state time of the load in the starting process.
As shown in fig. 3, fig. 3 shows a schematic diagram of transient feature extraction in the embodiment of the present invention.
The characteristic information completely describing a load mainly comprises voltage, current, power factor and harmonic wave, and most other characteristic information of the load, such as active power, reactive power, apparent power, impedance, admittance and the like, can be deduced based on the four quantities.
In the process of extracting the steady-state characteristics of the load, because each household load is connected with the bus in parallel in the operation of the actual non-invasive load monitoring system, the voltage of each load is basically the same as the voltage of the bus, useful information for distinguishing the load cannot be provided, and the difference of the current is extracted in the transient characteristic extraction link, so that almost all information for completely describing one load can be contained only by extracting the power factor and the harmonic wave. The power factor can obtain a corresponding numerical value through the acquisition module to serve as a comparison parameter.
The extracted steady state features are therefore selected to be the power factor of the load and the odd harmonic amplitudes of the current. Because the amplitude of the harmonic wave flowing through the higher order is too small and provides too little useful information, the invention only extracts the fundamental wave and the amplitudes of the harmonic waves of 3, 5, 7 and 9 orders of the current. The harmonic current is a general term of each cosine component with the frequency which is odd times of the frequency of the power frequency periodic current when a non-sinusoidal periodic current curve of a load is expanded according to the Fourier series. The load original current i (t) can be decomposed into
i(t)＝I 1 cos(ωt+θ 1 )+I 3 cos(3ωt+θ 3 )+...+I k cos(kωt+θ k )
In the formula, k is an odd number and represents the harmonic order, k =1 represents the fundamental wave, and the harmonic currents with k larger than 1 are called k harmonic currents; i is k Is the amplitude of the k harmonic current; omega is the angular frequency of the fundamental current; theta k Is the initial phase angle of the k harmonic current.
104, performing dimension reduction operation on each multi-dimensional load feature, and respectively constructing initial feature object points;
in order to avoid the problems that the dimension of the load characteristic is too large and the calculation cost is increased, the multidimensional load characteristic can be reduced to a two-dimensional coordinate form to construct initial characteristic object points to wait for subsequent clustering.
And 105, clustering the initial feature object points, and identifying the load type of each initial feature object point.
Optionally, step 105 may comprise the sub-steps of:
selecting any initial characteristic object point from the plurality of initial characteristic object points as a point to be judged;
determining a cluster growth radius corresponding to a point to be judged according to a preset target neighborhood parameter and a preset mapping relation;
performing DBSCAN clustering according to the cluster growth radius to obtain a target cluster corresponding to the point to be judged;
deleting initial characteristic object points belonging to the target cluster from the plurality of initial characteristic object points;
skipping to execute the step of selecting any initial characteristic object point from the plurality of initial characteristic object points as a point to be judged until no initial characteristic object point exists, and obtaining a plurality of target cluster clusters;
and determining the load type of each target cluster according to the comparison result of the multi-dimensional load characteristics corresponding to the points to be judged in each target cluster and the preset load characteristic categories.
In the embodiment of the invention, firstly, a better solution is found by utilizing an optimal retention mechanism of a genetic algorithm, and a global optimal solution is obtained by converting the better solution into the initial pheromone of the ant colony algorithm. Secondly, the genetic ant colony algorithm is used for finding out the global optimal parameters to be used as the input of the DBSCAN algorithm, and the defect that the DBSCAN algorithm is sensitive to the parameters is overcome. And thirdly, when the DBSCAN algorithm is used for calculation, the effective distance and the sigmod function are used for constructing an adaptive coefficient function, and the selection of the radius threshold value during the near point search in the cluster growth is optimized. And finally, clustering by using a self-adaptive threshold DBSCAN clustering algorithm to complete load identification.
It should be noted that, in the embodiment of the present invention, the cluster growth radius of the DBSCAN cluster may be adjusted through an adaptive threshold policy, and specifically, the adaptive threshold policy requires to construct a mapping relationship based on a specific parameter. The sigmoid function is a commonly used S-shaped nonlinear activation function, and can map a real number between 0 and 1. Considering the characteristic that the acquired data has strong sparsity, the sigmoid function has good matching performance on the data, and therefore the sigmoid function is selected as a basic model of the constructed mapping relation. The constructed mapping relation is
Wherein k is r 、c r 、ε r 、r 0 、
Are all model parameters; the position of any data point on the plane can be represented by coordinates (xi, yi),
radius range of data points. Epsilon r 、
Two parameters are the scaling factors in the model by dynamically adjusting epsilon r 、
Two parameters, the parameter range of the scaling factor is [0,1]In the meantime. By enumerating different values of the parameters, the radius of cluster growth can be influenced:
R * ＝R 0 ×f(r i )
R * the corrected radius parameter; r 0 Is an initial radius parameter; r is a radical of hydrogen i To search for data points of the same cluster point.
Further, the method comprises the following steps S11-S17:
s11, obtaining a plurality of initial samples and initial neighborhood parameters;
s12, carrying out DBSCAN clustering on the plurality of initial samples according to the neighborhood parameters to obtain an initial clustering result;
s13, judging whether the initial clustering result meets a preset clustering condition or not;
s14, if not, selecting new initial neighborhood parameters to execute cross operation and mutation operation, and judging whether the current iteration number reaches the maximum iteration number;
s15, if the maximum iteration times are not reached, skipping to execute the step of carrying out DBSCAN clustering on a plurality of initial samples according to neighborhood parameters to obtain an initial clustering result;
s16, if the maximum iteration times are reached, determining the initial neighborhood parameters at the current moment as initial values of the ant colony algorithm pheromones, and initializing the ant colony algorithm;
s17, calculating the state transition probability, updating the local pheromone until the optimal local pheromone is obtained, and determining the target neighborhood parameters of the optimal local pheromone.
In specific implementation, the adjustment of the neighborhood parameters and the subsequent DBSCAN clustering process can be realized through the following steps:
step 1: input sample set D = { x = 1 ,x 2 ,...,x m And the neighborhood parameters (ε, minPts). And performing mathematical description on the most available parameters according to the parameter information.
Step 2: initializing genetic algorithm parameters including population scale, crossover and mutation probability, evolution algebra, termination condition and the like.
And 3, generating a population by adopting a random initialization mode, and calculating the individual fitness. Setting iteration initial value as 1, and continuously generating new individuals by using genetic operators such as selection, crossover and mutation.
And 4, judging whether the current iteration times reach the maximum iteration times. And if the cycle ending condition is met, converting the optimal solution into an initial value of the ant colony algorithm pheromone. Otherwise, step 3 is executed.
And 5, initializing ant colony algorithm related parameters.
And 6, calculating the state transition probability and updating the local pheromone. The global pheromone is updated after the optimal solution is found.
And 7: and taking the calculated optimal solution as an input parameter of the DBSCAN algorithm.
And 8, traversing all points in the sample set, finding all core points, and adding the core points into the set omega.
And step 9: randomly selecting a core point from omega, finding out points connected by all densities of which the density of the core point can reach by using a formula (2) to form a cluster, and then removing all the core points in the cluster from omega.
And 10, repeatedly executing the step (9) until the omega is empty.
Step 11: and outputting a clustering result.
In the embodiment of the invention, the power information sequence of the power system to be identified is received in real time and is divided into a plurality of power information subsequences according to the preset length; sequentially detecting each power information subsequence by adopting a preset sliding window, and judging whether the power information subsequence has a load change event or not; if yes, extracting a plurality of multi-dimensional load characteristics from the power information subsequence to which the load change event belongs; performing dimension reduction operation on each multi-dimensional load characteristic, and respectively constructing initial characteristic object points; and clustering the initial characteristic object points, and identifying the load type of each initial characteristic object point. Therefore, the technical problem of low load identification accuracy caused by the European distance or the artificial parameter setting clustering scheme in the prior art is solved, and the load identification accuracy is effectively improved.
Referring to fig. 4, fig. 4 is a block diagram illustrating a non-intrusive load identification apparatus according to an embodiment of the present application.
The embodiment of the invention provides a non-invasive load identification device, which comprises:
the sequence dividing module 401 is configured to receive a power information sequence of the power system to be identified in real time, and divide the power information sequence into a plurality of power information subsequences according to a preset length;
an event determining module 402, configured to sequentially detect each power information subsequence by using a preset sliding window, and determine whether a load change event occurs in the power information subsequence;
a load feature extraction module 403, configured to, if yes, extract a plurality of multidimensional load features from the power information subsequence to which the load change event belongs;
a feature dimension reduction module 404, configured to perform dimension reduction on each multi-dimensional load feature, and respectively construct initial feature object points;
and a load type identification module 405, configured to cluster the initial feature object points, and identify a load type to which each initial feature object point belongs.
Optionally, the sliding window includes an average value calculation window and a variable point detection window which are connected in sequence; the event determining module 402 is specifically configured to:
calculating the sequence average value of the power information subsequence in an average value calculation window, and accumulating the load switching quantity in real time;
if the load switching number is not accumulated to a preset switching threshold value in the average value calculation window, judging that no load change event occurs in the power information subsequence;
calculating the initial sampling number of an average value calculation window and a first superposed sampling number of detectable sampling numbers of a variable point detection window;
taking the first superposed sampling number as an average value to calculate the initial sampling number of the window at the next moment;
if the load switching number is accumulated to a switching threshold value in the average value calculation window, judging that a load change event occurs in the power information subsequence;
calculating the initial sampling number of the average value calculation window at the current moment and a second superposed sampling number capable of detecting the sampling number;
and determining the difference value between the first superposed sampling number and the maximum time delay value as the initial sampling number of the average value calculation window at the next moment.
Optionally, the load type identifying module 404 is specifically configured to:
selecting any initial characteristic object point from the plurality of initial characteristic object points as a point to be judged;
determining a cluster growth radius corresponding to a point to be judged according to a preset target neighborhood parameter and a preset mapping relation;
performing DBSCAN clustering according to the cluster growth radius to obtain a target cluster corresponding to the point to be judged;
deleting initial characteristic object points belonging to the target cluster from the plurality of initial characteristic object points;
skipping to execute the step of selecting any initial characteristic object point from the plurality of initial characteristic object points as a point to be judged until no initial characteristic object point exists, and obtaining a plurality of target cluster clusters;
and determining the load type of each target cluster according to the comparison result of the multi-dimensional load characteristics corresponding to the points to be judged in each target cluster and the preset load characteristic categories.
Optionally, the apparatus further comprises:
the initial data acquisition module is used for acquiring a plurality of initial samples and initial neighborhood parameters;
the clustering module is used for carrying out DBSCAN clustering on the initial samples according to the neighborhood parameters to obtain an initial clustering result;
the clustering condition judging module is used for judging whether the initial clustering result meets a preset clustering condition;
the judging module is used for selecting new initial neighborhood parameters to execute cross operation and mutation operation if the current iteration times reach the maximum iteration times;
the cluster skipping module is used for skipping to execute the step of carrying out DBSCAN clustering on the initial samples according to the neighborhood parameters to obtain an initial clustering result if the maximum iteration times are not reached;
the ant colony optimization module is used for determining the initial neighborhood parameters at the current moment as the initial values of the ant colony algorithm pheromone and initializing the ant colony algorithm if the maximum iteration times are reached;
and the target neighborhood parameter determining module is used for calculating the state transition probability, updating the local pheromone until the optimal local pheromone is obtained, and determining the target neighborhood parameter by the optimal local pheromone.
An embodiment of the present invention provides an electronic device, which includes a memory and a processor, where the memory stores a computer program, and when the computer program is executed by the processor, the processor executes the steps of the non-intrusive load identification method according to any embodiment of the present invention.
Embodiments of the present invention provide a computer-readable storage medium, on which a computer program is stored, where the computer program is executed to implement a non-intrusive load identification method according to any embodiment of the present invention.
It can be clearly understood by those skilled in the art that, for convenience and brevity of description, the specific working processes of the above-described apparatuses and modules may refer to the corresponding processes in the foregoing method embodiments, and are not described herein again.
In the embodiments provided in the present invention, it should be understood that the disclosed apparatus and method may be implemented in other ways. For example, the above-described apparatus embodiments are merely illustrative, and for example, the division of the units is only one logical division, and other divisions may be realized in practice, for example, a plurality of units or components may be combined or integrated into another system, or some features may be omitted, or not executed. In addition, the shown or discussed mutual coupling or direct coupling or communication connection may be an indirect coupling or communication connection through some interfaces, devices or units, and may be in an electrical, mechanical or other form.
The units described as separate parts may or may not be physically separate, and parts displayed as units may or may not be physical units, may be located in one place, or may be distributed on a plurality of network units. Some or all of the units can be selected according to actual needs to achieve the purpose of the solution of the embodiment.
In addition, functional units in the embodiments of the present invention may be integrated into one processing unit, or each unit may exist alone physically, or two or more units are integrated into one unit. The integrated unit can be realized in a form of hardware, and can also be realized in a form of a software functional unit.
The integrated unit, if implemented in the form of a software functional unit and sold or used as a stand-alone product, may be stored in a computer readable storage medium. Based on such understanding, the technical solution of the present invention may be embodied in the form of a software product, which is stored in a storage medium and includes instructions for causing a computer device (which may be a personal computer, a server, or a network device) to execute all or part of the steps of the method according to the embodiments of the present invention. And the aforementioned storage medium includes: a U-disk, a removable hard disk, a Read-Only Memory (ROM), a Random Access Memory (RAM), a magnetic disk, or an optical disk, and various media capable of storing program codes.
The above-mentioned embodiments are only used for illustrating the technical solutions of the present invention, and not for limiting the same; although the present invention has been described in detail with reference to the foregoing embodiments, it will be understood by those of ordinary skill in the art that: the technical solutions described in the foregoing embodiments may still be modified, or some technical features may be equivalently replaced; and such modifications or substitutions do not depart from the spirit and scope of the corresponding technical solutions of the embodiments of the present invention.
Claims (10)
1. A method of non-intrusive load identification, comprising:
receiving a power information sequence of a power system to be identified in real time, and dividing the power information sequence into a plurality of power information subsequences according to a preset length;
sequentially detecting each power information subsequence by adopting a preset sliding window, and judging whether a load change event occurs in the power information subsequences;
if yes, extracting a plurality of multi-dimensional load characteristics from the power information subsequence to which the load change event belongs;
performing dimension reduction operation on each multi-dimensional load characteristic, and respectively constructing initial characteristic object points;
and clustering the initial feature object points, and identifying the load type of each initial feature object point.
2. The method of claim 1, wherein the sliding window comprises a mean calculation window and a change point detection window connected in sequence; the step of sequentially detecting each power information subsequence by adopting a preset sliding window and judging whether the power information subsequence has a load change event or not comprises the following steps:
calculating the sequence average value of the power information subsequence in the average value calculation window, and accumulating the load switching quantity in real time;
if the load switching number is not accumulated to a preset switching threshold value in the average value calculation window, judging that the power information subsequence has no load change event;
calculating a first superposed sampling number of the initial sampling number of the average value calculation window and the detectable sampling number of the variable point detection window;
taking the first superposed sampling number as the initial sampling number of the average value calculation window at the next moment;
if the load switching quantity is accumulated to the switching threshold value in the average value calculation window, judging that a load change event occurs in the power information subsequence;
calculating the initial sampling number of the average value calculation window at the current moment and a second superposed sampling number of the detectable sampling numbers;
and determining the difference value between the first superposed sampling number and the maximum time delay value as the initial sampling number of the average calculation window at the next moment.
3. The method according to claim 1, wherein the step of clustering the initial feature object points and identifying the load type to which each of the initial feature object points belongs comprises:
selecting any initial characteristic object point from the plurality of initial characteristic object points as a point to be judged;
determining a cluster growth radius corresponding to the point to be judged according to a preset target neighborhood parameter and a preset mapping relation;
performing DBSCAN clustering according to the cluster growth radius to obtain a target cluster corresponding to the point to be judged;
deleting initial feature object points belonging to the target cluster from the plurality of initial feature object points;
skipping to execute the step of selecting any initial characteristic object point from the plurality of initial characteristic object points as a point to be judged until no initial characteristic object point exists, and obtaining a plurality of target cluster clusters;
and determining the load type of each target cluster according to the comparison result of the multi-dimensional load characteristics corresponding to the points to be judged in each target cluster and the preset load characteristic categories.
4. The method according to any one of claims 1-3, further comprising:
acquiring a plurality of initial samples and initial neighborhood parameters;
performing DBSCAN clustering on the plurality of initial samples according to the neighborhood parameters to obtain an initial clustering result;
judging whether the initial clustering result meets a preset clustering condition or not;
if not, selecting new initial neighborhood parameters to execute cross operation and mutation operation, and judging whether the current iteration times reach the maximum iteration times;
if the maximum iteration times are not reached, skipping to execute the step of carrying out DBSCAN clustering on the plurality of initial samples according to the neighborhood parameters to obtain an initial clustering result;
if the maximum iteration times are reached, determining the initial neighborhood parameters of the current moment as the initial values of the ant colony algorithm pheromones, and initializing the ant colony algorithm;
and calculating the state transition probability, updating the local pheromone until the optimal local pheromone is obtained, and determining the target neighborhood parameters by the optimal local pheromone.
5. A non-intrusive load recognition device, comprising:
the sequence dividing module is used for receiving the power information sequence of the power system to be identified in real time and dividing the power information sequence into a plurality of power information subsequences according to a preset length;
the event judgment module is used for sequentially detecting each power information subsequence by adopting a preset sliding window and judging whether the power information subsequence has a load change event or not;
the load characteristic extraction module is used for extracting a plurality of multi-dimensional load characteristics from the power information subsequence to which the load change event belongs if the load characteristic extraction module is used for extracting the multi-dimensional load characteristics from the power information subsequence to which the load change event belongs;
the characteristic dimension reduction module is used for executing dimension reduction operation on each multi-dimensional load characteristic and respectively constructing initial characteristic object points;
and the load type identification module is used for clustering the initial characteristic object points and identifying the load type of each initial characteristic object point.
6. The apparatus of claim 5, wherein the sliding window comprises a mean calculation window and a change point detection window connected in sequence; the event judgment module is specifically configured to:
calculating the sequence average value of the power information subsequence in the average value calculation window, and accumulating the load switching quantity in real time;
if the load switching number is not accumulated to a preset switching threshold value in the average value calculation window, judging that the power information subsequence has no load change event;
calculating a first superposed sampling number of the initial sampling number of the average value calculation window and the detectable sampling number of the variable point detection window;
taking the first superposed sampling number as the initial sampling number of the average value calculation window at the next moment;
if the load switching quantity is accumulated to the switching threshold value in the average value calculation window, judging that a load change event occurs in the power information subsequence;
calculating the initial sampling number of the average value calculation window at the current moment and a second overlapped sampling number of the detectable sampling numbers;
and determining the difference value between the first superposed sampling number and the maximum time delay value as the initial sampling number of the average calculation window at the next moment.
7. The apparatus according to claim 5, wherein the load type identification module is specifically configured to:
selecting any initial characteristic object point from the plurality of initial characteristic object points as a point to be judged;
determining a cluster growth radius corresponding to the point to be judged according to a preset target neighborhood parameter and a preset mapping relation;
performing DBSCAN clustering according to the cluster growth radius to obtain a target cluster corresponding to the point to be judged;
deleting initial feature object points belonging to the target cluster from the plurality of initial feature object points;
skipping to execute the step of selecting any initial characteristic object point from the plurality of initial characteristic object points as a point to be judged until no initial characteristic object point exists, and obtaining a plurality of target cluster clusters;
and determining the load type of each target cluster according to the comparison result of the multi-dimensional load characteristics corresponding to the points to be judged in each target cluster and the preset load characteristic categories.
8. The apparatus of any of claims 5-7, further comprising:
the initial data acquisition module is used for acquiring a plurality of initial samples and initial neighborhood parameters;
the clustering module is used for carrying out DBSCAN clustering on the initial samples according to the neighborhood parameters to obtain an initial clustering result;
the clustering condition judging module is used for judging whether the initial clustering result meets a preset clustering condition or not;
the judging module is used for selecting new initial neighborhood parameters to execute cross operation and mutation operation if the current iteration times reach the maximum iteration times;
the cluster skipping module is used for skipping to execute the step of carrying out DBSCAN clustering on the initial samples according to the neighborhood parameters to obtain an initial clustering result if the maximum iteration times are not reached;
the ant colony optimization module is used for determining the initial neighborhood parameters at the current moment as the initial values of the ant colony algorithm pheromone and initializing the ant colony algorithm if the maximum iteration times are reached;
and the target neighborhood parameter determining module is used for calculating the state transition probability, updating the local pheromone until the optimal local pheromone is obtained, and determining the target neighborhood parameter by the optimal local pheromone.
9. An electronic device comprising a memory and a processor, the memory having stored thereon a computer program that, when executed by the processor, causes the processor to perform the steps of the non-intrusive load identification method as defined in any one of claims 1-4.
10. A computer-readable storage medium, on which a computer program is stored, which, when executed, implements the non-intrusive load identification method of any one of claims 1-4.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202211182000.9A CN115758188A (en) | 2022-09-27 | 2022-09-27 | Non-invasive load identification method, device, equipment and medium |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202211182000.9A CN115758188A (en) | 2022-09-27 | 2022-09-27 | Non-invasive load identification method, device, equipment and medium |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN115758188A true CN115758188A (en) | 2023-03-07 |
Family
ID=85351984
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202211182000.9A Pending CN115758188A (en) | 2022-09-27 | 2022-09-27 | Non-invasive load identification method, device, equipment and medium |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN115758188A (en) |
Cited By (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN116865451A (en) * | 2023-09-04 | 2023-10-10 | 湖南巨森电气集团有限公司 | Intelligent power consumption control management system and method |
| CN118114137A (en) * | 2024-04-30 | 2024-05-31 | 浙江大华技术股份有限公司 | Non-invasive load identification method, device and computer storage medium |
-
2022
- 2022-09-27 CN CN202211182000.9A patent/CN115758188A/en active Pending
Cited By (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN116865451A (en) * | 2023-09-04 | 2023-10-10 | 湖南巨森电气集团有限公司 | Intelligent power consumption control management system and method |
| CN116865451B (en) * | 2023-09-04 | 2023-11-28 | 湖南巨森电气集团有限公司 | Intelligent power consumption control management system and method |
| CN118114137A (en) * | 2024-04-30 | 2024-05-31 | 浙江大华技术股份有限公司 | Non-invasive load identification method, device and computer storage medium |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| Angelis et al. | NILM applications: Literature review of learning approaches, recent developments and challenges | |
| Wen et al. | A shape-based clustering method for pattern recognition of residential electricity consumption | |
| Deng et al. | A missing power data filling method based on improved random forest algorithm | |
| CN110956220B (en) | Non-invasive household appliance load identification method | |
| CN109325545A (en) | Low-voltage network topological structure method of calibration, device, equipment and storage medium | |
| CN115758188A (en) | Non-invasive load identification method, device, equipment and medium | |
| CN109840691B (en) | Non-invasive polynomial electric quantity estimation method based on deep neural network | |
| CN110380444B (en) | Capacity planning method for distributed wind power orderly access to power grid under multiple scenes based on variable structure Copula | |
| CN109470957B (en) | Non-invasive electrical appliance identification method based on Frey's distance | |
| CN111654392A (en) | Low-voltage distribution network topology identification method and system based on mutual information | |
| CN114444539A (en) | Power load identification method, apparatus, device, medium, and program product | |
| CN112989133B (en) | Graph data modeling power fingerprint identification method, storage medium and system for electrical equipment | |
| CN113659564A (en) | Low-voltage distribution network topology identification method and system based on voltage fluctuation feature clustering | |
| CN113010985A (en) | Non-invasive load identification method based on parallel AANN | |
| Domínguez et al. | Analysis of electric power consumption using Self-Organizing Maps. | |
| CN116467631A (en) | Power fingerprint identification model training method, power fingerprint identification method and device | |
| CN115528684A (en) | Ultra-short-term load prediction method and device and electronic equipment | |
| CN114186631A (en) | Load identification method based on non-invasive intelligent terminal | |
| Yan et al. | A clustering method for power time series curves based on improved self-organizing mapping algorithm | |
| Yanhui et al. | Forecasting hourly electricity demand with nonparametric functional data analysis | |
| CN113224748A (en) | Method for calculating line loss of low-voltage distribution station area | |
| CN111815022A (en) | Power load prediction method based on time-delay coordinate embedding method | |
| Liu et al. | Hierarchical identification method of household-transformer relationship based on adaptive piecewise aggregation approximation | |
| CN109064009A (en) | Substation's industry ratio acquisition methods and calculating equipment based on EMS data | |
| CN113131489B (en) | Non-invasive load identification method based on improved artificial fish swarm clustering algorithm |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination |