Electric water heater protection control system based on clustering processing

Abstract

The application discloses an electric water heater protection control system based on clustering processing, which comprises a data acquisition module, a basic clustering generation module, an optimal fuzzy matrix learning module, a sample screening module, an optimization strategy design module and a protection control module. The application belongs to the field of data processing and specifically refers to an electric water heater protection control system based on clustering processing. The scheme introduces a joint target of physical space Euclidean distance and entropy regularization to reduce misjudgment of a real fault boundary. A physical distance penalty term is introduced to improve the sensitivity to real fault state changes. The fault tolerance to sensor noise and boundary state is enhanced through fuzzy neighbors. Noise and fault points are removed in combination with fuzzy similarity and outlying degree to perform all-around abnormal filtering. The risk of equipment failure is reduced. Random coefficients and difference information are combined to adjust hyperparameters, so that good protection control effects can be achieved under different working conditions.

Description

TECHNICAL FIELD

[0001] The present application relates to the field of data processing, in particular to an electric water heater protection control system based on clustering processing. BACKGROUND

[0002] The electric water heater protection control system is a combination of a series of functional modules and technologies that guarantee safety in use, ensure stable operation and prolong the service life of the equipment. However, the general electric water heater protection control system has the problem of sensor measurement noise and operating state boundary, which leads to a sharp fluctuation in single division results and an increase in misjudgment rate. The general electric water heater protection control system also has the problem of improper identification of outlier data in electric water heater operating data, which makes it difficult to find short-term sudden abnormal working conditions, resulting in poor electric water heater protection control effect. SUMMARY

[0003] In view of the above situation, in order to overcome the defects of the prior art, the present application provides an electric water heater protection control system based on clustering processing. In view of the problem of sensor measurement noise and operating state boundary in the general electric water heater protection control system, which leads to a sharp fluctuation in single division results and an increase in misjudgment rate, the present application introduces a joint target of physical space Euclidean distance and entropy regularization to reduce misjudgment of the real fault boundary; introduces a physical distance penalty term to improve the sensitivity to real fault state changes and reduce misjudgment; enhances the fault tolerance to sensor noise and boundary state through fuzzy neighbors, reducing the risk of false positives and false negatives; in view of the problem of improper identification of outlier data in electric water heater operating data in the general electric water heater protection control system, which makes it difficult to find short-term sudden abnormal working conditions, resulting in poor electric water heater protection control effect, the present application combines fuzzy similarity and outlier degree to remove noise and fault points for all-round abnormal filtering; both single-point abnormal alarm and rare working condition capture based on state transition probability are used to reduce the risk of equipment failure; random coefficients and difference information are combined to adjust hyperparameters, ensuring good protection control effect under different working conditions.

[0004] The technical scheme adopted by the present application is as follows: the electric water heater protection control system based on clustering processing provided by the present application comprises a data acquisition module, a basic clustering generation module, an optimal fuzzy matrix learning module, a sample screening module, an optimization strategy design module and a protection control module;

[0005] The data acquisition module acquires electric water heater heating element operating data and constructs an electric water heater sample set;

[0006] The basic clustering generation module runs K-means clustering in parallel to generate a basic clustering result for the electric water heater sample set and construct a co-association matrix;

[0007] The optimal fuzzy matrix learning module constructs an initial fuzzy co-association matrix based on a co-association matrix, and iteratively learns an optimal fuzzy co-association matrix through alternating optimization to obtain a preliminary clustering result of the electric water heater sample set;

[0008] The sample screening module calculates sample outlying degree to remove abnormal points and counts state transition probability of a clustering label based on the preliminary clustering result of the electric water heater sample set, triggers rare abnormality early warning, and obtains a screening clustering result;

[0009] The optimization strategy design module optimizes the screening clustering result to obtain a final clustering result of the electric water heater sample set;

[0010] The protection control module performs protection control on real-time collected electric water heater operation data based on the final clustering result.

[0011] Further, the data acquisition module synchronously samples historical sensor signals of the electric water heater heating element at a fixed sampling frequency to form a historical time series data set; each dimension is linearly normalized according to a maximum range; the time series data set is divided according to a window length w and a step s to obtain the electric water heater sample set, and an operation mode is assigned to the electric water heater sample set as a label, and the label does not participate in dimension operation.

[0012] Further, the basic clustering generation module runs K-means on the electric water heater sample set m times to obtain m basic clustering results; a co-association matrix is constructed; each basic clustering obtains a co-association matrix, and for the kth basic clustering result, a co-association matrix is constructed , the formula used is: ; wherein, is a matrix element, i and j are matrix row and column indexes; is a label; and are the ith sample and the jth sample in the electric water heater sample set, respectively; a hard similarity label between sample pairs in the electric water heater sample set is constructed.

[0013] Further, the optimal fuzzy matrix learning module specifically includes the following contents:

[0014] An initial fuzzy co-association matrix; the initial fuzzy co-association matrix is obtained based on the co-association matrix according to an average strategy, and the formula used is: ; wherein, is the initial fuzzy co-association matrix;

[0015] Alternating optimization of the overall objective; under the condition of satisfying the fuzzy index r and the rank constraint, is minimized, and the overall objective is represented as: ; ; wherein, is the Laplacian matrix, and control the balance between the difference with the original measurement and the entropy regularization; is the weight of the kth base clustering result in the final fusion; is the element in the current fuzzy co-association matrix S; i and j are the matrix row and column indices; rank(·) is the rank of the matrix; n is the total number of samples; c is the total number of clusters; is the diagonal of

[0016] equivalent transformation; with the help of the Lagrange multiplier , the overall objective is decomposed and combined with the spectral structure matrix F to obtain the equivalent objective, and the formula is: ; ; where F is the spectral structure matrix composed of the smallest c eigenvectors of ; Tr(·) is the trace of the matrix; I is the identity matrix; T is the matrix transpose;

[0017] alternating three-step iteration; repeat the following three steps until convergence, and obtain the fuzzy co-association matrix after convergence, directly extract the c cluster numbers from the connected components of the fuzzy co-association matrix; the cluster number is which cluster the sample belongs to; take the label with the most number in the same cluster as the cluster label; obtain the preliminary clustering result of the electric water heater sample set: (1) update F to extract the low-dimensional spectral embedding of the latest similarity graph, and the formula is: , F takes the smallest c eigenvectors; (2) update S to adjust the similarity of samples in the electric water heater sample set by combining physical distance, spectral distance and basic co-association information; after writing the sub-problem as a quadratic form, independently solve it row by row, which is equivalent to: ; ; where and are the row vectors of F; is the element of the co-association matrix ; (3) update the weight to dynamically allocate the importance of each base clustering according to the current fusion result, and the formula is: ; where t is the index of the base clustering; is the co-association matrix obtained by the tth base clustering.

[0018] Further, the sample screening module screens the preliminary clustering result to obtain a screened clustering result; the screening process is: for the clustering members, calculate the electric water heater sample outlying degree by using the fuzzy matrix element, and the formula is: ; when , set The data marked as abnormal data is deleted, and an abnormal condition early warning detection is performed, a state transition probability matrix P is constructed, a cluster label obtained by each iteration of the electric water heater sample is counted, and a state transition frequency is obtained, and a formula used is: ; if , a rare abnormality alarm is triggered; wherein, is a sample outlying degree, is an outlying threshold value; is the kth cluster; is a state transition probability matrix element; is the number of times that the sample is counted in all iterations from the cluster label u to the cluster label g; is the number of times that the sample is counted in all iterations from the cluster label u to all other cluster labels; is a transition threshold value.

[0019] Further, the optimization strategy design module sets a label ratio B, and for the screening clustering result, the minimum ratio of the label with the largest number in the cluster to all labels in the cluster is D, if D is less than B, it indicates that the current clustering result of the electric water heater sample set is not ideal, and optimization is performed; the optimization content is that based on the running number m, the fuzzy index r, the balance parameter and , the label ratio D, the window length w and the step s establish a search space; a genetic algorithm is used for parameter optimization, D obtained based on the position of the genetic individual is used as an individual fitness value; a crossover operator is changed, a random coefficient and difference information are combined to enhance diversity, binary codes of the parent generation are converted into decimals, and arithmetic crossover and difference mutation are performed, and a formula used for specific operation is: ; ; wherein, and are decimal values of the parent chromosome; and are decimal values of the generated offspring chromosome; ; ; then the offspring is mapped back to the binary code and returned to the genetic algorithm optimization; until D obtained by an individual is not less than B, the final clustering result of the electric water heater sample set corresponding to this time is returned to the protection control module.

[0020] Further, based on the final clustering result, the protection control module collects electric water heater running data in real time, distributes the real-time collected electric water heater running data based on the minimum distance principle, uses the cluster label as the label of the real-time collected electric water heater running data, and predefines the running mode corresponding to each label; according to the label triggered control strategy, the protection control of the electric water heater is realized.

[0021] The present application has the following beneficial effects by using the above scheme:

[0022] (1) For the general electric water heater protection control system, the sensor measurement noise and the running state boundary cause the single division result to fluctuate sharply, and the false judgment rate increases. The scheme reduces the false judgment of the real fault boundary by introducing the joint target of physical space Euclidean distance and entropy regularization; the sensitivity to the real fault state change is improved by introducing the physical distance penalty term, and the false judgment is reduced; the fault tolerance to sensor noise and boundary state is enhanced by fuzzy neighbor, and the risk of false alarm and missed alarm is reduced.

[0023] (2) For the general electric water heater protection control system, the identification of outliers in the electric water heater operation data is improper, it is difficult to find short-term sudden abnormal working conditions, and the electric water heater protection control effect is poor. The scheme combines fuzzy similarity and outlier degree to remove noise and fault points, and performs all-round abnormal filtering; both single-point abnormal alarm and rare working condition based on state transition probability are captured to reduce the risk of equipment failure; random coefficient and difference information are combined to adjust the hyperparameters, so that good protection control effect can be realized under different working conditions. BRIEF DESCRIPTION OF DRAWINGS

[0024] Figure 1 The flowchart of the electric water heater protection control system based on clustering processing provided by the present application is shown.

[0025] Figure 2 The flowchart of the optimal fuzzy matrix learning module is shown.

[0026] The accompanying drawings are used to provide a further understanding of the present application, and constitute a part of the specification, together with the embodiments of the present application, to explain the present application, and do not constitute a limitation of the present application. DETAILED DESCRIPTION

[0027] The technical solutions in the embodiments of the present application will be described clearly and completely below with reference to the drawings in the embodiments of the present application. Obviously, the described embodiments are only a part of the embodiments of the present application, not all the embodiments. Based on the embodiments in the present application, all other embodiments obtained by those skilled in the art without creative labor are within the scope of protection of the present application.

[0028] In the description of the present application, it should be understood that the terms "up", "down", "front", "back", "left", "right", "top", "bottom", "inner", "outer" and the like indicate the orientation or positional relationship based on the orientation or positional relationship shown in the drawings, and are only for the convenience of describing the present application and simplifying the description, and do not indicate or imply that the device or element referred to must have a particular orientation, be constructed and operated in a particular orientation, therefore it cannot be understood as a limitation of the present application.

[0029] Embodiment one, refer toFigure 1 The application provides an electric water heater protection control system based on clustering processing, which comprises a data acquisition module, a basic clustering generation module, an optimal fuzzy matrix learning module, a sample screening module, an optimization strategy design module and a protection control module.

[0030] The data acquisition module acquires electric water heater heating element operation data, and constructs an electric water heater sample set; and sends the data to the basic clustering generation module.

[0031] The basic clustering generation module runs K-means clustering in parallel, generates a basic clustering result for the electric water heater sample set, and constructs a co-association matrix; and sends the data to the optimal fuzzy matrix learning module.

[0032] The optimal fuzzy matrix learning module constructs an initial fuzzy co-association matrix based on the co-association matrix, obtains an optimal fuzzy co-association matrix through alternating optimization and iterative learning, and obtains a preliminary clustering result for the electric water heater sample set; and sends the data to the sample screening module.

[0033] The sample screening module calculates sample outlying degree to remove abnormal points and calculates state transition probability of a clustering label based on the preliminary clustering result of the electric water heater sample set, triggers rare abnormality early warning, and obtains a screening clustering result; and sends the data to the optimization strategy design module.

[0034] The optimization strategy design module optimizes the screening clustering result, and obtains a final clustering result for the electric water heater sample set; and sends the data to the protection control module.

[0035] The protection control module performs protection control on real-time acquired electric water heater operation data based on the final clustering result.

[0036] Embodiment two, refer to Figure 1 The data acquisition module synchronously samples historical sensor signals of temperature, current, voltage, water pressure and flow of the electric water heater heating element at a fixed sampling frequency, and forms a historical time series data set; each dimension is linearly normalized according to the maximum range difference to reduce the influence of dimension; the time series data set is divided according to the window length w and the step s, and an electric water heater sample set is obtained to enhance real-time performance and diversity; and the electric water heater sample set is assigned with an operation mode as a label, and the label does not participate in dimension operation.

[0037] Embodiment three, refer to Figure 1, the embodiment is based on the above embodiment, the basic clustering generation module runs m times K-means on the electric water heater sample set in parallel to obtain m basic clustering results, each basic clustering result represents the division of the samples in the electric water heater sample set belonging to different operating states, the adaptability to the number of faults is enhanced by using a random k value; a co-association matrix is constructed; each basic clustering obtains a co-association matrix, for the kth basic clustering result, a co-association matrix is constructed , the formula used is: ; wherein, is a matrix element, i and j are matrix row and column indices; is a label; and are the ith sample and the jth sample in the electric water heater sample set, respectively; a hard similarity label between the sample pairs in the electric water heater sample set is constructed.

[0038] Embodiment four, refer to Figure 1 and Figure 2 , the embodiment is based on the above embodiment, the optimal fuzzy matrix learning module specifically includes the following contents:

[0039] An initial fuzzy co-association matrix; based on the co-association matrix, an initial fuzzy co-association matrix is obtained according to the average strategy, and the formula used is: ; wherein, is an initial fuzzy co-association matrix; by averaging the basic clustering results, the error influence of single clustering of the electric water heater sample set is preliminarily reduced;

[0040] Alternatingly optimizing the overall objective; under the condition of meeting the fuzzy index r and the rank constraint, the minimum is introduced, the Euclidean distance from the physical space, if two sensor samples have large differences in actual measurement values, they should not be considered as similar states, and the overall objective is represented as: ; ; wherein, is a Laplacian matrix, and control the balance between the difference from the original measurement and the entropy regularization; is the weight of the kth basic clustering result in the final fusion; is an element in the current fuzzy co-association matrix S; i and j are matrix row and column indices; rank(·) is the rank of the matrix; n is the total number of samples; is the diagonal; c is the total number of clusters; is to ensure that the Laplacian matrix has exactly c zero eigenvalues, so as to obtain c connected components corresponding to c clusters; the first term fuses multiple partitions and suppresses single clustering errors; the second term introduces the measurement physical distance, if the sensor measurement values are significantly different, then The third term adaptively adjusts the importance weight; meanwhile, the consistency of multiple electric water heater sample sets, the difference of real measurements and the weight adaptation are considered;

[0041] equivalent conversion; by means of Lagrange multiplier , the overall objective is decomposed and combined with the spectral structure matrix F to obtain an equivalent objective, and the formula used is: ; ; wherein F is a spectral structure matrix composed of the smallest c eigenvectors of ; Tr(·) is the trace of a matrix; I is a unit matrix; T is a matrix transpose;

[0042] alternating three-step iteration; the following three steps are repeated until convergence, and the fuzzy co-association matrix is obtained after convergence, and the cluster number is directly extracted from the connected component of the fuzzy co-association matrix; the cluster number is which cluster the sample belongs to; the label with the most number in the same cluster is taken as the cluster label; the preliminary clustering result of the electric water heater sample set is obtained: (1) update F to extract the low-dimensional spectral embedding of the latest similarity graph, and the formula used is: , F takes the smallest c eigenvectors of ; (2) update S to combine the physical distance, spectral distance and basic co-association information to finely adjust the sample similarity in the electric water heater sample set; after writing the sub-problem as a quadratic form, it is independently solved in a closed form, which is equivalent to: ; ; wherein and are row vectors of F; is an element of the co-association matrix ; (3) update the weight, dynamically allocate the importance of each basic clustering according to the current fusion result, and the formula used is: ; wherein t is the index of the basic clustering; is the co-association matrix obtained by the tth basic clustering; each step of iteration has a closed form and an efficient solution, which is suitable for real-time operation of a microcontroller; the fuzzy neighbor determined by r, the entropy term and the weight enhances the fault tolerance to sensor noise and boundary state; the rank constraint ensures that exactly c stable states are divided, avoiding false reporting of redundant categories.

[0043] By performing the above operation, for the problem that the single division result fluctuates sharply and the false positive rate increases due to sensor measurement noise and running state boundary of the general electric water heater protection control system, the scheme reduces the false positive rate of the real fault boundary by introducing a joint objective of physical space Euclidean distance and entropy regularization; the physical distance penalty term is introduced to improve the sensitivity to real fault state changes and reduce false positives; the fuzzy neighbor enhances the fault tolerance to sensor noise and boundary state, reducing the risk of false positives and false negatives.

[0044] Embodiment five, refer to Figure 1 , based on the above embodiment, the sample screening module screens the preliminary clustering results to obtain screened clustering results; the screening process is: for the clustering members, the fuzzy matrix element is used to calculate the electric water heater sample outlying degree, and the formula used is: ; when , the abnormal data is marked for deletion; and abnormal condition early warning detection is performed, and a state transition probability matrix P is constructed, the state transition frequency of the cluster label obtained by each iteration of the electric water heater sample is counted, and the formula used is: ; if , a rare abnormality alarm is triggered; wherein, is the sample outlying degree, is the outlying threshold; is the kth cluster; is the state transition probability matrix element; is the number of times that the sample is counted in all iterations from the cluster label u to the cluster label g; is the number of times that the sample is counted in all iterations from the cluster label u to all other cluster labels; is the transition threshold.

[0045] Embodiment six, refer to Figure 1 , based on the above embodiment, the optimization strategy design module sets the label ratio B, and sets that for the screened clustering results, the minimum ratio of the label with the most number in the cluster to all labels in the cluster is D, if D is less than B, it indicates that the current clustering result of the electric water heater sample set is not ideal, and optimization is performed; the optimization content: based on the running number m, the fuzzy index r, the balance parameter and , the label ratio D, the window length w and the step size s establish a search space; genetic algorithm is used for parameter optimization, D obtained based on the position of the genetic individual is used as the individual fitness value; the crossover operator is changed, the random coefficient and the difference information are combined to enhance the diversity, the binary code of the parent generation is converted to decimal, and arithmetic crossover and difference mutation are performed, which can introduce more exploration on the basis of preserving the parent information, thereby more effectively searching the solution space, and the specific operation formula is: ; ; wherein, and are the decimal values of the parent chromosomes; and are the decimal values of the generated offspring chromosomes; ; ; Then the offspring is mapped back to the binary encoding returned to the genetic algorithm optimization; until there is an individual D obtained is not less than B, the final clustering result of the corresponding water heater sample set at this time is returned to the protection control module.

[0046] By performing the above operation, for the general electric water heater protection control system, there is an improper identification of outlier data in the electric water heater operation data, it is difficult to find short-term sudden abnormal working conditions, resulting in poor electric water heater protection control effect, the scheme combines fuzzy similarity and outlier degree to remove noise and fault points, and performs all-around abnormal filtering; both single-point abnormal alarm and rare working condition capture based on state transition probability to reduce the risk of equipment failure; adjust the hyperparameters combined with random coefficients and difference information to ensure good protection control effect under different working conditions.

[0047] Embodiment seven, see Figure 1 Based on the above embodiment, the protection control module based on the final clustering result, real-time acquisition of electric water heater operation data, based on the minimum distance principle, the real-time acquisition of electric water heater operation data is distributed, the cluster label is used as the label of the real-time acquisition of electric water heater operation data, and the running mode corresponding to each label is defined in advance; according to the label, the control strategy is triggered to realize the protection control of electric water heater; the clustering result is directly mapped to the control strategy; the running mode corresponding to the clustering label is represented as: ; The execution signal is: normal heating (l=1), maintain heating operation; heating warning (l=3), buzzer prompt, UI display current overload risk; standby (l=2), turn off the main heating element, low-power maintenance; leakage fault (l=4), immediately cut off the heating circuit power, open the safety valve, buzzer alarm and remote notification; dry burning protection (l=5), turn off the heating element, detect the water flow and perform necessary cooling treatment.

[0048] Although embodiments of the present application have been shown and described, it will be understood by those having ordinary skill in the art that various changes, modifications, substitutions and alterations can be made thereto without departing from the principles and spirit of the present application.

[0049] The above describes the present application and its embodiments, which are not restrictive, and the drawings shown are only one of the embodiments of the present application, and the actual structure is not limited thereto. In general, if a person skilled in the art is inspired, without departing from the purpose of the present application, without creative design, similar structure and embodiments of the technical solution can be designed, which shall belong to the protection scope of the present application.

Claims

1. A protection and control system for electric water heaters based on clustering processing, characterized in that: The system includes a data acquisition module, a basic clustering generation module, an optimal fuzzy matrix learning module, a sample selection module, an optimization strategy design module, and a protection and control module. The data acquisition module collects operating data of the heating element of the electric water heater and constructs a sample set of electric water heaters. The basic clustering generation module runs K-means clustering in parallel to generate basic clustering results for the electric water heater sample set and constructs a co-association matrix; The optimal fuzzy matrix learning module constructs an initial fuzzy co-correlation matrix based on the co-correlation matrix, and obtains the optimal fuzzy co-correlation matrix through alternating optimization and iterative learning, thus obtaining the preliminary clustering results of the electric water heater sample set; The sample screening module calculates the outlier degree of the samples based on the preliminary clustering results of the electric water heater sample set, removes outliers, and counts the state transition probability of the cluster labels to trigger a rare anomaly warning and obtain the screening clustering results. The optimization strategy design module optimizes the screening clustering results to obtain the final clustering results of the electric water heater sample set; The protection and control module performs protection and control on the real-time collected electric water heater operation data based on the final clustering results; The sample screening module filters the preliminary clustering results to obtain the filtered clustering results. The screening process is as follows: for cluster members, the outlier degree of the electric water heater samples is calculated using fuzzy matrix elements, and the formula used is: ;when At that time, Mark the data as abnormal and delete it. Abnormal operating condition early warning detection is performed, and a state transition probability matrix P is constructed. For the cluster labels obtained in each iteration of the electric water heater sample, the state transition frequency is statistically analyzed using the following formula: ;like This triggered a rare anomaly alarm; among them, It is the outlier rate of the sample. It is the outlier threshold; It is the k-th cluster; These are elements of the state transition probability matrix; It represents the number of times a sample moves from cluster label u to cluster label g across all iterations; It represents the number of times a statistical sample moves from cluster label u to all other cluster labels throughout all iterations; It is the transition threshold; It is the i-th sample in the electric water heater sample set; The optimization strategy design module sets a label ratio B. Let D be the minimum ratio of the number of labels with the highest number in a cluster to all labels in that cluster for filtering clustering results. If D is less than B, it indicates that the current clustering result for the electric water heater sample set is not ideal, and optimization is required. The optimization content includes: based on the number of runs m, the fuzzy index r, and the balance parameter. and The search space is established using the label ratio B, window length w, and step size s; a genetic algorithm is used for parameter optimization, with D obtained based on the genetic individual's position as the individual's fitness value; the crossover operator is modified, combining random coefficients and differential information to enhance diversity, converting the parent's binary encoding to decimal, and performing arithmetic crossover and differential mutation. The specific formula used in this operation is as follows: ; ;in, and It is the decimal value of the parent's chromosome; and It is the decimal value of the chromosomes of the generated offspring; ; Then the offspring are mapped back to binary encoding and returned to the genetic algorithm for optimization; until there is an individual whose D is not less than B, the final clustering result of the corresponding electric water heater sample set is returned to the protection control module.

2. The electric water heater protection control system based on clustering processing according to claim 1, characterized in that: The optimal fuzzy matrix learning module specifically includes the following: Initial fuzzy co-correlation matrix; Obtain the initial fuzzy co-correlation matrix based on the co-correlation matrix using an averaging strategy; Alternately optimize the overall objective; under the conditions of satisfying the fuzzy exponent r and rank constraints; Equivalent transformation; using Lagrange multipliers The overall objective is decomposed and combined with the spectral structure matrix F to obtain the equivalent objective; Alternating three-step iteration.

3. The electric water heater protection control system based on clustering processing according to claim 2, characterized in that: The alternating three-step iteration is to repeat the following three steps until convergence. After convergence, a fuzzy co-correlation matrix is ​​obtained. c cluster numbers are directly extracted from the connected components of the fuzzy co-correlation matrix. The label with the most number of labels in the same cluster is taken as the cluster label. The preliminary clustering results of the electric water heater sample set are obtained: First step, update F and extract the low-dimensional spectral embedding of the latest similarity map. F is the spectral structure matrix; the second step is to update S, and adjust the sample similarity in the electric water heater sample set by combining physical distance, spectral distance and basic co-correlation information; S is the fuzzy co-correlation matrix; after writing the subproblem in quadratic form, solve it independently row by row in closed form; the third step is to update the weights.

4. The electric water heater protection control system based on clustering processing according to claim 3, characterized in that: The basic clustering generation module runs K-means m times in parallel on the electric water heater sample set to obtain m basic clustering results; Construct the co-association matrix; each basic cluster yields a co-association matrix, and for the k-th basic cluster result, construct the co-association matrix. .

5. The electric water heater protection control system based on clustering processing according to claim 4, characterized in that: The data acquisition module synchronously samples the historical sensor signals of the heating element of the electric water heater at a fixed sampling frequency to form a historical time series dataset; it linearly normalizes each dimension according to the maximum range; it divides the time series dataset according to the window length w and the step size s to obtain the electric water heater sample set, and assigns the operating mode as a label to the electric water heater sample set. The label does not participate in the dimension calculation.

6. The electric water heater protection control system based on clustering processing according to claim 5, characterized in that: The protection and control module collects real-time operating data of the electric water heater based on the final clustering result, allocates the real-time operating data of the electric water heater based on the minimum distance principle, uses cluster labels as labels for the real-time operating data of the electric water heater, and predefines the operating mode corresponding to each label; and realizes the protection and control of the electric water heater according to the label trigger control strategy.

Images