CN109309382B - Short-term power load prediction method - Google Patents
Short-term power load prediction method Download PDFInfo
- Publication number
- CN109309382B CN109309382B CN201811068735.2A CN201811068735A CN109309382B CN 109309382 B CN109309382 B CN 109309382B CN 201811068735 A CN201811068735 A CN 201811068735A CN 109309382 B CN109309382 B CN 109309382B
- Authority
- CN
- China
- Prior art keywords
- power load
- neural network
- short
- stage
- prediction model
- 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.)
- Active
Links
Images
Classifications
-
- H—ELECTRICITY
- H02—GENERATION; CONVERSION OR DISTRIBUTION OF ELECTRIC POWER
- H02J—CIRCUIT ARRANGEMENTS OR SYSTEMS FOR SUPPLYING OR DISTRIBUTING ELECTRIC POWER; SYSTEMS FOR STORING ELECTRIC ENERGY
- H02J3/00—Circuit arrangements for ac mains or ac distribution networks
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/04—Architecture, e.g. interconnection topology
- G06N3/045—Combinations of networks
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/08—Learning methods
- G06N3/084—Backpropagation, e.g. using gradient descent
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
- G06Q10/00—Administration; Management
- G06Q10/04—Forecasting or optimisation specially adapted for administrative or management purposes, e.g. linear programming or "cutting stock problem"
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
- G06Q50/00—Systems or methods specially adapted for specific business sectors, e.g. utilities or tourism
- G06Q50/06—Electricity, gas or water supply
-
- H—ELECTRICITY
- H02—GENERATION; CONVERSION OR DISTRIBUTION OF ELECTRIC POWER
- H02J—CIRCUIT ARRANGEMENTS OR SYSTEMS FOR SUPPLYING OR DISTRIBUTING ELECTRIC POWER; SYSTEMS FOR STORING ELECTRIC ENERGY
- H02J2203/00—Indexing scheme relating to details of circuit arrangements for AC mains or AC distribution networks
- H02J2203/20—Simulating, e g planning, reliability check, modelling or computer assisted design [CAD]
-
- H—ELECTRICITY
- H02—GENERATION; CONVERSION OR DISTRIBUTION OF ELECTRIC POWER
- H02J—CIRCUIT ARRANGEMENTS OR SYSTEMS FOR SUPPLYING OR DISTRIBUTING ELECTRIC POWER; SYSTEMS FOR STORING ELECTRIC ENERGY
- H02J3/00—Circuit arrangements for ac mains or ac distribution networks
- H02J3/003—Load forecast, e.g. methods or systems for forecasting future load demand
-
- Y—GENERAL TAGGING OF NEW TECHNOLOGICAL DEVELOPMENTS; GENERAL TAGGING OF CROSS-SECTIONAL TECHNOLOGIES SPANNING OVER SEVERAL SECTIONS OF THE IPC; TECHNICAL SUBJECTS COVERED BY FORMER USPC CROSS-REFERENCE ART COLLECTIONS [XRACs] AND DIGESTS
- Y04—INFORMATION OR COMMUNICATION TECHNOLOGIES HAVING AN IMPACT ON OTHER TECHNOLOGY AREAS
- Y04S—SYSTEMS INTEGRATING TECHNOLOGIES RELATED TO POWER NETWORK OPERATION, COMMUNICATION OR INFORMATION TECHNOLOGIES FOR IMPROVING THE ELECTRICAL POWER GENERATION, TRANSMISSION, DISTRIBUTION, MANAGEMENT OR USAGE, i.e. SMART GRIDS
- Y04S10/00—Systems supporting electrical power generation, transmission or distribution
- Y04S10/50—Systems or methods supporting the power network operation or management, involving a certain degree of interaction with the load-side end user applications
Abstract
The invention provides a short-term power load forecasting method, which comprises the steps of establishing a short-term power load forecasting model based on a front-stage and back-stage BP cascade neural network, clustering historical power load data by adopting an improved maximum deviation similarity criterion clustering algorithm, carrying out hierarchical clustering on class centers, obtaining similar load classes of a forecasting day through class cluster combination, and taking load data of a Z class as training data of the back-stage BP neural network, thereby improving the convergence speed and the forecasting precision of the neural network. The short-term power load prediction method provided by the invention has the advantages of high calculation speed and high prediction precision, and can better meet the demand of a power grid on power load prediction.
Description
Technical Field
The invention relates to the field of big data, in particular to a short-term power load prediction method.
Background
In the big data era, by researching a plurality of factors such as the operating characteristics, the power load data, the natural conditions and the social influence of the power system, under the condition of meeting a certain precision requirement, the power load data curve of one or more days in the future is predicted, which is an important content in the economic dispatching of the power system. The traditional BP neural network has the defects of long neural network training time, easiness in falling into local optimization, low precision, incapability of predicting the power utilization rule of a user and the like when the power load is predicted, so that the key of the power load prediction problem is how to reduce the neural network training time and improve the power load prediction precision.
Disclosure of Invention
The invention overcomes the defects that the neural network training time is long, the neural network training is easy to fall into local optimization, the precision is low, the power utilization rule of a user cannot be predicted and the like in the prior art, and provides the short-term power load prediction method.
The technical scheme of the invention is summarized as follows:
a short term power load forecasting method comprising the steps of:
s1: determining an input variable, an output variable and a training mode of a preceding-stage BP neural network;
s2: determining a prediction day, and collecting historical power load data of N consecutive days before the prediction day;
s3: establishing a preceding stage BP neural network short-term power load prediction model, training the preceding stage BP neural network short-term power load prediction model by using historical power load data, and recording the output of the preceding stage BP neural network short-term power load prediction model to obtain a power load characteristic vector V of a prediction day;
s4: clustering analysis is carried out on historical power load data by using an improved maximum deviation similarity criterion clustering algorithm to obtain a clustering result C1;
S5: clustering result C by using hierarchical clustering algorithm1Class center X ofiClustering to obtain a clustering result C2;
S6: clustering result C2In the same class XiCorresponding C1The cluster in (1) is merged to obtain a clustering result C3And calculate C3Class center X ofi′;
S7: calculating a similar day category Z of a power load curve of a predicted day;
s8: establishing a short-term power load prediction model of a back-stage BP neural network;
s9: normalizing the Z-th class power load data obtained in the step S7, taking the normalized Z-th class power load data as input, and training a rear-stage BP neural network short-term power load prediction model;
s10: calculating a total training error epsilon of the BP neural network;
s11: if ε < ε0Or when the training frequency of the back-stage BP neural network reaches the set maximum training frequency, switching to S11, otherwise, switching to S9;
s12: predicting the power load data of the predicted day by using the trained back-stage BP neural network, and outputting a power load prediction curve of the predicted day;
s13: and (6) ending.
Further, step S1 specifically includes the following steps:
s11: determining the input variables of the input layer nodes of the preceding stage BP neural network prediction modelPredicting the related power load data of N consecutive days;
s12: determining the output variable of the output layer node of the preceding stage BP neural network prediction modelIs the relevant power load data for the predicted day;
s13: determining a training mode of the preceding stage BP neural network prediction model: and taking the related power load data of the previous day in the N consecutive days of the prediction day as input, and taking the related power load data of the next day as output until all the training samples of the previous N days are trained.
Further, the established early-stage BP neural network short-term power load prediction model in step S3 is as follows:
wherein i is 1,21,j=1,2,...,m1,k=1,2,...,n1,l=1,2,...,M1,n1The number of nodes of the input layer, m, of the short-term power load prediction model of the preceding BP neural network1Is the node number, M, of the output layer of the short-term power load prediction model of the preceding-stage BP neural network1The number of nodes of a hidden layer of a short-term power load prediction model of a preceding BP neural network; y iskjIs the output corresponding to the kth input sample of the preceding stage BP neural network short-term power load prediction model, wijIs the weight from the hidden layer to the output layer of the short-term power load prediction model of the preceding BP neural network, bijIs the bias from the hidden layer of the short-term power load prediction model of the preceding BP neural network to the output layer, and delta is the neuron excitation from the hidden layer of the short-term power load prediction model of the preceding BP neural network to the output layerThe activity function, δ, is expressed as the Relu function as follows:
δ(x)=max(λx,0)
where λ is the linear factor of the Relu function.
alThe input value of the I-th hidden layer node of the short-term power load prediction model of the preceding-stage BP neural network is expressed as follows:
wherein i is 1,21,k=1,2,...,n1,l=1,2,...,M1,wkiIs the connection weight value, x, from the input layer to the output layer of the short-term power load prediction model of the preceding BP neural networkkIs an input variable of a short-term power load prediction model of a preceding-stage BP neural network, bkiThe method is the bias from an input layer to a hidden layer of a short-term power load prediction model of a preceding-stage BP neural network, wherein i is 1,21,k=1,2,...,n1,l=1,2,...,M1,n1Is the number of nodes of the input layer, M, of the short-term power load prediction model of the preceding-stage BP neural network1The number of nodes of the hidden layer of the short-term power load prediction model of the preceding-stage BP neural network is determined.
Further, the components of the power load feature vector V on the predicted day described in step S3 are derived from the output y of the preceding BP neural network, that is:
wherein the content of the first and second substances,is a predicted value of the power load data on the predicted day.
Further, step S4 specifically includes the following steps:
s41: drawing an input power load curve of a maximum deviation similarity criterion clustering algorithm, and setting the ith power load data as xi=(xi1,xi2,...,xik,...,xip),i=1,2,...,n11,2, p, where x isikIs the power load value at the kth time point of the ith power load data;
s42: setting control parameters of an improved maximum deviation similarity criterion clustering algorithm, wherein the control parameters comprise a maximum deviation gamma, a maximum deviation point allowable deviation delta, a lowest similarity factor alpha and a continuous deviation factor beta;
s43: calculating two power load curves xiAnd xjAbsolute difference s at the kth time pointijkThe calculation formula is as follows:
sijk=|xik-xjk|
wherein x isikIs the power load curve xiPower load value, x, at the k-th point in timejkPower load curve xjA power load value at a kth time point;
s44: calculating xiAnd xjNumber of similarity points nijAnd the maximum number of continuous deviation points mij,nijIs to satisfy sijkS is less than or equal to gammaijkNumber of (1), mijIs that gamma is less than sijkS of < δijkWherein, i, j is 1,21,k=1,2,...,m1,mijThe expression is as follows:
wherein m isijIs the maximum number of consecutive deviation points,is the power load xiAnd xjAt the k-th0The absolute difference of the individual points in time,is the power load xiAnd xjAt the k-th0The absolute difference of +1 time points,is the power load xiAnd xjAt the k-th0(ii) the absolute difference of + s-1 time points, γ is the maximum deviation, δ is the maximum deviation point allowed deviation;
s45: judging x according to the maximum deviation similarity criterioniAnd xjWhether the two are similar or not is judged according to the following steps:
n0≤nij≤m0
wherein n isijIs the power load curve xiAnd xjNumber of similarity points, n0=[α×m],m0=[β×m]Alpha is a similarity factor, beta is a continuous deviation factor, and alpha is more than or equal to 0 and less than or equal to 1, and beta is more than or equal to 0 and less than or equal to 1-alpha;
if the above formula is true, then xiAnd xjSimilarly, if the above formula is false, then xiAnd xjAre not similar;
s46: calculating xiAnd xjTotal distance d (x) ofi,xj) The calculation formula is as follows:
wherein s isijkIs the power load xiAnd xjAbsolute difference at the kth time point;
s47: calculating xiS (x) ofi) Obtaining a clustering result C1The method comprises the following specific steps: for all i, j ═ 1,21And i is less than or equal to j, with xiAs a comparison center, nij、mijAre each independently of n0、m0Comparing, all x satisfying the maximum deviation similarity criterioniIs divided into s (x)i) In, let s (x)i)=s(xi)∪{xjAnd x isjRemoving from the original power load data set U;
s48: calculating xiS (x) ofi) Class center X ofiCalculatingThe formula is as follows:
further, step S5 specifically includes the following steps:
s51: initializing a hierarchical clustering algorithm, and clustering a result C1Class center X ofiClassifying the cluster into a cluster R;
s52: calculation of X in RiAnd XjTwo elements X corresponding to the maximum value of the distance of (1)aAnd XbIs mixing XaAnd XbInto two different clusters R1And R2Wherein the distance calculation formula is as follows:
s53: calculating other elements X in the original cluster RiTo XaAnd XbEuclidean distance d (X)i,Xa) And d (X)i,Xb) If d (X)i,Xa)<d(Xi,Xb) Then X will beiFall under R1In, otherwise, fall under R2;
S54: for class cluster R1And R2Steps S52 and S53 are performed until all the class clusters R are finally dividediMiddle two elements XiAnd XjIs less than RiAverage distance M ofiλ times of (i), i.e.:
maxd(Xi,Xj)<λMi
wherein M isiThe calculation formula of (2) is as follows:
wherein R is a cluster RiNumber of elements of (c)iIs a cluster-like RiClass center of (1), XjIs a cluster-like RiAn element of (1);
stopping the hierarchical clustering algorithm to obtain a clustering result C2The number of clusters is K.
Further, step S6 specifically includes the following steps:
s61: clustering result C2In the same class XiCorresponding C1Merging the clusters in (1), i.e. merging the elements in the clusters into one cluster to obtain a clustering result C3;
S62: calculating a clustering result C3Class center X ofi′。
Further, step S7 specifically includes the following steps:
s71: initializing a hierarchical clustering algorithm, and clustering a result C3Quasi center X 'of'iAnd the power load characteristic vector V of the forecast day is classified into a cluster R;
s72: calculating the element X in RiAnd XjTwo elements X corresponding to the maximum value of the distance of (1)aAnd XbIs mixing XaAnd XbInto two different clusters R1And R2Wherein the distance calculation formula is as follows:
s73: calculating other elements X in the original cluster RiTo XaAnd XbEuclidean distance d (X)i,Xa) And d (X)i,Xb) If d (X)i,Xa)<d(Xi,Xb) Then X will beiFall under R1In, otherwise, fall under R2;
S74: for class cluster R1And R2Executing the steps S52 and S53 until the K clusters are finally divided, stopping the hierarchical clustering algorithm to obtain a clustering result C4;
S75: clustering result C4The power load characteristic vector V of the predicted day and the power load characteristic vector V of the predicted day are class centers X 'of the same class cluster'iCorresponding power load data typeThe power load curve for the predicted day resembles the day class Z.
Further, the short-term power load prediction model of the back-stage BP neural network described in step S8 is as follows:
wherein i is 1,22,j=1,2,...,m2,k=1,2,...,n2,l=1,2,3...M2,n2The number of nodes of the input layer of the short-term power load prediction model of the back-stage BP neural network, m2Is the number of nodes of the output layer of the short-term power load prediction model of the back-stage BP neural network, M2The number of nodes of a hidden layer of a short-term power load prediction model of a back-stage BP neural network, ykjIs the output corresponding to the kth input sample, wijIs the weight from the hidden layer to the output layer of the short-term power load prediction model of the back-stage BP neural network, bijThe bias from a hidden layer of a short-term power load prediction model of the back-stage BP neural network to an output layer is adopted, delta is a neuron activation function from the hidden layer of the short-term power load prediction model of the back-stage BP neural network to the output layer, and delta adopts a tanh function and is expressed as follows:
alis the input value of the l-th hidden layer node of the short-term power load prediction model of the back-stage BP neural network, alThe calculation formula of (a) is as follows:
wherein, l ═ 1,2,32,k=1,2,...,n2,i=1,2...M2,wkiIs the connection weight value, x, from the input layer to the output layer of the short-term power load prediction model of the back-stage BP neural networkkIs input of a short-term power load prediction model of a back-stage BP neural networkVariable bkiAnd predicting the bias from the input layer to the hidden layer of the model for the short-term power load of the back-stage BP neural network.
Further, the normalization formula described in step S9 is as follows:
wherein x isikIs the power load value, x, at the kth time point of the power load curve on the ith daymaxAnd xminRespectively, the maximum and minimum values of the class Z power load data.
Compared with the prior art, the invention has the beneficial effects that: the invention combines the improved maximum deviation similarity criterion algorithm with the BP neural network, and takes the load data of the Z type as the training data of the BP neural network, thereby accelerating the training speed of the neural network, improving the prediction precision, and overcoming the defects of long training time, easy falling into local optimum, lower precision and the like of the BP neural network. The short-term power load prediction method provided by the invention has the advantages of high calculation speed and good prediction effect, and can better meet the demand of power load prediction of a power grid company.
Drawings
FIG. 1 is a flow chart of the present invention;
fig. 2 shows a result of power load prediction according to an embodiment of the present invention.
Detailed Description
The drawings are for illustrative purposes only and are not to be construed as limiting the patent;
for the purpose of better illustrating the embodiments, certain features of the drawings may be omitted, enlarged or reduced, and do not represent the size of an actual product;
it will be understood by those skilled in the art that certain well-known structures in the drawings and descriptions thereof may be omitted.
The invention is further described below with reference to the accompanying drawings:
referring to fig. 1, a short-term power load prediction method of the present invention includes the steps of:
s1, determining input variables, output variables and training modes of the preceding-stage BP neural network, specifically as follows:
s11: determining the input variables of the input layer nodes of the preceding stage BP neural network prediction modelIs the related power load data of N days before the prediction day (N is a preset positive integer, N is more than or equal to 30 and less than or equal to 1000), and four related data are selected, namely N1The specific formula is 4 as follows: x is the number of1Is the daily average power load value, x2Is the daily maximum power load value, x3Is the daily minimum power load value, x4Is a daily electric quantity value;
s12: determining the output variable of the output layer node of the preceding stage BP neural network prediction modelIs the relevant power load data of the predicted day, four relevant data are selected, namely m1The specific formula is 4 as follows: y is1Is the daily average power load predicted value, y2Is the daily maximum power load predicted value, y3Is the daily minimum power load predicted value, y4Is a daily electricity consumption predicted value;
s13: determining a training mode of the preceding stage BP neural network prediction model: and taking the related power load data of the previous day in the N consecutive days of the prediction day as input, and taking the related power load data of the next day as output until all the training samples of the previous N days are trained.
S2: and determining a prediction day, and collecting historical power load data of N days before the prediction day, including a daily average power load prediction value, a daily maximum power load prediction value, a daily minimum power load prediction value and a daily electric quantity prediction value.
S3: establishing a short-term power load prediction model of a preceding-stage BP neural network, as follows:
wherein i is 1,21,j=1,2,...,m1,k=1,2,...,n1,l=1,2,...,M1,n1The number of nodes of the input layer, m, of the short-term power load prediction model of the preceding BP neural network1Is the node number, M, of the output layer of the short-term power load prediction model of the preceding-stage BP neural network1The number of nodes of a hidden layer of a short-term power load prediction model of a preceding BP neural network; y iskjIs the output corresponding to the kth input sample of the preceding stage BP neural network short-term power load prediction model, wijIs the weight from the hidden layer to the output layer of the short-term power load prediction model of the preceding BP neural network, bijThe method is characterized in that the bias from a hidden layer of a short-term power load prediction model of a preceding-stage BP neural network to an output layer is adopted, delta is a neuron activation function from the hidden layer of the short-term power load prediction model of the preceding-stage BP neural network to the output layer, and delta adopts a Relu function and is expressed as follows:
δ(x)=max(λx,0)
wherein λ is a linear factor of the Relu function;
alis the input value of the first hidden layer node of the short-term power load prediction model of the preceding BP neural network, alThe calculation formula of (a) is as follows:
wherein k is 1,21,i=1,2...M1,l=1,2,...,M1,wkiIs the connection weight value, x, from the input layer to the output layer of the short-term power load prediction model of the preceding BP neural networkkIs an input variable of a short-term power load prediction model of a preceding-stage BP neural network, bkiThe bias from the input layer of the short-term power load prediction model of the preceding-stage BP neural network to the hidden layer is adopted.
S4: training a preceding stage BP neural network short-term power load prediction model by using historical power load data, recording the output of the preceding stage BP neural network short-term power load prediction model, and obtaining a power load level characteristic vector V of a prediction day, wherein the component of the V is derived from the output y of the preceding stage BP neural network, namely:
wherein the content of the first and second substances,is a predicted value of the power load data on the predicted day, and four pieces of the data, m, are selected1The specific formula is 4 as follows: v. of1' is a daily average power load prediction value of a prediction day, v2' is a daily maximum power load prediction value of a predicted day, v3' is a daily minimum power load prediction value of a prediction day, v4' is a daily electricity consumption prediction value of the prediction day.
S5: clustering analysis is carried out on historical power load data by using an improved maximum deviation similarity criterion clustering algorithm to obtain a clustering result C1The method comprises the following specific steps:
s51: drawing an input power load curve of a maximum deviation similarity criterion clustering algorithm, and setting the ith power load data as xi=(xi1,xi2,...,xik,...,xip),i=1,2,...,n11,2, p, where x isikIs the power load value at the kth time point of the ith power load data;
s52: setting control parameters of an improved maximum deviation similarity criterion clustering algorithm, wherein the control parameters comprise a maximum deviation gamma, a maximum deviation point allowable deviation delta, a lowest similarity factor alpha and a continuous deviation factor beta;
s53: calculating two power load curves xiAnd xjAbsolute difference s at the kth time pointijkThe calculation formula is as follows:
sijk=|xik-xjk|
wherein x isikIs the power load curve xiPower load value, x, at the k-th point in timejkPower load curve xjElectric load value at k-th time point;
S54: calculating xiAnd xjNumber of similarity points nijAnd the maximum number of continuous deviation points mij,nijIs to satisfy sijkS is less than or equal to gammaijkNumber of (1), mijIs that gamma is less than sijkS of < δijkWherein, i, j is 1,21,k=1,2,...,m1,mijThe expression is as follows:
wherein m isijIs the maximum number of consecutive deviation points,is the power load xiAnd xjAt the k-th0The absolute difference of the individual points in time,is the power load xiAnd xjAt the k-th0The absolute difference of +1 time points,is the power load xiAnd xjAt the k-th0(ii) the absolute difference of + s-1 time points, γ is the maximum deviation, δ is the maximum deviation point allowed deviation;
s55: judging x according to the maximum deviation similarity criterioniAnd xjWhether the two are similar or not is judged according to the following steps:
n0≤nij≤m0
wherein n isijIs the power load curve xiAnd xjNumber of similarity points, n0=[α×m],m0=[β×m]Alpha is a similarity factor, beta is a continuous deviation factor, and alpha is more than or equal to 0 and less than or equal to 1, and beta is more than or equal to 0 and less than or equal to 1-alpha;
if the above formula is true, then xiAnd xjSimilarly, if the above formula is false, then xiAnd xjAre not similar;
s56: calculating xiAnd xjTotal distance d (x) ofi,xj) The calculation formula is as follows:
wherein s isijkIs the power load xiAnd xjThe absolute difference at the kth time point, i, j ═ 1,2.. n1;
S57: calculating xiS (x) ofi) Obtaining a clustering result C1The method comprises the following specific steps: for all i, j ═ 1,21And i is less than or equal to j, with xiAs a comparison center, nij、mijAre each independently of n0、m0Comparing, all x satisfying the maximum deviation similarity criterioniIs divided into s (x)i) In, let s (x)i)=s(xi)∪{xjAnd x isjRemoving from the original power load data set U;
s58: calculating xiS (x) ofi) Class center X ofiThe calculation formula is as follows:
s6: clustering result C by using hierarchical clustering algorithm1Class center X ofiClustering to obtain a clustering result C2The method comprises the following specific steps;
s61: initializing a hierarchical clustering algorithm, and clustering a result C1Class center X ofiClassifying the cluster into a cluster R;
s62: calculation of X in RiAnd XjTwo elements X corresponding to the maximum value of the distance of (1)aAnd XbIs mixing XaAnd XbInto two different clusters R1And R2Wherein the distance calculation formula is as follows:
s63: calculating other elements X in the original cluster RiTo XaAnd XbEuclidean distance d (X)i,Xa) And d (X)i,Xb) If d (X)i,Xa)<d(Xi,Xb) Then X will beiFall under R1In, otherwise, fall under R2;
S64: for class cluster R1And R2Steps S52 and S53 are performed until all the class clusters R are finally dividediMiddle two elements XiAnd XjIs less than RiAverage distance M ofiλ times of (i), i.e.:
maxd(Xi,Xj)<λMi
wherein M isiThe calculation formula of (2) is as follows:
wherein R is a cluster RiNumber of elements of (c)iIs a cluster-like RiClass center of (1), XjIs a cluster-like RiAn element of (1);
stopping the hierarchical clustering algorithm to obtain a clustering result C2The number of clusters is K.
S7: clustering result C2In the same class XiCorresponding C1The cluster in (1) is merged to obtain a clustering result C3And calculate C3Class center X ofi' the concrete steps are as follows:
s71: clustering result C2In the same class XiCorresponding C1Merging the clusters in (1), i.e. merging the elements in the clusters into one cluster to obtain a clustering result C3;
S72: calculating a clustering result C3Class center X ofi′。
S8: calculating a similar day category Z of a power load curve of a predicted day;
s81: initializing a hierarchical clustering algorithm, and clustering a result C3Quasi center X 'of'iAnd the power load characteristic vector V of the forecast day is classified into a cluster R;
s82: calculating the element X in RiAnd XjTwo elements X corresponding to the maximum value of the distance of (1)aAnd XbIs mixing XaAnd XbInto two different clusters R1And R2Wherein the distance calculation formula is as follows:
s83: calculating other elements X in the original cluster RiTo XaAnd XbEuclidean distance d (X)i,Xa) And d (X)i,Xb) If d (X)i,Xa)<d(Xi,Xb) Then X will beiFall under R1In, otherwise, fall under R2;
S84: for class cluster R1And R2Executing the steps S52 and S53 until the K clusters are finally divided, stopping the hierarchical clustering algorithm to obtain a clustering result C4;
S85: clustering result C4The power load characteristic vector V of the middle and predicted days is a class center X of the same class clusteriThe corresponding power load data category is the power load curve similar day category Z of the prediction day.
S9: establishing a short-term power load prediction model of a back-stage BP neural network, wherein the model comprises the following steps:
wherein i is 1,22,j=1,2,...,m2,k=1,2,...,n2,l=1,2,3...M2,n2Is a back-stage BP neural network short-term power loadNumber of nodes of input layer of load prediction model, m2Is the number of nodes of the output layer of the short-term power load prediction model of the back-stage BP neural network, M2The number of nodes of a hidden layer of a short-term power load prediction model of a back-stage BP neural network, ykjIs the output corresponding to the kth input sample, wijIs the weight from the hidden layer to the output layer of the short-term power load prediction model of the back-stage BP neural network, bijThe bias from a hidden layer of a short-term power load prediction model of the back-stage BP neural network to an output layer is adopted, delta is a neuron activation function from the hidden layer of the short-term power load prediction model of the back-stage BP neural network to the output layer, and delta adopts a tanh function and is expressed as follows:
alis the input value of the l-th hidden layer node of the short-term power load prediction model of the back-stage BP neural network, alThe calculation formula of (a) is as follows:
wherein, l ═ 1,2,32,k=1,2,...,n2,i=1,2...M2,wkiIs the connection weight value, x, from the input layer to the output layer of the short-term power load prediction model of the back-stage BP neural networkkIs an input variable of a short-term power load prediction model of a back-stage BP neural network, bkiAnd predicting the bias from the input layer to the hidden layer of the model for the short-term power load of the back-stage BP neural network.
S10: normalizing the Z-th class power load data obtained in the step S8 to be input, training a short-term power load prediction model of the back-stage BP neural network, wherein the normalization formula is as follows:
wherein x isikIs the power load of the ith dayPower load value, x, at the k-th time point of the curvemaxAnd xminThe maximum value and the minimum value of the Z-th class power load data are respectively;
s11: calculating a total training error epsilon of the BP neural network;
s12: if ε < ε0Or the training times of the back-stage BP neural network reach the set maximum training times, then S12 is switched; otherwise, turning to S10;
s13: predicting the power load data of the predicted day by using the trained back-stage BP neural network, and outputting a power load prediction curve of the predicted day, wherein the prediction result of the embodiment is shown in FIG. 2;
s14: and (6) ending.
Specifically, historical power load data of a workshop 2018, month 1 and 2018, month 5 and month 15 of a certain company are selected as experimental data, the format of the historical power load data is that power loads are sampled every 15 minutes from 0.00 of the day to 23.45 of the day, 96 points of power load data are counted in one day to form a power load curve of the day, and p is 96.
Specifically, in the preceding-stage BP neural network power load similar day type prediction model, the number of nodes of an input layer is 4, the number of nodes of an output layer is 4, the number of nodes of a hidden layer is 7, namely n1=4,m1=4,M1=7。
Specifically, control parameters of the improved maximum deviation similarity criterion clustering algorithm are set, the maximum deviation gamma is set to be 0.10, the maximum deviation point allowable deviation delta is set to be 0.25, the similarity factor alpha is set to be 0.80, and the continuous deviation factor beta is set to be 0.20, namely, gamma is 0.10, delta is 0.25, alpha is 0.80, and beta is 0.20.
Specifically, the hierarchical clustering algorithm is used for clustering the result C1Class center X ofiClustering to obtain a clustering result C2The hierarchical clustering algorithm stopping condition in (1): until all clusters R of the final partitioniMiddle two elements XiAnd XjIs less than RiAverage distance M ofiλ, where λ is 1.2.
In particular, a back-stage BP neural networkIn the short-term power load prediction model, the number of nodes of an input layer is 10, the number of nodes of an output layer is 3, the number of nodes of a hidden layer is 8, namely n2=4,m2=4,M2When training the short-term power load prediction model of the BP neural network at the later stage, the data adopts a rolling input mode, namely the input of the first training data set is { x1,1,x1,2,...,x1,10The output is { x }1,11,x1,12,x1,13}; the second set of training data is input as { x1,4,x1,5,...,x1,13The output is { x }1,14,x1,15,x1,16}; and so on to guide all training samples to finish training.
The same or similar reference numerals correspond to the same or similar parts;
the terms describing positional relationships in the drawings are for illustrative purposes only and are not to be construed as limiting the patent;
it should be understood that the above-described embodiments of the present invention are merely examples for clearly illustrating the present invention, and are not intended to limit the embodiments of the present invention. Other variations and modifications will be apparent to persons skilled in the art in light of the above description. And are neither required nor exhaustive of all embodiments. Any modification, equivalent replacement, and improvement made within the spirit and principle of the present invention should be included in the protection scope of the claims of the present invention.
Claims (9)
1. A method for short term power load prediction, comprising the steps of:
s1: determining an input variable, an output variable and a training mode of a preceding-stage BP neural network;
s2: determining a prediction day, and collecting historical power load data of N consecutive days before the prediction day;
s3: establishing a preceding stage BP neural network short-term power load prediction model, training the preceding stage BP neural network short-term power load prediction model by using historical power load data, and recording the output of the preceding stage BP neural network short-term power load prediction model to obtain a power load characteristic vector V of a prediction day;
s4: clustering analysis is carried out on historical power load data by using an improved maximum deviation similarity criterion clustering algorithm to obtain a clustering result C1;
S5: clustering result C by using hierarchical clustering algorithm1Class center X ofiClustering to obtain a clustering result C2;
S6: clustering result C2In the same class XiCorresponding C1The cluster in (1) is merged to obtain a clustering result C3And calculate C3Quasi center X 'of'i;
S7: calculating a similar day category Z of a power load curve of a predicted day;
s8: establishing a short-term power load prediction model of a back-stage BP neural network;
s9: normalizing the Z-th class power load data obtained in the step S7, taking the normalized Z-th class power load data as input, and training a rear-stage BP neural network short-term power load prediction model;
s10: calculating a total training error epsilon of the BP neural network;
s11: if ε < ε0Or the training times of the back BP neural network reach the set maximum training times, then S12 is switched, otherwise S9, epsilon0Is a threshold value;
s12: predicting the power load data of the predicted day by using the trained back-stage BP neural network, and outputting a power load prediction curve of the predicted day;
s13: finishing;
the established short-term power load prediction model of the preceding-stage BP neural network in step S3 is as follows:
wherein, i is 1,21,j=1,2,...,m1,k=1,2,...,n1,l=1,2,...,M1,n1Is a preceding stage BP neural netInput layer node number m of short-term power load prediction model1Is the node number, M, of the output layer of the short-term power load prediction model of the preceding-stage BP neural network1The number of nodes of a hidden layer of a short-term power load prediction model of a preceding BP neural network; y iskjIs the output corresponding to the kth input sample of the preceding stage BP neural network short-term power load prediction model, wijIs the weight from the hidden layer to the output layer of the short-term power load prediction model of the preceding BP neural network, bijThe method is characterized in that the bias from a hidden layer of a short-term power load prediction model of a preceding-stage BP neural network to an output layer is adopted, delta is a neuron activation function from the hidden layer of the short-term power load prediction model of the preceding-stage BP neural network to the output layer, and delta adopts a Relu function and is expressed as follows:
δ(x)=max(λx,0)
wherein λ is a linear factor of the Relu function;
alis the input value of the first hidden layer node of the short-term power load prediction model of the preceding BP neural network, alThe calculation formula of (a) is as follows:
wherein k is 1,21,i=1,2,...,M1,l=1,2,...,M1,wkiIs the connection weight value, x, from the input layer to the output layer of the short-term power load prediction model of the preceding BP neural networkkIs an input variable of a short-term power load prediction model of a preceding-stage BP neural network, bkiThe bias from the input layer of the short-term power load prediction model of the preceding-stage BP neural network to the hidden layer is adopted.
2. The method of claim 1, wherein the step S1 includes the following steps:
s11: determining input variables of input layer nodes of the preceding BP neural network Predicting the related power load data of N consecutive days;
s12: determining the output variable of the output layer node of the preceding stage BP neural network Is the relevant power load data for the predicted day;
s13: determining the training mode of the preceding stage BP neural network: and taking the related power load data of the previous day in the N consecutive days of the prediction day as input, and taking the related power load data of the next day as output until all the training samples of the previous N days are trained.
3. The method according to claim 1, wherein the components of the power load feature vector V of the prediction day in step S3 are derived from the output y of the preceding BP neural network, namely:
4. The method of claim 1, wherein the step S4 includes the following steps:
s41: drawing an input power load curve of a maximum deviation similarity criterion clustering algorithm, and setting the ith power load data as xi=(xi1,xi2,...,xik,...,xip),i=1,2,...,n11,2, p, where x isikIs the power load value at the kth time point of the ith load data;
s42: setting control parameters of an improved maximum deviation similarity criterion clustering algorithm, wherein the control parameters comprise a maximum deviation gamma, a maximum deviation point allowable deviation delta, a lowest similarity factor alpha and a continuous deviation factor beta;
s43: calculating two power load curves xiAnd xjAbsolute difference s at the kth time pointijkThe calculation formula is as follows:
sijk=|xik-xjk|
wherein x isikIs the power load curve xiPower load value, x, at the k-th point in timejkIs the power load curve xjA power load value at a kth time point;
s44: calculating xiAnd xjNumber of similarity points nijAnd the maximum number of continuous deviation points mij,nijIs to satisfy sijkS is less than or equal to gammaijkNumber of (1), mijIs that gamma is less than sijkS of < δijkWherein k is 1,21,mijThe expression is as follows:
wherein m isijIs the maximum number of consecutive deviation points,is the power load xiAnd xjAt the k-th0The absolute difference of the individual points in time,is the power load xiAnd xjAt the k-th0The absolute difference of +1 time points,is the power load xiAnd xjAt the k-th0(ii) the absolute difference of + s-1 time points, γ is the maximum deviation, δ is the maximum deviation point allowed deviation;
s45: judging x according to the maximum deviation similarity criterioniAnd xjWhether the two are similar or not is judged according to the following steps:
n0≤nij≤m0
wherein n isijIs the power load curve xiAnd xjNumber of similarity points, n0=[α×m],m0=[β×m]Alpha is a similarity factor, beta is a continuous deviation factor, and alpha is more than or equal to 0 and less than or equal to 1, and gamma is more than or equal to 0 and less than or equal to 1-alpha;
if the above formula is true, then xiAnd xjSimilarly, if the above formula is false, then xiAnd xjAre not similar;
s46: calculating xiAnd xjTotal distance d (x) ofi,xj) The calculation formula is as follows:
wherein s isijkIs the power load xiAnd xjAbsolute difference at the kth time point;
s47: calculating xiS (x) ofi) Obtaining a clustering result C1The method comprises the following specific steps: for all i, j ═ 1,21And i is less than or equal to j, with xiAs a comparison center, nij、mijAre each independently of n0、m0Comparing, all x satisfying the maximum deviation similarity criterioniIs divided into s (x)i) In, let s (x)i)=s(xi)∪{xjAnd x isjRemoving from the historical power load data set U;
s48: calculating xiS (x) ofi) Class center X ofiCalculatingThe formula is as follows:
5. the method of claim 1, wherein the step S5 includes the following steps:
s51: initializing a hierarchical clustering algorithm, and clustering a result C1Class center X ofiClassifying the cluster into a cluster R;
s52: calculation of X in RiAnd XjTwo elements X corresponding to the maximum value of the distance of (1)aAnd XbIs mixing XaAnd XbInto two different clusters R1And R2Wherein the distance calculation formula is as follows:
s53: calculating other elements X in the original cluster RiTo XaAnd XbEuclidean distance d (X)i,Xa) And d (X)i,Xb) If d (X)i,Xa)<d(Xi,Xb) Then X will beiFall under R1In, otherwise, fall under R2;
S54: for class cluster R1And R2Steps S52 and S53 are performed until all the class clusters R are finally dividediMiddle two elements XiAnd XjIs less than Riλ times the average distance of (a), i.e.:
maxd(Xi,Xj)<λMi
wherein M isiThe calculation formula of (2) is as follows:
wherein R is a cluster RiNumber of elements of (c)iIs a cluster-like RiClass center of (1), XjIs a cluster-like RiAn element of (1);
stopping the hierarchical clustering algorithm to obtain a clustering result C2The number of clusters is K.
6. The method of claim 1, wherein the step S6 includes the following steps:
s61: clustering result C2In the same class XiCorresponding C1Merging the clusters in (1), i.e. merging the elements in the clusters into one cluster to obtain a clustering result C3;
S62: calculating a clustering result C3Quasi center X 'of'i。
7. The method of claim 1, wherein the step S7 includes the following steps:
s71: initializing a hierarchical clustering algorithm, and clustering a result C3Quasi center X 'of'iAnd the power load characteristic vector V of the forecast day is classified into a cluster R;
s72: calculating the element X in RiAnd XjTwo elements X corresponding to the maximum value of the distance of (1)aAnd XbIs mixing XaAnd XbInto two different clusters R1And R2Wherein the distance calculation formula is as follows:
s73: calculating other elements X in the original cluster RiTo XaAnd XbEuclidean distance d (X)i,Xa) And d (X)i,Xb) If d (X)i,Xa)<d(Xi,Xb) Then X will beiFall under R1In, otherwise, fall under R2;
S74: for class cluster R1And R2Executing the steps S72 and S73 until the K clusters are finally divided, stopping the hierarchical clustering algorithm to obtain a clustering result C4;
S75: clustering result C4The power load characteristic vector V of the predicted day and the power load characteristic vector V of the predicted day are class centers X 'of the same class cluster'iThe corresponding power load data category is the power load curve similar day category Z of the prediction day.
8. The method according to claim 1, wherein the step S8 is implemented by using the following model for predicting the short-term power load of the back-stage BP neural network:
wherein, i is 1,22,j=1,2,...,m2,k=1,2,...,n2,l=1,2,...,M2,n2The number of nodes of the input layer of the short-term power load prediction model of the back-stage BP neural network, m2Is the number of nodes of the output layer of the short-term power load prediction model of the back-stage BP neural network, M2The number of nodes of a hidden layer of a short-term power load prediction model of a back-stage BP neural network, ykjIs the output corresponding to the kth input sample, wijIs the weight from the hidden layer to the output layer of the short-term power load prediction model of the back-stage BP neural network, bijThe bias from a hidden layer of a short-term power load prediction model of the back-stage BP neural network to an output layer is adopted, delta is a neuron activation function from the hidden layer of the short-term power load prediction model of the back-stage BP neural network to the output layer, and delta adopts a tanh function and is expressed as follows:
clis the input value of the l-th hidden layer node of the short-term power load prediction model of the back-stage BP neural network,clthe calculation formula of (a) is as follows:
wherein, l is 1,22,k=1,2,...,n2,i=1,2,...,M2,vkiIs the connection weight value from the input layer to the output layer of the short-term power load prediction model of the back-stage BP neural network, dkIs an input variable of a short-term power load prediction model of a back-stage BP neural network, ekiAnd predicting the bias from the input layer to the hidden layer of the model for the short-term power load of the back-stage BP neural network.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201811068735.2A CN109309382B (en) | 2018-09-13 | 2018-09-13 | Short-term power load prediction method |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201811068735.2A CN109309382B (en) | 2018-09-13 | 2018-09-13 | Short-term power load prediction method |
Publications (2)
Publication Number | Publication Date |
---|---|
CN109309382A CN109309382A (en) | 2019-02-05 |
CN109309382B true CN109309382B (en) | 2022-03-04 |
Family
ID=65224468
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN201811068735.2A Active CN109309382B (en) | 2018-09-13 | 2018-09-13 | Short-term power load prediction method |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN109309382B (en) |
Families Citing this family (11)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN109934395B (en) * | 2019-03-04 | 2023-05-02 | 三峡大学 | Multi-hydropower-region short-term power load prediction method based on time-sharing and regional meteorological data |
CN110135623A (en) * | 2019-04-12 | 2019-08-16 | 天津大学 | Load prediction input feature vector screening technique based on hierarchical clustering and neural network |
CN110071502B (en) * | 2019-04-24 | 2022-08-12 | 广东工业大学 | Calculation method for short-term power load prediction |
CN110263860A (en) * | 2019-06-21 | 2019-09-20 | 广东工业大学 | A kind of freeway traffic flow prediction technique and device |
CN110728395A (en) * | 2019-09-05 | 2020-01-24 | 广州供电局有限公司 | Main transformer short-term power load calculation method and device, computer and storage medium |
CN111325413B (en) * | 2020-03-23 | 2022-04-01 | 电子科技大学 | Short-term load interval prediction method based on neural network and proportionality coefficient method |
CN111476435B (en) * | 2020-04-30 | 2024-02-09 | 国核电力规划设计研究院有限公司 | Charging pile load prediction method based on density peak value |
CN112365363A (en) * | 2020-10-14 | 2021-02-12 | 国网四川省电力公司电力科学研究院 | Calculation method for similarity of power load curves |
CN112990597B (en) * | 2021-03-31 | 2024-02-27 | 国家电网有限公司 | Ultra-short-term prediction method for industrial park power consumption load |
CN113222216A (en) * | 2021-04-14 | 2021-08-06 | 国网江苏省电力有限公司营销服务中心 | Method, device and system for predicting cooling, heating and power loads |
CN113688910A (en) * | 2021-08-26 | 2021-11-23 | 国网江苏省电力有限公司常州供电分公司 | Power grid operation multi-source information fusion method based on k-BP algorithm |
Citations (1)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN108229754A (en) * | 2018-01-31 | 2018-06-29 | 杭州电子科技大学 | Short-term load forecasting method based on similar day segmentation and LM-BP networks |
Family Cites Families (5)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN104008427A (en) * | 2014-05-16 | 2014-08-27 | 华南理工大学 | Central air conditioner cooling load prediction method based on BP neural network |
CN105090001B (en) * | 2015-07-20 | 2017-05-03 | 北京化工大学 | Reciprocating compressor signal complete period phase correction method based on similarity analysis |
CN105550744A (en) * | 2015-12-06 | 2016-05-04 | 北京工业大学 | Nerve network clustering method based on iteration |
CN108074004A (en) * | 2016-11-12 | 2018-05-25 | 华北电力大学(保定) | A kind of GIS-Geographic Information System short-term load forecasting method based on gridding method |
CN106600059B (en) * | 2016-12-13 | 2020-07-24 | 北京邮电大学 | Intelligent power grid short-term load prediction method based on improved RBF neural network |
-
2018
- 2018-09-13 CN CN201811068735.2A patent/CN109309382B/en active Active
Patent Citations (1)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN108229754A (en) * | 2018-01-31 | 2018-06-29 | 杭州电子科技大学 | Short-term load forecasting method based on similar day segmentation and LM-BP networks |
Non-Patent Citations (1)
Title |
---|
基于最大偏差相似性准则的BP神经网络短期电力负荷预测算法;罗育辉 等;《计算机应用研究(优先出版)》;20180810;第36卷(第11期);第1-6页 * |
Also Published As
Publication number | Publication date |
---|---|
CN109309382A (en) | 2019-02-05 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN109309382B (en) | Short-term power load prediction method | |
CN109754113B (en) | Load prediction method based on dynamic time warping and long-and-short time memory | |
CN108876054B (en) | Short-term power load prediction method based on improved genetic algorithm optimization extreme learning machine | |
CN110674999A (en) | Cell load prediction method based on improved clustering and long-short term memory deep learning | |
CN109934375B (en) | Power load prediction method | |
CN109063911A (en) | A kind of Load aggregation body regrouping prediction method based on gating cycle unit networks | |
CN109861211B (en) | Dynamic reconfiguration method of power distribution network based on data driving | |
CN106971238A (en) | The Short-Term Load Forecasting Method of Elman neutral nets is obscured based on T S | |
Corzo et al. | Knowledge-based modularization and global optimization of artificial neural network models in hydrological forecasting | |
CN109255726A (en) | A kind of ultra-short term wind power prediction method of Hybrid Intelligent Technology | |
CN112329990A (en) | User power load prediction method based on LSTM-BP neural network | |
CN110188221B (en) | Shape distance-based load curve hierarchical clustering method | |
CN113361801A (en) | Hybrid rolling short-term load prediction method based on GAN network and semi-supervised regression | |
CN111461921B (en) | Load modeling typical user database updating method based on machine learning | |
CN111882114B (en) | Short-time traffic flow prediction model construction method and prediction method | |
Tian et al. | Daily power demand prediction for buildings at a large scale using a hybrid of physics-based model and generative adversarial network | |
Lee et al. | Probabilistic wind power forecasting based on the laplace distribution and golden search | |
CN113205223A (en) | Electric quantity prediction system and prediction method thereof | |
CN109460872B (en) | Mobile communication user loss imbalance data prediction method | |
CN109272058B (en) | Integrated power load curve clustering method | |
Luo et al. | Elevator traffic flow prediction with least squares support vector machines | |
CN114595542A (en) | Line loss prediction method | |
CN112052985B (en) | Middle-short-term low-voltage prediction method based on lightgbm | |
Guo et al. | Wind speed forecasting of genetic neural model based on rough set theory | |
CN116485075A (en) | FTS-based power grid load prediction method |
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 | ||
GR01 | Patent grant | ||
GR01 | Patent grant | ||
TR01 | Transfer of patent right |
Effective date of registration: 20230902 Address after: Room 302, Building 4, No. 63 Punan Road, Huangpu District, Guangzhou City, Guangdong Province, 510000 Patentee after: GUANGZHOU AIPU ELECTRON TECHNOLOGY Co.,Ltd. Address before: 510006 Dongfeng East Road, Yuexiu District, Guangzhou, Guangdong 729 Patentee before: GUANGDONG University OF TECHNOLOGY |
|
TR01 | Transfer of patent right |