CN112734135B - Power load prediction method, intelligent terminal and computer readable storage medium - Google Patents
Power load prediction method, intelligent terminal and computer readable storage medium Download PDFInfo
- Publication number
- CN112734135B CN112734135B CN202110102621.0A CN202110102621A CN112734135B CN 112734135 B CN112734135 B CN 112734135B CN 202110102621 A CN202110102621 A CN 202110102621A CN 112734135 B CN112734135 B CN 112734135B
- Authority
- CN
- China
- Prior art keywords
- load
- predicted
- historical
- data
- power
- 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
- 238000000034 method Methods 0.000 title claims abstract description 57
- 230000005611 electricity Effects 0.000 claims abstract description 32
- 238000013145 classification model Methods 0.000 claims abstract description 16
- 239000011159 matrix material Substances 0.000 claims description 84
- 238000012549 training Methods 0.000 claims description 53
- 238000003066 decision tree Methods 0.000 claims description 34
- 230000008569 process Effects 0.000 claims description 18
- 238000004422 calculation algorithm Methods 0.000 claims description 13
- 238000007637 random forest analysis Methods 0.000 claims description 6
- 238000004364 calculation method Methods 0.000 claims description 5
- 238000005457 optimization Methods 0.000 claims description 3
- 238000003062 neural network model Methods 0.000 description 6
- 210000002569 neuron Anatomy 0.000 description 4
- 238000012545 processing Methods 0.000 description 4
- 238000005070 sampling Methods 0.000 description 4
- 230000006870 function Effects 0.000 description 3
- 239000004973 liquid crystal related substance Substances 0.000 description 2
- 238000012986 modification Methods 0.000 description 2
- 230000004048 modification Effects 0.000 description 2
- 238000012360 testing method Methods 0.000 description 2
- 239000002699 waste material Substances 0.000 description 2
- 230000008859 change Effects 0.000 description 1
- 238000001564 chemical vapour infiltration Methods 0.000 description 1
- 238000004590 computer program Methods 0.000 description 1
- 238000007405 data analysis Methods 0.000 description 1
- 238000011161 development Methods 0.000 description 1
- 238000010586 diagram Methods 0.000 description 1
- 230000000694 effects Effects 0.000 description 1
- 238000011478 gradient descent method Methods 0.000 description 1
- 238000003064 k means clustering Methods 0.000 description 1
- 238000002372 labelling Methods 0.000 description 1
- 230000003287 optical effect Effects 0.000 description 1
- 238000010248 power generation Methods 0.000 description 1
- 230000009467 reduction Effects 0.000 description 1
- 230000000007 visual effect Effects 0.000 description 1
Images
Classifications
-
- 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
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N5/00—Computing arrangements using knowledge-based models
- G06N5/01—Dynamic search techniques; Heuristics; Dynamic trees; Branch-and-bound
-
- 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
- G06Q10/06—Resources, workflows, human or project management; Enterprise or organisation planning; Enterprise or organisation modelling
- G06Q10/067—Enterprise or organisation modelling
-
- 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—Information and communication technology [ICT] specially adapted for implementation of business processes of specific business sectors, e.g. utilities or tourism
- G06Q50/06—Energy or water supply
Landscapes
- Engineering & Computer Science (AREA)
- Business, Economics & Management (AREA)
- Strategic Management (AREA)
- Human Resources & Organizations (AREA)
- Economics (AREA)
- Theoretical Computer Science (AREA)
- General Physics & Mathematics (AREA)
- Physics & Mathematics (AREA)
- Entrepreneurship & Innovation (AREA)
- General Business, Economics & Management (AREA)
- Tourism & Hospitality (AREA)
- Marketing (AREA)
- Development Economics (AREA)
- Operations Research (AREA)
- Health & Medical Sciences (AREA)
- Game Theory and Decision Science (AREA)
- Quality & Reliability (AREA)
- Artificial Intelligence (AREA)
- Data Mining & Analysis (AREA)
- Software Systems (AREA)
- General Engineering & Computer Science (AREA)
- Computational Linguistics (AREA)
- Mathematical Physics (AREA)
- Computing Systems (AREA)
- Evolutionary Computation (AREA)
- Public Health (AREA)
- Water Supply & Treatment (AREA)
- General Health & Medical Sciences (AREA)
- Primary Health Care (AREA)
- Educational Administration (AREA)
- Supply And Distribution Of Alternating Current (AREA)
- Management, Administration, Business Operations System, And Electronic Commerce (AREA)
Abstract
The invention discloses a power load prediction method, an intelligent terminal and a computer readable storage medium, wherein the method comprises the following steps: acquiring prediction characteristic data corresponding to time to be predicted, wherein the prediction characteristic data comprises meteorological data and a time type; inputting the predicted characteristic data into a trained classification model and classifying the electricity utilization mode of the time to be predicted through the classification model to obtain a predicted electricity utilization mode corresponding to the time to be predicted; inputting the prediction characteristic data into a trained load prediction model and predicting the electric load of the time to be predicted through the load prediction model to obtain an initial electric load curve corresponding to the time to be predicted; and determining a predicted power load curve corresponding to the time to be predicted according to the predicted power consumption mode and the initial power load curve. The invention can accurately predict the electrical load.
Description
Technical Field
The invention relates to the technical field of power data analysis, in particular to a power load prediction method, an intelligent terminal and a computer readable storage medium.
Background
Because the electric energy is not storable, in order to maintain the grid frequency and ensure a balanced supply and demand relationship between the generated energy and the used amount, the electric power system must predict the electric power load in advance and make a power generation plan according to the prediction result. The objective of power system load prediction is to improve the prediction accuracy and provide a safe and reliable power supply at the lowest possible operating cost. Because the user side, that is to say the power consumption end, has a lot of power consumption factors, in daily electricity generation process, must guarantee that some thermal power generating units are in rotatory standby state. If the electric load of the user side is suddenly reduced, the thermal power generating unit can carry out load shedding, the safety of a power grid is greatly influenced, and huge energy waste is caused. Therefore, short-term load prediction is an important task of the power system, and the accuracy of the prediction result directly influences the stability of the power system and also influences the operation cost of a power grid enterprise and the safety of the power grid.
However, due to many external factors such as climate change, social activities, and residential living habits, the prediction of the power load is highly non-linear and unpredictable, and thus, accurate short-term power load prediction has been a problem in the power industry. Therefore, the development of a short-term power load high-precision prediction method has very important significance for reducing energy waste and optimizing the operation of a power system.
Disclosure of Invention
The invention mainly aims to provide a power load prediction method, an intelligent terminal and a computer readable storage medium, and aims to solve the problem of low short-term power utilization prediction accuracy in the prior art.
In order to achieve the above object, the present invention provides a power load prediction method, including the steps of:
acquiring predicted characteristic data corresponding to time to be predicted, wherein the predicted characteristic data comprises meteorological data and time types;
inputting the prediction characteristic data into a trained classification model and classifying the electricity utilization mode of the time to be predicted through the classification model to obtain a prediction electricity utilization mode corresponding to the time to be predicted;
inputting the prediction characteristic data into a trained load prediction model and predicting the electric load of the time to be predicted through the load prediction model to obtain an initial electric load curve prediction electric load corresponding to the time to be predicted;
and predicting the electric load according to the predicted electric load mode and the initial electric load curve, and determining a predicted electric load curve corresponding to the time to be predicted.
Optionally, in the power load prediction method, the classification model includes a plurality of classification decision trees trained based on a random forest algorithm; the training process of the classification decision tree specifically comprises the following steps:
acquiring historical load data and historical characteristic data corresponding to each preset historical time;
clustering the historical load data to generate a target clustering set, wherein the target clustering set comprises a plurality of power utilization mode sets, and each power utilization mode set comprises historical load data corresponding to the same power utilization mode;
for each electricity utilization mode set, marking the historical characteristic data according to the historical time corresponding to each historical load data in the electricity utilization mode set to obtain the electricity utilization mode corresponding to each historical characteristic data;
and aiming at each preset decision tree, selecting training characteristic data in the historical characteristic data to input into the decision tree and splitting the decision tree according to the Gini index until the power utilization modes corresponding to the historical characteristic data of each node in the decision tree are the same, thereby obtaining the classification decision tree.
Optionally, the power load prediction method, wherein the clustering the historical load data to generate a target cluster set specifically includes:
according to the current clustering number, randomly determining an initial clustering center in the historical load data, wherein when clustering is performed for the first time, the clustering number is 2;
calculating a kth intermediate membership matrix corresponding to the historical load data according to a preset fuzzy C-means clustering algorithm and the initial clustering center, wherein k is the clustering frequency;
when the cluster number is smaller than a preset cluster number threshold value, adding one to the cluster number, and repeatedly determining a kth intermediate membership matrix;
when the clustering number is larger than or equal to the clustering number threshold value, determining a target membership matrix from a first intermediate membership matrix to a (K-1) th intermediate membership matrix, wherein K is the clustering number threshold value;
and clustering the historical load data according to the target membership matrix and a clustering center corresponding to the target membership matrix to obtain a target clustering set.
Optionally, the method for predicting a power load, wherein when the cluster number is greater than or equal to the threshold cluster number, determining a target membership matrix in a first to (K-1) th intermediate membership matrices specifically includes:
calculating a clustering effectiveness index corresponding to each of the first to (K-1) th intermediate membership degree matrixes;
and determining a target clustering set from the first intermediate membership matrix to the (K-1) th intermediate membership matrix according to the clustering effectiveness index.
Optionally, the power load prediction method includes a plurality of load prediction submodels, where the load prediction submodels correspond to the power consumption modes one to one.
Optionally, the power load prediction method, wherein the historical load data includes a historical load curve corresponding to each historical time, and the historical load curve includes a historical load characteristic value; the training process of the load prediction submodel specifically comprises the following steps:
for each preset initial model, inputting training curve data into training characteristic data corresponding to the initial model in the historical characteristic data in the initial model, and classifying the training curve data historical characteristic data through the initial model to obtain a training load characteristic value corresponding to the training curve data historical characteristic data, wherein the training curve data are data corresponding to the initial model in the historical characteristic data;
and performing parameter optimization on the initial model according to the training load characteristic value and the historical load characteristic value corresponding to the historical characteristic data until the initial model is converged to obtain a loss load prediction submodel.
Optionally, in the power load prediction method, the initial power load curve prediction power load includes a candidate predicted power load curve corresponding to each load prediction submodel; the step of inputting the prediction characteristic data into a trained load prediction model and predicting the electric load of the time to be predicted through the load prediction model to obtain an initial electric load curve prediction electric load corresponding to the time to be predicted specifically comprises the following steps:
inputting the predicted characteristic data into each load prediction submodel to obtain a predicted load characteristic value corresponding to the predicted characteristic data;
and calculating a candidate electric load curve prediction electric load corresponding to the load prediction submodel according to the electric mode set corresponding to the load prediction submodel and the prediction load characteristic value.
Optionally, in the power load prediction method, the predicted power utilization mode includes an estimated power utilization mode corresponding to the time to be predicted, where the predicted characteristic data is classified by each of the classification decision trees; the predicting the power load according to the predicted power consumption mode and the initial power load curve and determining the predicted power load curve corresponding to the time to be predicted specifically comprise:
calculating a mode proportion value corresponding to each power utilization mode according to all the estimated power utilization modes;
and for each power utilization mode, according to the mode proportion value corresponding to the power utilization mode, carrying out weighted summation on the candidate power utilization load curve predicted load curve corresponding to the power utilization mode to obtain the predicted power utilization load curve corresponding to the time to be predicted.
In addition, to achieve the above object, the present invention further provides an intelligent terminal, wherein the intelligent terminal includes: a memory, a processor and a power load prediction program stored on the memory and executable on the processor, the power load prediction program when executed by the processor implementing the steps of the power load prediction method as described above.
Further, to achieve the above object, the present invention also provides a computer-readable storage medium storing a power load prediction program that realizes the steps of the power load prediction method as described above when executed by a processor.
The method comprises the steps of dividing power utilization prediction into mode prediction and load prediction, firstly obtaining to-be-predicted characteristic data corresponding to-be-predicted time, determining a corresponding predicted power utilization mode according to the predicted characteristic data, predicting an initial power utilization load curve corresponding to the predicted characteristic data according to the predicted characteristic data, and then adjusting the predicted power utilization load according to the predicted power utilization mode, so that the power utilization load curve corresponding to the to-be-predicted time is obtained. Because the prediction of the power utilization mode is added on the basis of the conventional load prediction, and the result of the load prediction is adjusted on the basis of the power utilization mode, the prediction result is more accurate.
Drawings
FIG. 1 is a flow chart of a preferred embodiment of a power load prediction method according to the present invention;
fig. 2 is a schematic operating environment diagram of an intelligent terminal according to a preferred embodiment of the invention.
Detailed Description
In order to make the objects, technical solutions and advantages of the present invention clearer and clearer, the present invention is further described in detail below with reference to the accompanying drawings and examples. It should be understood that the specific embodiments described herein are merely illustrative of the invention and do not limit the invention.
In the power load prediction method according to the preferred embodiment of the present invention, the power load prediction method may be performed by an intelligent terminal, and the intelligent terminal includes a terminal such as a smart television and a smart phone. As shown in fig. 1, the power load prediction method includes the steps of:
step S100, obtaining prediction characteristic data corresponding to time to be predicted, wherein the prediction characteristic data comprises meteorological data and time types.
Specifically, the predicted characteristic data corresponding to the time to be predicted is obtained first, and in this embodiment, the predicted characteristic data refers to data according to characteristics related to the electrical load, such as meteorological data and time type. Taking meteorological data as an example, when the weather is hot, a household needs an electric fan and an air conditioner, so that the power consumption load is increased compared with that under the normal temperature condition. The meteorological data in the present embodiment includes temperature, wind power, humidity, and the like. The time type refers to the time characteristics of the time to be predicted and the power load, for example, the power load in late night is definitely different from the power load in daytime, and the power loads in working days, weekends and holidays are also different.
In addition, the time to be predicted in the embodiment is based on the day as a unit, that is, the power load of a certain day is predicted, but in practical application, the time to be predicted may be divided in more detail or roughly according to the actual demand, for example, the time to be predicted is a certain hour of a certain day, or a power load of a certain week.
Step S200, inputting the prediction characteristic data into a trained classification model and classifying the power utilization mode of the time to be predicted through the classification model to obtain a prediction power utilization mode corresponding to the time to be predicted.
Specifically, the predicted feature data is input into a trained classification model, the classification model used in the embodiment is a classification with supervised learning, and the classification model calculates a probability value that the predicted feature data is a preset power consumption mode, so that the predicted power consumption mode is obtained. In the first method of obtaining the predicted power consumption pattern according to the present embodiment, the power consumption pattern having the highest probability value is directly used as the predicted power consumption pattern.
Further, in this embodiment, the classification model includes a plurality of classification decision trees trained based on a random forest algorithm; the classification decision tree is obtained by training the decision tree on the basis of a random forest algorithm, and the process of obtaining the classification decision tree in the embodiment is as follows:
and A10, acquiring historical load data and historical characteristic data corresponding to preset historical time.
Specifically, historical load data corresponding to each preset historical time is acquired, and since the present embodiment takes a day as a unit of time to be predicted, the historical load data acquired during training is also taken in a day unit, and X ═ X is set1,x2,…,xNAnd N represents the total number of days the data set contains. Further, in this embodiment, the historical load data for each historical time, i.e., each day, includes a normalized value of the historical load curve for that day, where the historical load curve is xi＝{xi1,xi2,…,ximAnd m represents a load sampling point of one day, and then the load value corresponding to each sampling point in the historical load curve is divided by the maximum value in the historical load curve, so that the historical load curve is normalized, and the historical load data is obtained.
The historical characteristic data are meteorological data and time types corresponding to the acquired historical time, and the set of the historical characteristic data is represented as Y ═ Y1,y2,…,yNN represents the total number of days included in the data set, and the historical characteristic data corresponding to each historical time can be represented as yi＝{yi1,yi2,…,yin,yin+1,yin+2,yin+3And the former n values are weather characteristic data, and the latter three values are encoding results of time types in this embodiment, where the encoding results in this embodiment adopt a one-hot encoding manner, and the time types are three time types, namely weekday, weekend, holiday and holiday, but when applied, the encoding results may also include time types of monday to sunday, morning, afternoon, and the like. For example, a working day of [0,0,1 ]]And the weekend is [0,1,0 ]]Holidays of [1,0]。
This example collected 697 days of load data and profile data for market a. In the present embodiment, the sampling period in the load curve, that is, the interval between sampling points is 1 hour, so that each load data includes 24 normalized load values, and the load values are normalized to obtain the load data. The characteristic data includes meteorological data including 5 characteristics of maximum wind speed, minimum wind speed, maximum temperature, minimum temperature, and surface average air pressure, and a time type. The dimension of the load data X is 24 × 730, and the dimension of the feature data Y is 8 × 730.
To evaluate the accuracy of the prediction method employed in this embodiment, load data of 486 days (486/697 about 0.7) was randomly selected as historical load data and its corresponding characteristic data as historical characteristic data in a 7:3 manner in this embodiment. The remaining 211 days are taken as the time to be predicted, and the corresponding load data and characteristic data are taken as the test load data and the test characteristic data for evaluating the accuracy of the method adopted by the embodiment.
A20, clustering the historical load data to generate a target clustering set, wherein the target clustering set comprises a plurality of electricity utilization mode sets, and each electricity utilization mode set comprises historical load data corresponding to the same electricity utilization mode.
Specifically, a target cluster set is generated by clustering historical load data, wherein the target cluster set comprises a plurality of power utilization mode sets. In this embodiment, the clustering process may adopt a supervised or semi-supervised clustering algorithm, or an unsupervised clustering algorithm. Such as K-Means clustering, mean shift clustering algorithms, hierarchical clustering.
Further, in this embodiment, an improved fuzzy C-means clustering algorithm is adopted as a way of clustering historical load data in the clustering process, and the specific process is as follows:
and B10, randomly determining an initial cluster center in the historical load data according to the current cluster number.
Specifically, before clustering is performed, the current number of clusters is obtained, and the number of clusters increases with the increase of the clustering frequency, and in this embodiment, when clustering is performed for the first time, the number of clusters is 2, that is, the number of clusters c is 2 for the first time. Random selection from historical load datasetc samples as initial clustering center thetajWherein j ∈ [1, c ]]。
And B20, clustering the historical load data according to a preset fuzzy C-means clustering algorithm and the initial clustering center to obtain a kth clustering set corresponding to the current clustering number.
Specifically, a Fuzzy C-Means (FCM) algorithm is a clustering algorithm introducing Fuzzy power, samples are classified by calculating the probability that a sample belongs to a certain class as a membership degree, and FCM implements clustering by minimizing a target function and a preset constraint condition. The process is as follows:
updating the (r-1) th clustering center and the (r-1) th membership matrix to obtain the r th clustering center and the r th initial membership matrix, wherein the updating process comprises the following steps:
calculating the amount of the (r-1) th clustering center according to the (r-1) th membership matrix, wherein (r-1) is iteration times, when the iteration is performed for the first time, the first clustering center is a primary clustering center, and the first initial membership matrix is a membership matrix corresponding to the historical load data constructed by random numbers;
updating the (r-1) th clustering center according to a preset clustering center updating formula to obtain an r-th clustering center, updating the formula according to a preset membership value, and updating the (r-1) th membership matrix to obtain an r-th initial membership matrix;
when the r initial membership matrix is determined to accord with a preset constraint condition, taking the r initial membership matrix as a k intermediate membership matrix and outputting the k intermediate membership matrix;
and when the r initial membership matrix is determined not to meet the constraint condition, iteratively updating the r clustering center and the r initial membership matrix.
In this embodiment, taking (r-1) as an example for description, after determining a plurality of initial clustering centers according to the clustering number c, a first initial membership matrix U is constructed by using random numbersN×c＝{uij1, N, j 1, c, wherein uijFor the ith historical load data pairMembership value of j cluster centers, random number [0,1 ]]And N is the data quantity of the historical load data corresponding to all historical time. According to the membership value between each historical load data and the initial clustering center in the membership matrix, the corresponding volume of each first clustering center can be calculated, and the calculation formula of the volume is
Because the initial clustering center is randomly selected and the first membership value is randomly distributed and has a large difference from the real situation, the initial clustering center needs to be updated to obtain the second clustering center, wherein the updating formula of the clustering center is
q is a preset ambiguity value. Meanwhile, the reliability of the membership value in the first initial membership matrix can be evaluated according to the quantity, so that the first membership value can be updated according to the calculated quantity to obtain a second membership value, and the updating formula of the membership value is
Wherein DTW (x)i,θj) Is the dynamic time warping distance.
And after the second initial membership matrix is obtained, judging whether the second initial membership matrix meets a preset constraint rule or not. If yes, the second initial membership matrix is used as a k-th intermediate membership matrix and output; and if not, continuously updating the membership value and the clustering center until the r initial membership matrix meets the constraint rule, and taking the r initial membership matrix as a k intermediate membership matrix and outputting the k intermediate membership matrix.
Further, the constraint conditions adopted in this embodiment are
Or R is larger than or equal to R, wherein epsilon is a preset iteration error threshold, and R is an iteration time threshold.
And B30, when the cluster number is smaller than a preset cluster number threshold value, adding one to the cluster number, and repeatedly determining the k-th intermediate membership matrix.
Specifically, a cluster number threshold value K is preset, when the cluster number is smaller than the preset cluster number threshold value, the cluster number is increased by one, taking the current cluster number as 2 as an example, the cluster number after the addition is 3, then the initial cluster center is confirmed again in the historical load data with the cluster number of 3, and a corresponding second intermediate membership matrix is calculated when the cluster number is 3.
And B40, when the cluster number is larger than or equal to the cluster number threshold value, determining a target membership matrix from the first intermediate membership matrix to the (K-1) th intermediate membership matrix.
Specifically, when the cluster number is greater than or equal to the cluster number threshold, iteratively increasing the cluster number from 2 to K-th intermediate membership matrix in the process that the cluster number is K to determine a target membership matrix. Because an intermediate membership matrix is output every time the clustering number is increased, each intermediate membership matrix is different certainly, and the intermediate membership matrix which is the most accurate in calculation, namely the target membership matrix, is selected. In the present embodiment, K is 10, and thus the first to ninth membership matrices are obtained.
In this embodiment, the method for determining the target membership matrix is as follows:
calculating a clustering effectiveness index corresponding to each of the first to (K-1) th intermediate membership degree matrixes;
and determining a target clustering set from the first intermediate membership matrix to the (K-1) th intermediate membership matrix according to the clustering validity index.
Specifically, for each Cluster set, a Cluster Validity Index (CVI) corresponding to the Cluster set is calculated, in this embodiment, a calculation formula of the Cluster Validity Index is
Wherein c is the number of clusters,
and the minimum value of the dynamic time warping distance of the corresponding clustering center of the intermediate membership matrix is obtained. And then selecting a middle membership matrix corresponding to the minimum value in all the CVIs as a target membership matrix, and recording the number of clusters corresponding to the target membership matrix as c. In this embodiment, c ═ 5, that is, five electricity utilization patterns obtained through training of historical load data. Due to the cluster center thetajThe historical load data is the historical load data corresponding to a certain day, and the historical load data in the embodiment is the load value which is acquired and normalized every hour in 24 hours, so that the cluster center theta isjCan be identified as a vector of 24 numerical values combined. Table 1 below shows cluster centers of 5 power consumption patterns determined based on the historical load data, where each column indicates a cluster center corresponding to the power consumption pattern. And table 2 shows membership values of the historical load data corresponding to the cluster centers of different electricity utilization modes.
TABLE 1
TABLE 2
And B50, clustering the historical load data according to the target membership matrix and the clustering center corresponding to the target membership matrix to obtain a target clustering set.
Specifically, according to a target membership matrix and a clustering center corresponding to the target membership matrix, defuzzification processing is performed on the historical load data, and for each historical load data, the clustering center corresponding to the maximum value in the membership values corresponding to the historical load data is used as a central point, so that each historical load data is clustered, and a target clustering set is obtained. For example, the membership values of the historical load data to the clustering centers A, B, C, D and E corresponding to the target membership matrix are 0.1, 0.2, 0.1, 0.5 and 0.1, respectively, so that the historical load data a is classified as the category of the clustering center D. And when the clustering is finished, the category corresponding to each clustering center is used as an electricity utilization mode, so that c × electricity utilization mode sets are obtained.
And A30, labeling the historical characteristic data according to the historical time corresponding to each historical load data in the power consumption mode set aiming at each power consumption mode set to obtain the power consumption mode corresponding to each historical characteristic data.
Specifically, for each electricity utilization pattern set, for example, the electricity utilization pattern set a, the historical time corresponding to the historical load data in the electricity utilization pattern set a is labeled, for example, 3 months and 3 months, and 5 days, and the labeled electricity utilization pattern is the electricity utilization pattern a, so that each historical characteristic data is labeled to obtain the electricity utilization pattern corresponding to each historical characteristic data, that is, the historical characteristic data corresponding to 3 months and 3 months, and 5 days are labeled as the electricity utilization pattern a.
A40, aiming at each preset decision tree, selecting training feature data in the historical feature data to input into the decision tree, and splitting the decision tree according to the Gini index until the power utilization modes corresponding to the historical feature data of each node in the decision tree are the same, so as to obtain the classification decision tree.
Specifically, in this embodiment, a random forest model is constructed by using a random forest packet of Python, and parameters are set as follows: the n _ trees is 200, that is, 200 decision trees are preset, wherein the number of the decision trees can be adjusted according to actual needs. And then randomly and replaceably extracting a plurality of training feature data from all the historical feature data to obtain training feature data sets with the same number as the decision trees. Each training feature data set is used to train a decision tree. In this embodiment, each training feature data set includes 400 training feature data.
And for each decision tree, starting to split downwards from the root node, selecting the best characteristic for splitting each time according to the Gini index until the training characteristic data of all the nodes are from the same class, namely the training characteristic data of the node corresponds to the same electricity utilization mode, and stopping splitting to obtain a classification decision tree.
Step S300, inputting the prediction characteristic data into a trained load prediction model and predicting the electric load of the time to be predicted through the load prediction model to obtain an initial electric load curve corresponding to the time to be predicted.
Specifically, the obtained prediction characteristic data is input into a trained load prediction model, and the load prediction model is used for calculating an initial power load curve corresponding to the prediction characteristic data according to the input prediction characteristic data. The initial model before the load prediction model training is preferably a neural network model.
In this embodiment, the load prediction model includes a plurality of load prediction submodels, and the load prediction submodels correspond to the electricity consumption modes one to one. That is, each power consumption mode corresponds to one load prediction submodel. Thereby avoiding the reduction of prediction accuracy caused by the difference of the power utilization modes. In an implementation manner of obtaining the predicted electrical load in this embodiment, after obtaining the predicted electrical load mode, the predicted characteristic data is input into a load prediction model corresponding to the predicted electrical load mode, so as to obtain a corresponding initial electrical load curve.
In this embodiment, the training process of the load prediction model is as follows:
inputting training curve data into each preset initial model, classifying the training curve data through the initial model, and obtaining a training load characteristic value corresponding to each historical characteristic data, wherein the training curve data are data corresponding to the initial model in the historical characteristic data;
and according to the training load characteristic value and the historical load characteristic value corresponding to the historical characteristic data, performing parameter optimization on the initial model until the initial model is converged to obtain a load prediction submodel.
Specifically, in this embodiment, the number of neural network models equal to the number of power consumption pattern types is preset, that is, five neural network models are preset, each neural network model includes 5 layers, the first layer is an input layer, and the number of neurons is equal to 8; the middle 3 layers are hidden layers, and each layer comprises 10 neurons; the last layer is an output layer, which is composed of two neurons. Wherein the number of neurons in the input layer is determined by the number of types of predicted feature data, i.e., m + 3. In addition, a training-related parameter learning rate γ, a batch size batch _ size, and a training target minimum error ∈ are set in advance. In the present embodiment, the learning rate γ is 0.001, the batch size batch _ size is 64, and the training target minimum error ∈ is 1 e-5;
in a first implementation manner of this embodiment, the training curve data corresponding to the initial model in the historical feature data is directly determined according to the power consumption mode corresponding to the initial model, for example, the power consumption mode a, and when the corresponding initial model is trained, only the historical feature data corresponding to the power consumption mode a is used as the training curve data corresponding to the initial model. In a second implementation manner of this embodiment, the training curve data corresponding to the initial model is determined according to the membership value corresponding to each piece of historical feature data. For example, if the membership value of the historical characteristic data a to the cluster center of the electricity usage pattern a is 0.99, and the membership value of the historical characteristic data B to the cluster center of the electricity usage pattern a is 0.1, the historical characteristic data a is selected as the training curve data corresponding to the initial model with a high probability, and the historical characteristic data B is selected as the training curve data corresponding to the initial model with a low probability.
The initial model classifies the input training curve data to obtain a corresponding training load curve. Because the historical load data includes the historical load curve corresponding to each historical time, and the difficulty in predicting the curve individually is high, in this embodiment, the historical load characteristic value is used as a value for describing a training load curve. The historical load characteristic value refers to a numerical value which can describe the characteristic of the historical load curve, such as a peak value, a valley value, an average value and the like. The present embodiment is described taking as an example both the peak value and the bottom value as the historical load characteristic value. The training load characteristic value refers to a load characteristic value obtained by classifying input training characteristic data.
Then, based on a preset loss function, according to a loss value between the training load curve and the historical load data corresponding to the actually input training curve data, the loss function in this embodiment is a formula for calculating a root mean square error value of a peak value and a valley value. And then, according to the loss value, optimizing the parameters of the neural network model by adopting a random gradient descent method until the neural network model converges to obtain a load prediction submodel. In each iteration, the above process is repeated
Next, N is the number of historical load data. The convergence rule in this embodiment is that the loss value is less than the training target minimum error ∈ 1 e-5.
Further, in a second implementation manner of obtaining the predicted electrical load in this embodiment, the predicted electrical load includes a predicted electrical load corresponding to each load prediction submodel, and a process of obtaining the predicted electrical load is as follows:
inputting the predicted characteristic data into each load prediction submodel to obtain a predicted load characteristic value corresponding to the predicted characteristic data;
and calculating a candidate power load curve corresponding to the load forecasting sub-model according to the power consumption mode set corresponding to the load forecasting sub-model and the forecasting load characteristic value.
Specifically, for each of the load prediction submodels, for example, the load prediction submodel corresponding to the electricity consumption mode a, the predicted characteristic data is input to the load prediction submodel, and the predicted load submodel performs electricity consumption load prediction on the predicted characteristic data to obtain a predicted load characteristic value corresponding to the predicted characteristic data. In this embodiment, the predicted peak and the predicted valley correspond to the time to be predicted.
Because the historical electric loads in the electric mode set corresponding to the load forecasting submodel have the same electric power trend and electric power trend, the candidate electric load curve corresponding to the load forecasting submodel can be calculated according to the electric power mode set and the obtained forecasting peak value and forecasting valley value. And obtaining a candidate power load curve corresponding to each load forecasting submodel, thereby obtaining an initial power load curve.
Further, in this embodiment, the formula in the candidate electrical load curve corresponding to the load forecasting submodel is calculated as
Wherein j is 1,2jRepresents the predicted peak, valley, obtained by the jth load predictor modeljThe predicted trough value obtained by the jth load predictor model is shown.
And S400, predicting the electric load according to the predicted electric load mode and the initial electric load curve, and determining a predicted electric load curve corresponding to the time to be predicted.
Specifically, in this embodiment, after the predicted power consumption mode and the initial power consumption load curve are obtained, the initial power consumption load curve is adjusted according to the predicted power consumption mode, and the overall trend and the trend of the power consumption load curve are closer to the historical load data in the predicted power consumption mode, so that the predicted power consumption load curve corresponding to the time to be predicted is obtained.
Further, in this embodiment, the predicted power utilization mode includes an estimated power utilization mode corresponding to the time to be predicted, which is obtained by classifying the predicted feature data by each of the classification decision trees. And firstly, calculating a mode ratio corresponding to each power consumption mode according to all the estimated power consumption modes. For example, the predicted power consumption pattern L corresponding to the time to be predicted is L ═ L1,l2,...,ln_treesH, wherein ln_treesIs the n-thThe predicted power consumption mode corresponding to each classification decision tree. Counting and predicting mode ratio corresponding to each power consumption mode in power consumption modes
Wherein
njThe power consumption mode is a mode proportion value corresponding to the jth power consumption mode. In this embodiment, a total of estimated power consumption modes of 200 classification decision trees are obtained, and the mode ratio values corresponding to the 5 power consumption modes are respectively: 0.3195,0.0589,0.4729,0.0076 and 0.1411.
For each power utilization mode, according to the mode ratio corresponding to the power utilization mode, carrying out weighted summation on the candidate power utilization load curve corresponding to the power utilization mode to obtain the candidate power utilization load curve corresponding to the time to be predicted, wherein the calculation formula is
In this embodiment, the 211-day-to-be-predicted time is taken as an example, the Mean Absolute Percentage Error (MAPE) between the predicted power load curve and the real power load curve is calculated, and the Mean value of the MAPE corresponding to the 211-day-to-be-predicted time is 3.1%.
Further, as shown in fig. 2, based on the power load prediction method, the present invention also provides an intelligent terminal, where the intelligent terminal includes a processor 10, a memory 20, and a display 30. Fig. 2 shows only some of the components of the smart terminal, but it is to be understood that not all of the shown components are required to be implemented, and that more or fewer components may be implemented instead.
The memory 20 may be an internal storage unit of the intelligent terminal in some embodiments, such as a hard disk or a memory of the intelligent terminal. The memory 20 may also be an external storage device of the Smart terminal in other embodiments, such as a plug-in hard disk, a Smart Media Card (SMC), a Secure Digital (SD) Card, a Flash memory Card (Flash Card), and the like, which are provided on the Smart terminal. Further, the memory 20 may also include both an internal storage unit and an external storage device of the smart terminal. The memory 20 is used for storing application software installed in the intelligent terminal and various data, such as program codes of the installed intelligent terminal. The memory 20 may also be used to temporarily store data that has been output or is to be output. In one embodiment, the memory 20 stores a power load prediction program 40, and the power load prediction program 40 can be executed by the processor 10 to implement the power load prediction method of the present application.
The processor 10 may be a Central Processing Unit (CPU), microprocessor or other data Processing chip in some embodiments, and is used for executing program codes stored in the memory 20 or Processing data, such as executing the power load prediction method.
The display 30 may be an LED display, a liquid crystal display, a touch-sensitive liquid crystal display, an OLED (Organic Light-Emitting Diode) touch panel, or the like in some embodiments. The display 30 is used for displaying information at the intelligent terminal and for displaying a visual user interface. The components 10-30 of the intelligent terminal communicate with each other via a system bus.
In one embodiment, when processor 10 executes power load prediction program 40 in memory 20, the following steps are implemented:
acquiring predicted characteristic data corresponding to time to be predicted, wherein the predicted characteristic data comprises meteorological data and time types;
inputting the predicted characteristic data into a trained classification model and classifying the electricity utilization mode of the time to be predicted through the classification model to obtain a predicted electricity utilization mode corresponding to the time to be predicted;
inputting the prediction characteristic data into a trained load prediction model and predicting the electric load of the time to be predicted through the load prediction model to obtain an initial electric load curve prediction electric load corresponding to the time to be predicted;
and predicting the electric load according to the predicted electric load mode and the initial electric load curve, and determining a predicted electric load curve corresponding to the time to be predicted.
The present invention also provides a computer-readable storage medium, wherein the computer-readable storage medium stores a power load prediction program, which when executed by a processor implements the steps of the power load prediction method as described above.
Of course, it can be understood by those skilled in the art that all or part of the processes in the methods of the embodiments described above can be implemented by instructing relevant hardware (such as a processor, a controller, etc.) by a computer program, and the program can be stored in a computer-readable storage medium, and when executed, the program can include the processes of the methods described above. The computer readable storage medium may be a memory, a magnetic disk, an optical disk, etc.
It will be understood that the invention is not limited to the examples described above, but that modifications and variations will occur to those skilled in the art in light of the above teachings, and that all such modifications and variations are considered to be within the scope of the invention as defined by the appended claims.
Claims (7)
1. A power load prediction method, comprising:
acquiring prediction characteristic data corresponding to time to be predicted, wherein the prediction characteristic data comprises meteorological data and a time type;
inputting the predicted characteristic data into a trained classification model and classifying the electricity utilization mode of the time to be predicted through the classification model to obtain a predicted electricity utilization mode corresponding to the time to be predicted;
inputting the prediction characteristic data into a trained load prediction model and predicting the electric load of the time to be predicted through the load prediction model to obtain an initial electric load curve corresponding to the time to be predicted;
determining a predicted power load curve corresponding to the time to be predicted according to the predicted power consumption mode and the initial power load curve;
the classification model comprises a plurality of classification decision trees obtained based on random forest algorithm training; the training process of the classification decision tree specifically comprises the following steps:
acquiring historical load data and historical characteristic data corresponding to each preset historical time;
clustering the historical load data to generate a target clustering set, wherein the target clustering set comprises a plurality of power utilization mode sets, and each power utilization mode set comprises historical load data corresponding to the same power utilization mode;
for each power utilization mode set, marking the historical characteristic data according to the historical time corresponding to each historical load data in the power utilization mode set to obtain the power utilization mode corresponding to each historical characteristic data;
aiming at each preset decision tree, selecting training characteristic data in the historical characteristic data to input into the decision tree and splitting the decision tree according to a Gini index until power utilization modes corresponding to the historical characteristic data of each node in the decision tree are the same to obtain a classification decision tree;
the clustering the historical load data to generate a target cluster set specifically includes:
according to the current clustering number, randomly determining an initial clustering center in the historical load data, wherein when clustering is performed for the first time, the clustering number is 2;
calculating a kth intermediate membership matrix corresponding to the historical load data according to a preset fuzzy C-means clustering algorithm and the initial clustering center, wherein k is the clustering frequency;
updating the (r-1) th clustering center and the (r-1) th membership matrix to obtain the r th clustering center and the r th initial membership matrix, wherein the updating process comprises the following steps:
calculating the amount of the (r-1) th clustering center according to the (r-1) th membership matrix, wherein (r-1) is iteration times, when the iteration is performed for the first time, the first clustering center is a primary clustering center, and the first initial membership matrix is a membership matrix corresponding to the historical load data constructed by random numbers;
updating the (r-1) th clustering center according to a preset clustering center updating formula to obtain an r-th clustering center, updating the formula according to a preset membership value, and updating the (r-1) th membership matrix to obtain an r-th initial membership matrix;
when the r initial membership matrix is determined to accord with a preset constraint condition, taking the r initial membership matrix as a k intermediate membership matrix and outputting the k intermediate membership matrix;
when the r initial membership matrix is determined not to meet the constraint condition, iteratively updating the r clustering center and the r initial membership matrix;
determining a plurality of initial clustering centers according to the clustering number c, and constructing a first initial membership matrix U by random numbersN×c＝{uij1, N; j-1.. c, wherein uijThe membership value of the ith historical load data to the jth clustering center is a random number of [0, 1%]Random number of interval, N is the data volume of historical load data corresponding to all historical time;
according to the membership value between each historical load data and the initial clustering center in the membership matrix, the corresponding mass of each first clustering center can be calculated, and the calculation formula of the mass is as follows:
updating the initial clustering center to obtain a second clustering center, wherein the clustering center updating formula is as follows:
wherein, the first and the second end of the pipe are connected with each other,q is a predetermined ambiguity value, xi＝{xi1，xi2，...，ximThe historical load curve is used as the data;
according to the amount, updating the first membership value to obtain a second membership value, wherein the updating formula of the membership value is as follows:
wherein DTW (x)i，θj) Is the dynamic time warping distance, fkThe amount of the k-th cluster center, θzUpdating a formula for the clustering center when j takes a value of Z;
when the cluster number is smaller than a preset cluster number threshold value, adding one to the cluster number, and repeatedly determining a kth intermediate membership matrix;
when the cluster number is larger than or equal to the cluster number threshold value, determining a target membership matrix from a first intermediate membership matrix to a (K-1) th intermediate membership matrix, wherein K is the cluster number threshold value;
clustering the historical load data according to the target membership matrix and a clustering center corresponding to the target membership matrix to obtain a target clustering set;
the initial power load curve comprises a candidate power load curve corresponding to each load forecasting submodel; the step of inputting the prediction characteristic data into a trained load prediction model and predicting the electric load of the time to be predicted through the load prediction model to obtain an initial electric load curve corresponding to the time to be predicted specifically comprises the following steps:
inputting the predicted characteristic data into each load prediction submodel to obtain a predicted load characteristic value corresponding to the predicted characteristic data;
calculating a candidate power load curve corresponding to the load forecasting submodel according to the power consumption mode set corresponding to the load forecasting submodel and the forecasting load characteristic value;
the predicted load characteristic value corresponding to the predicted characteristic data is a predicted peak value and a predicted valley value corresponding to the time to be predicted;
the formula in the candidate electric load curve corresponding to the load forecasting submodel is calculated as follows:
wherein j is 1,2*，peakjRepresents the predicted peak, valley, obtained by the jth load predictor modeljThe predicted valley value obtained by the jth load predictor model is shown.
2. The method according to claim 1, wherein when the cluster number is greater than or equal to the cluster number threshold, determining a target membership matrix from a first intermediate membership matrix to a (K-1) th intermediate membership matrix specifically includes:
calculating a clustering effectiveness index corresponding to each of the first to (K-1) th intermediate membership degree matrixes;
and determining a target clustering set from the first intermediate membership matrix to the (K-1) th intermediate membership matrix according to the clustering validity index.
3. The power load prediction method of claim 1, wherein the load prediction model comprises a plurality of load prediction submodels, and the load prediction submodels correspond to the power usage patterns one to one.
4. The power load prediction method according to claim 3, wherein the historical load data includes a historical load profile corresponding to each of the historical times, the historical load profile including a historical load characteristic value; the training process of the load prediction submodel specifically comprises the following steps:
inputting training curve data into each preset initial model, classifying the training curve data through the initial model, and obtaining a training load characteristic value corresponding to each historical characteristic data, wherein the training curve data are data corresponding to the initial model in the historical characteristic data;
and performing parameter optimization on the initial model according to the training load characteristic value and the historical load characteristic value corresponding to the historical characteristic data until the initial model is converged to obtain a loss load prediction submodel.
5. The power load prediction method according to claim 1, wherein the predicted power utilization pattern includes a predicted power utilization pattern corresponding to the time to be predicted, which is obtained by classifying the predicted characteristic data by each of the classification decision trees; the predicting the power load according to the predicted power utilization mode and the initial power load curve and determining the predicted power load curve corresponding to the time to be predicted specifically include:
calculating a mode proportion value corresponding to each power utilization mode according to all the estimated power utilization modes;
and for each power consumption mode, carrying out weighted summation on the candidate power consumption load curve corresponding to the power consumption mode according to the mode ratio corresponding to the power consumption mode to obtain a predicted power consumption load curve corresponding to the time to be predicted.
6. An intelligent terminal, characterized in that, intelligent terminal includes: memory, a processor and a power load prediction program stored on the memory and executable on the processor, the power load prediction program when executed by the processor implementing the steps of the power load prediction method according to any one of claims 1-5.
7. A computer-readable storage medium, characterized in that the computer-readable storage medium stores a power load prediction program which, when executed by a processor, realizes the steps of the power load prediction method according to any one of claims 1 to 5.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202110102621.0A CN112734135B (en) | 2021-01-26 | 2021-01-26 | Power load prediction method, intelligent terminal and computer readable storage medium |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202110102621.0A CN112734135B (en) | 2021-01-26 | 2021-01-26 | Power load prediction method, intelligent terminal and computer readable storage medium |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN112734135A CN112734135A (en) | 2021-04-30 |
| CN112734135B true CN112734135B (en) | 2022-07-15 |
Family
ID=75594025
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202110102621.0A Active CN112734135B (en) | 2021-01-26 | 2021-01-26 | Power load prediction method, intelligent terminal and computer readable storage medium |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN112734135B (en) |
Families Citing this family (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN113516295B (en) * | 2021-05-25 | 2023-01-20 | 广东电网有限责任公司广州供电局 | Space load prediction method and system for rapid power restoration after disaster |
| CN113673587B (en) * | 2021-08-12 | 2024-02-02 | 国网湖北省电力有限公司电力科学研究院 | Household basic daily electricity quantity calculation method |
| CN113819581B (en) * | 2021-08-23 | 2022-11-18 | 广西电网有限责任公司电力科学研究院 | Summer air conditioner load capacity and air temperature sensitivity analysis method |
| CN113705929B (en) * | 2021-09-15 | 2024-05-07 | 中国南方电网有限责任公司 | Spring festival holiday load prediction method based on load characteristic curve and typical characteristic value fusion |
| CN116485049B (en) * | 2023-06-25 | 2024-04-19 | 陕西银河电力仪表股份有限公司 | Electric energy metering error prediction and optimization system based on artificial intelligence |
Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN107145976A (en) * | 2017-04-28 | 2017-09-08 | 北京科技大学 | A kind of method for predicting user power utilization load |
| CN110071502A (en) * | 2019-04-24 | 2019-07-30 | 广东工业大学 | A kind of calculation method of short-term electric load prediction |
| CN111028100A (en) * | 2019-11-29 | 2020-04-17 | 南方电网能源发展研究院有限责任公司 | Refined short-term load prediction method, device and medium considering meteorological factors |
| CN111832796A (en) * | 2020-02-29 | 2020-10-27 | 上海电力大学 | Fine classification and prediction method and system for residential electricity load mode |
| CN112215426A (en) * | 2020-10-16 | 2021-01-12 | 国网山东省电力公司信息通信公司 | Short-term power load prediction method |
Family Cites Families (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN104881706B (en) * | 2014-12-31 | 2018-05-25 | 天津弘源慧能科技有限公司 | A kind of power-system short-term load forecasting method based on big data technology |
| CN105989420B (en) * | 2015-02-12 | 2020-07-17 | 西门子公司 | Method for determining electricity utilization behavior characteristics of user, and method and device for predicting electricity utilization load of user |
| GB2543281A (en) * | 2015-10-13 | 2017-04-19 | British Gas Trading Ltd | System for energy consumption prediction |
| CN105512768A (en) * | 2015-12-14 | 2016-04-20 | 上海交通大学 | User electricity consumption relevant factor identification and electricity consumption quantity prediction method under environment of big data |
| CN108846517B (en) * | 2018-06-12 | 2021-03-16 | 清华大学 | Integration method for predicating quantile probabilistic short-term power load |
| US11315044B2 (en) * | 2018-11-08 | 2022-04-26 | Vmware, Inc. | Multi dimensional scale analysis using machine learning |
-
2021
- 2021-01-26 CN CN202110102621.0A patent/CN112734135B/en active Active
Patent Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN107145976A (en) * | 2017-04-28 | 2017-09-08 | 北京科技大学 | A kind of method for predicting user power utilization load |
| CN110071502A (en) * | 2019-04-24 | 2019-07-30 | 广东工业大学 | A kind of calculation method of short-term electric load prediction |
| CN111028100A (en) * | 2019-11-29 | 2020-04-17 | 南方电网能源发展研究院有限责任公司 | Refined short-term load prediction method, device and medium considering meteorological factors |
| CN111832796A (en) * | 2020-02-29 | 2020-10-27 | 上海电力大学 | Fine classification and prediction method and system for residential electricity load mode |
| CN112215426A (en) * | 2020-10-16 | 2021-01-12 | 国网山东省电力公司信息通信公司 | Short-term power load prediction method |
Non-Patent Citations (2)
| Title |
|---|
| 基于小波回归分析法的短期负荷预测模型研究;闫冬梅等;《长春师范学院学报(自然科学版)》;20100420;第29卷(第04期);第20-24页 * |
| 基于聚类分析与随机森林的短期负荷滚动预测;荀港益;《智能城市》;20180514;第4卷(第09期);第9-11页 * |
Also Published As
| Publication number | Publication date |
|---|---|
| CN112734135A (en) | 2021-04-30 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN112734135B (en) | Power load prediction method, intelligent terminal and computer readable storage medium | |
| CN113962364B (en) | Multi-factor power load prediction method based on deep learning | |
| Shang et al. | Short-term load forecasting based on PSO-KFCM daily load curve clustering and CNN-LSTM model | |
| Poczęta et al. | Learning fuzzy cognitive maps using structure optimization genetic algorithm | |
| CN111027772A (en) | Multi-factor short-term load prediction method based on PCA-DBILSTM | |
| CN104598986A (en) | Big data based power load prediction method | |
| CN111028100A (en) | Refined short-term load prediction method, device and medium considering meteorological factors | |
| CN116186548B (en) | Power load prediction model training method and power load prediction method | |
| CN113255973A (en) | Power load prediction method, power load prediction device, computer equipment and storage medium | |
| Mustafa et al. | A new fuzzy grach model to forecast stock market technical analysis | |
| CN113887801A (en) | Building cold load prediction method, system, equipment and readable storage medium | |
| CN113361785A (en) | Power distribution network short-term load prediction method and device, terminal and storage medium | |
| CN113240153A (en) | Photovoltaic power generation data prediction method and device, computing equipment and storage medium | |
| CN114066071A (en) | Power parameter optimization method based on energy consumption, terminal equipment and storage medium | |
| CN112330052A (en) | Distribution transformer load prediction method | |
| CN113762591B (en) | Short-term electric quantity prediction method and system based on GRU and multi-core SVM countermeasure learning | |
| Yu et al. | Short-term cooling and heating loads forecasting of building district energy system based on data-driven models | |
| CN116011633B (en) | Regional gas consumption prediction method, regional gas consumption prediction system, regional gas consumption prediction equipment and Internet of things cloud platform | |
| CN117390550A (en) | Low-carbon park carbon emission dynamic prediction method and system considering emission training set | |
| Zhai et al. | Combining PSO-SVR and Random Forest Based Feature Selection for Day-ahead Peak Load Forecasting. | |
| KR102478684B1 (en) | Method for predicting energy consumption for using ensemble learning, and computing apparatus for performing the method | |
| CN114444763A (en) | Wind power prediction method based on AFSA-GNN | |
| CN114169763A (en) | Measuring instrument demand prediction method, system, computing device and storage medium | |
| CN112330017A (en) | Power load prediction method, power load prediction device, electronic device, and storage medium | |
| Gajowniczek et al. | Seasonal peak demand classification with machine learning techniques |
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 |