CN110674993A - User load short-term prediction method and device - Google Patents
User load short-term prediction method and device Download PDFInfo
- Publication number
- CN110674993A CN110674993A CN201910916662.6A CN201910916662A CN110674993A CN 110674993 A CN110674993 A CN 110674993A CN 201910916662 A CN201910916662 A CN 201910916662A CN 110674993 A CN110674993 A CN 110674993A
- Authority
- CN
- China
- Prior art keywords
- prediction
- load
- users
- model
- user
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Pending
Links
Images
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
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/23—Clustering techniques
- G06F18/232—Non-hierarchical techniques
- G06F18/2321—Non-hierarchical techniques using statistics or function optimisation, e.g. modelling of probability density functions
- G06F18/23213—Non-hierarchical techniques using statistics or function optimisation, e.g. modelling of probability density functions with fixed number of clusters, e.g. K-means clustering
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/24—Classification techniques
- G06F18/241—Classification techniques relating to the classification model, e.g. parametric or non-parametric approaches
-
- 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
-
- 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
- G06N3/044—Recurrent networks, e.g. Hopfield 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
Abstract
The embodiment of the application discloses a user load short-term prediction method and a user load short-term prediction device, wherein the method comprises the following steps: classifying users, and classifying the users into four categories of good power utilization stability, poor power utilization stability, low temperature sensitivity and high temperature sensitivity according to the power utilization stability and the temperature sensitivity of the users; predicting four types of users by adopting a plurality of prediction models, and selecting the prediction model with the highest efficiency as the main prediction model of the current type of users; and respectively adopting corresponding main prediction models to carry out short-term load prediction on the four types of users. The method classifies the users according to the power utilization characteristics of the users, and then selects the load prediction algorithm suitable for the characteristics of the users to predict, so that the situation that the prediction precision is reduced when a single load prediction method is applied to the users with different characteristics is avoided, meanwhile, the method has the characteristic of high prediction speed, and the problem that the prediction result with high precision and high speed is difficult to guarantee simultaneously for a large number of user loads in the current load prediction field is solved.
Description
Technical Field
The application relates to the technical field of power load prediction, in particular to a user load short-term prediction method and device.
Background
For the analysis of the power market demand, the short-term load prediction of the power system plays an important role, not only guarantees the safe and economic operation of the power system, but also is the basis for making a dispatching plan, a demand response mechanism and a trading plan under the market environment. Accurate power utilization prediction can economically and reasonably arrange the start and stop of a unit in a power grid, the safety and stability of the operation of the power grid are kept, the popularization of a market competition mechanism is facilitated, the further reformation of a power market is promoted, and therefore on the premise that the normal life and production activities of the society are guaranteed, the friendly interaction of the supply and demand of a power system is promoted, and the economic benefit and the social benefit are improved. However, in the current load prediction field, it is difficult to maintain a high-precision prediction result for a large amount of user loads, in order to improve the load prediction precision, most load prediction algorithms sacrifice prediction time to a certain extent, and the prediction model takes much time to train the model, thus occupying more computing resources.
Disclosure of Invention
The embodiment of the application provides a user load short-term prediction method and device, and solves the problem that in the current load prediction field, a prediction result with high precision and high speed cannot be ensured simultaneously aiming at a large number of user loads
In view of this, the first aspect of the present application provides a short-term user load prediction method, including:
classifying users, and classifying the users into four categories of good power utilization stability, poor power utilization stability, low temperature sensitivity and high temperature sensitivity according to the power utilization stability and the temperature sensitivity of the users;
predicting four types of users by adopting a plurality of prediction models, and selecting the prediction model with the highest efficiency as the main prediction model of the current type of users;
and respectively adopting corresponding main prediction models to carry out short-term load prediction on the four types of users.
Preferably, before classifying the user, the method further includes:
collecting historical load data of a user and temperature data of a corresponding date;
and identifying the user by adopting a clustering analysis method based on the historical load data and the temperature data, and determining the power utilization stability and the temperature sensitivity of the user.
Preferably, the respectively performing short-term load prediction on the four types of users by using the corresponding main prediction models specifically comprises:
aiming at four types of users with good power utilization stability, poor power utilization stability, low temperature sensitivity and high temperature sensitivity, an improved support vector machine prediction model, a multi-model fuzzy comprehensive model, a clustered LSTM prediction model and a load partition prediction model are respectively adopted to carry out short-term load prediction.
Preferably, the multi-model fuzzy comprehensive prediction model comprises:
obtaining typical load patterns reflecting the characteristics of the user load patterns by adopting a load pattern analysis algorithm based on clustering, and grouping the typical load patterns according to the distance from the sample to the centroid;
establishing a unit sub-prediction model corresponding to each typical load mode group by the neural network based on a neural network learning algorithm;
and analyzing and determining the weight of the unit sub-prediction model according to the similarity between the input at the moment to be predicted and the corresponding input of the typical load mode class corresponding to each unit sub-prediction model.
And adding the products of the weight and the predicted value of each unit sub-prediction model to obtain a multi-model fuzzy comprehensive prediction result.
Preferably, after the load short-term prediction is performed on the four types of users by using the corresponding main prediction models, the method further includes:
and storing the prediction result into a database and/or carrying out graphical display.
A second aspect of the present application provides a short-term user load prediction apparatus, including:
the classification unit is used for classifying the users, and classifying the users into four categories of good power utilization stability, poor power utilization stability, low temperature sensitivity and high temperature sensitivity according to the power utilization stability and the temperature sensitivity of the users;
the selection unit is used for predicting four types of users by adopting a plurality of prediction models, and selecting the prediction model with the highest efficiency as the main prediction model of the current type of users;
and the prediction unit is used for performing short-term load prediction on the four types of users by adopting corresponding main prediction models.
Preferably, the method further comprises the following steps:
the acquisition unit is used for acquiring historical load data of a user and temperature data of a corresponding date;
and the identification unit is used for identifying the user by adopting a clustering analysis method based on the historical load data and the temperature data and determining the power utilization stability and the temperature sensitivity of the user.
Preferably, the prediction unit is specifically configured to: aiming at four types of users with good power utilization stability, poor power utilization stability, low temperature sensitivity and high temperature sensitivity, an improved support vector machine prediction model, a multi-model fuzzy comprehensive model, a clustered LSTM prediction model and a load partition prediction model are respectively adopted to carry out short-term load prediction.
Preferably, the multi-model fuzzy comprehensive prediction model comprises:
the grouping unit is used for obtaining typical load patterns reflecting the characteristics of the user load patterns by adopting a load pattern analysis algorithm based on clustering, and grouping the typical load patterns according to the distance from the sample to the center of mass;
the establishing unit is used for establishing a unit sub-prediction model corresponding to each typical load mode group of the neural network based on a neural network learning algorithm;
and the analysis unit is used for analyzing and determining the weight of the unit sub-prediction model according to the similarity between the input at the moment to be predicted and the corresponding input of the typical load mode class corresponding to each unit sub-prediction model.
And the integration unit is used for adding the products of the weight and the predicted value of each unit sub-prediction model to obtain a multi-model fuzzy comprehensive prediction result.
Preferably, the method further comprises the following steps:
and the storage or display unit is used for storing the prediction result into a database and/or carrying out graphical display.
According to the technical scheme, the embodiment of the application has the following advantages:
the embodiment of the application provides a user load short-term prediction method which comprises the steps of classifying users, and classifying the users into four categories of good power utilization stability, poor power utilization stability, low temperature sensitivity and high temperature sensitivity according to the power utilization stability and the temperature sensitivity of the users; predicting four types of users by adopting a plurality of prediction models, and selecting the prediction model with the highest efficiency as the main prediction model of the current type of users; and respectively adopting corresponding main prediction models to carry out short-term load prediction on the four types of users.
According to the user load short-term prediction method provided by the embodiment of the application, users are classified according to the electricity utilization characteristics of the users, and then the load prediction algorithm suitable for the characteristics of the users is selected for prediction, so that the situation that the prediction precision is reduced when a single load prediction method is applied to users with different characteristics is avoided, meanwhile, the method has the characteristic of high prediction speed, and the problem that the prediction result with high precision and high speed is difficult to guarantee simultaneously for a large number of user loads in the current load prediction field is solved.
Drawings
FIG. 1 is a flowchart of a method for short-term user load prediction in a first embodiment of the present application;
FIG. 2 is a flowchart of a method for short term user load prediction in a second embodiment of the present application;
fig. 3 is a flowchart of load prediction for a user with low temperature sensitivity according to an embodiment of the present disclosure.
Detailed Description
In order to make the technical solutions of the present application better understood, the technical solutions in the embodiments of the present application will be clearly and completely described below with reference to the drawings in the embodiments of the present application, and it is obvious that the described embodiments are only a part of the embodiments of the present application, and not all of the embodiments. All other embodiments, which can be derived by a person skilled in the art from the embodiments given herein without making any creative effort, shall fall within the protection scope of the present application.
The application provides a user load short-term prediction method in a first aspect.
Referring to fig. 1, fig. 1 is a flowchart illustrating a method for short-term user load prediction in a first embodiment of the present application, which specifically includes:
and 101, classifying users, and classifying the users into four categories, namely good power utilization stability, poor power utilization stability, low temperature sensitivity and high temperature sensitivity according to the power utilization stability and the temperature sensitivity of the users.
In actual engineering application, a single prediction algorithm cannot meet load prediction requirements of massive users, and high-precision prediction results are difficult to maintain for user loads of different types and different electricity utilization characteristics. On the other hand, in order to improve the load prediction accuracy, most load prediction algorithms sacrifice prediction time to a certain extent, and the prediction model takes more time to train the model, thereby occupying more calculation resources. Therefore, the embodiment of the present application provides a prediction method for predicting different types of users by using different prediction models, and firstly, the users need to be classified.
And 102, predicting the four types of users by adopting a plurality of prediction models, and selecting the prediction model with the highest efficiency as the main prediction model of the current type of users.
The existing user load short-term prediction models are various, and in order to select a prediction model with high precision and high speed (namely high efficiency) for different types of users, the users with different types need to be predicted by using a plurality of prediction models respectively, so that an optimal model is selected as a main prediction model of the users.
And 103, respectively carrying out short-term load prediction on the four types of users by adopting corresponding main prediction models.
After the suitable models are respectively selected for the four types of users, the corresponding models can be directly used for short-term load prediction.
The user load short-term prediction method provided by the embodiment of the application is used for classifying users according to the electricity utilization characteristics of the users aiming at general user loads, and then selecting the load prediction algorithm suitable for the characteristics of the users to predict, so that the situation that the prediction precision is reduced when a single load prediction method is applied to users with different characteristics is avoided, meanwhile, the method has the characteristic of high prediction speed, and the problem that the prediction result with high precision and high speed is difficult to guarantee simultaneously aiming at a large number of user loads in the current load prediction field is solved.
Referring to fig. 2, fig. 2 is a flowchart illustrating a method for short-term user load prediction in a second embodiment of the present application, which specifically includes:
For example, the historical load data of the user is the load value of 96 sampling points (one sampling point in 15 minutes) each day 120 days before the prediction day (support vector machine prediction and multi-model fuzzy comprehensive prediction) or 60 days (cluster LSTM prediction and load partition prediction). However, in practical applications, the real-time load collection system updates the historical load data in batches, which may result in data missing in the last three days, and the closer to the current moment, the more load data that is not collected. Therefore, the load value at the missing moment of the last three days can be correspondingly supplemented by adopting the data (updated and complete) of the last fourth day, so that complete load historical data for training the prediction model can be obtained.
In addition, in consideration of the difference in power consumption characteristics between the weekday and the holiday, date type data may be included: the date type reflects the nature of the date, and Monday through Sunday are represented by 1-7, respectively.
The temperature data is the average air temperature per day corresponding to the historical load data.
The data are divided into data for prediction model training and data for prediction. The historical load data is training data, the date type and temperature data before the prediction day (excluding the prediction day) are training data, and the date type and temperature data seven days after the prediction day (including the prediction day) are prediction data.
And 202, identifying the user by adopting a clustering analysis method based on the historical load data and the temperature data, and determining the power utilization stability and the temperature sensitivity of the user.
The user electricity consumption behavior identification method provided by the patent with the application number of CN201910075483.4 can be adopted, the electricity stability and the temperature sensitivity are used as two indexes of classification, and the classification is carried out on the user through cluster analysis. And obtaining that the user belongs to the load with good power utilization stability, poor power utilization stability, low temperature sensitivity or high temperature sensitivity.
And step 203, classifying the users, and classifying the users into four categories of good power utilization stability, poor power utilization stability, low temperature sensitivity and high temperature sensitivity according to the power utilization stability and the temperature sensitivity of the users.
And step 204, adopting a plurality of prediction models for predicting four types of users, and selecting the prediction model with the highest efficiency as the main prediction model of the current type of users.
And step 205, aiming at four types of users with good power utilization stability, poor power utilization stability, low temperature sensitivity and high temperature sensitivity, respectively adopting an improved support vector machine prediction model, a multi-model fuzzy comprehensive model, a clustered LSTM prediction model and a load partition prediction model to carry out load short-term prediction.
(1) Improved support vector machine prediction
For the user load with good power utilization stability, the prediction speed is increased on the premise of ensuring the accuracy requirement as much as possible, the running time of the whole algorithm is shortened, and the operation resources are saved. For this purpose, a least square method improved support vector machine algorithm is adopted for load prediction.
The method of the least square support vector machine is to train and predict after preprocessing data, the training process also follows the principle of minimizing structural risk, the inequality constraint is changed into the equality constraint, the empirical risk is changed into the quadratic from the first power of deviation, the problem of solving quadratic programming is converted into the solution of a linear equation set, the insensitive loss function is avoided, the calculation complexity is greatly reduced, and the operation speed is higher than that of the general support vector machine.
For a given training data set { (x)k,yk) 1, 2., N }, where x isk∈RnTo input data, yk∈RnIs the output value. Using a non-linear mapping phi (-) to transform samples from the original space RnMapping to a feature space phi (x)i) And constructing an optimal decision function in a high-dimensional feature space:
in the formula (I), the compound is shown in the specification,is the kernel-space mapping function, ω is the weight vector and b is a constant.
Using the principle of minimizing the structural risk, finding ω, b is the minimization, and the objective optimization function is:
the constraint conditions are as follows:
k=1,...,N
wherein gamma is a normalization parameter; e.g. of the typekIs the relaxation variable.
To solve the above problem, a lagrangian function is defined:
in the formula, Lagrange multiplier alphakE.g. R. The above formula is optimized, i.e. omega, b, ek,αkIs equal to 0.
The above problem translates to solving a linear equation:
wherein y ═ y.. multidot.yt]T,It=[1,...,1]T,α=[α1,...αt]T,Where α, b is obtained by solving the above equation.
The function estimate of the least squares support vector machine is:
in the formula, K (x, x)k) For the kernel Function, the present invention selects a gaussian Radial Basis Function (RBF) as the kernel Function, that is:
where σ is a width parameter of the function.
(2) Multi-model fuzzy comprehensive prediction
For users with poor power utilization stability, the power utilization mode of the users needs to be deeply analyzed, so that the prediction precision is improved. Therefore, a multi-model fuzzy comprehensive prediction method is adopted to predict the load of the users.
The multi-model fuzzy comprehensive prediction method mainly comprises 4 links:
1) and obtaining typical load patterns reflecting the characteristics of the user load patterns by adopting a load pattern analysis algorithm based on clustering, and grouping the typical load patterns according to the distance from the sample to the centroid.
Firstly, a load pattern analysis algorithm based on K-means clustering is given, a small amount of outlier abnormal data can be eliminated by the algorithm, a typical load pattern reflecting the characteristics of the user load pattern is obtained, and grouping is carried out according to the distance as a similarity measure:
i, randomly extracting K objects from an object set to serve as initial clustering centers;
II, respectively calculating Euclidean distances from all the objects to all the clusters, and after mutual comparison, attributing the objects to the class with the minimum distance;
III, calculating a mean value of each category according to the initial classification obtained by the II to update a clustering center;
IV, repeating II and III according to the new clustering center until all object attributions are not changed;
v outlier identification: i, calculating the maximum distance maxD and the minimum distance minD from the sample to the corresponding centroid, and then calculating the difference ranging between the maximum distance and the minimum distance; ii determining a proportionality coefficient U; and iii, when the distance between the sample and the corresponding centroid is greater than range D U, judging the data as outlier data, and removing the outlier data from the training sample.
2) And establishing a unit sub-prediction model corresponding to each typical load mode group by the neural network based on a neural network learning algorithm.
Then, a neural network learning algorithm based on conjugate gradient is given, and the RBF neural network is trained respectively to establish unit sub-prediction models corresponding to each typical load pattern group.
As a common neural network model, the RBF network has the characteristics of simple topological structure and high learning speed. The selection of parameters such as the central component cji (t), the width dji (t) of the RBF neural network directly influences the prediction performance. Model training was performed according to the following steps:
i, initializing parameters of a neural network;
II, calculating the output of the hidden layer neuron and the output layer neuron;
III, calculating the root mean square error output by the network, finishing training if the root mean square error is smaller than a set error threshold, and otherwise, entering the next step;
IV, iterative calculation is carried out, and weight, center and width parameters are adjusted;
and V, if the iteration times are larger than the set iteration time upper limit, finishing the training, otherwise, adding one to the iteration times, and returning to the step II.
3) And analyzing and determining the weight of the unit sub-prediction model according to the similarity between the input at the moment to be predicted and the corresponding input of the typical load mode class corresponding to each unit sub-prediction model.
Then, a sub-model weighting method based on similarity is provided:
a weighting method based on similarity analysis is adopted, namely, the weight of the sub-model is determined according to the similarity analysis between the input of the moment to be predicted and the corresponding input of the typical load model class corresponding to each sub-model.
Assuming that the similarity from the time input x to be predicted to the j-th class typical load pattern class center cj (x1, x2, …, xd) is:
then the jth sub-model weight is:
in the formula, qhThe weight of the h component of the input data at the time to be predicted can be selected according to the principle of big or small
4) And adding the products of the weight and the predicted value of each unit sub-prediction model to obtain a multi-model fuzzy comprehensive prediction result.
And adding the single model predicted values obtained by calculation according to the weights to obtain the prediction result of the multi-model fuzzy comprehensive algorithm.
(3) Clustered LSTM prediction
For users with low temperature sensitivity, considering that the size of the electric load is less influenced by the temperature and is possibly more influenced by other factors such as holidays (date types), product sales and the like, a multi-model LSTM-based short-term load prediction method is adopted, and the prediction result is finally obtained by a load cluster analysis, a load-related factor-based prediction classifier, a multi-model load LSTM load prediction algorithm and the like. For a specific load prediction method, reference may be made to patent application No. CN 201910133236.5.
A flow chart of the prediction of this method is shown in fig. 3. Because the output of the neural network is very sensitive to the input data, the input data needs to be normalized and converted into data in the range of (0, 1), and a one hot encoding method is adopted for the temperature and date types. The input matrix X is formed by combining an active load value curve L (sampled every 15 minutes), a predicted value T of the air temperature on the day of the prediction day, and a predicted day and week number C (1-7), that is, X is [ L, T, C ]. The forecast value is adopted for the temperature acquisition of the day of forecast, and the weather forecast technology is mainly considered to make more accurate forecast within 24-72 h.
(4) Load partition prediction
For the user load with high temperature sensitivity, a load partition prediction method based on meteorological information is adopted. Along with the expansion of the geographical range, the distribution difference of external parameters is obvious, the accuracy of the simple single-region aggregation prediction result is reduced, and the error between the aggregation result and the actual situation is larger. In order to obtain a more accurate user load aggregation prediction model and more accurately describe the working characteristics and power requirements of a load group in a larger geographic range, a partition aggregation method is adopted on the basis of single-region aggregation in consideration of different parameter regional distributions.
According to the weather partitioning method based on the Thiessen polygon, the geographical range of the user load group participating in aggregation is divided into areas, the only weather observation station provides weather parameter information such as external environment temperature and the like for each sub-area, parameter differences of different geographic positions of the load can be reflected more accurately, the loads with similar parameter distribution characteristics of each area are aggregated, and then the aggregation results of the sub-areas are subjected to secondary aggregation on the basis, so that the load aggregation results are more accurate and reliable, and the prediction precision is improved.
Therefore, after the user load is partitioned according to the meteorological information, the cluster LSTM prediction is carried out on the user load in the sub-regions, and then the predicted user load of the whole region is obtained by adding the prediction results of the sub-regions.
And step 206, storing the prediction result into a database and/or carrying out graphical display.
After the user is identified and classified, the prediction result obtained by predicting according to the corresponding prediction algorithm is the load value of 96 points per day on the prediction date and six days (seven days in total) after the prediction date. The prediction result can be stored in a database, and can also be referred by a user through a direct calling or graphical display method.
Example (b):
the method comprises the steps of adopting load data of Fushan City in Guangdong province in 2017 years to carry out prediction and method verification, using one user of each of four users with good power utilization stability, poor power utilization stability, low temperature sensitivity and high temperature sensitivity obtained by user identification as a prediction object, selecting 8-month-1 day in 2017 as a prediction day, predicting to obtain load data of 8-month-1-8-month-7 days, comparing the load data with actual load data, and calculating the average relative error of seven-day prediction results, wherein the obtained results are shown in the following table.
TABLE 1 prediction case error Table
According to the table, the improved support vector machine algorithm can predict the user with better power utilization stability with the smallest error, and similarly, the multi-model fuzzy comprehensive prediction, the clustering LSTM and the load partition prediction are respectively used for predicting the user with poorer power utilization stability, the user with lower temperature sensitivity and the user with higher temperature sensitivity with the smallest error. On the other hand, embodiments also illustrate that a single algorithm is difficult to maintain consistently high accuracy for user loads of different characteristics.
The computer used for testing adopts Intel i5-7200U, 2.5GHz and 2.71GHz dual-core processors, the running memory is 4GB, and the running time of each algorithm obtained by testing is shown in the following table.
TABLE 2 predicted temporal comparisons
In conclusion, the user load short-term prediction method considering the user electricity utilization characteristic classification can effectively predict users with different electricity utilization characteristics, and has higher precision on user load prediction with different electricity utilization stationarities and different temperature sensitivities. Meanwhile, the method is high in prediction speed, suitable for engineering application and obvious in short-term rapid prediction advantage for massive users.
A second aspect of the present application provides a short-term user load prediction apparatus, including:
the classification unit is used for classifying the users, and classifying the users into four categories of good power utilization stability, poor power utilization stability, low temperature sensitivity and high temperature sensitivity according to the power utilization stability and the temperature sensitivity of the users;
the selection unit is used for predicting four types of users by adopting a plurality of prediction models, and selecting the prediction model with the minimum error value as the main prediction model of the current type of users;
and the prediction unit is used for performing short-term load prediction on the four types of users by adopting corresponding main prediction models.
Further, the method also comprises the following steps:
the acquisition unit is used for acquiring historical load data of a user and temperature data of a corresponding date;
and the identification unit is used for identifying the user by adopting a clustering analysis method based on the historical load data and the temperature data and determining the power utilization stability and the temperature sensitivity of the user.
Further, the prediction unit is specifically configured to: aiming at four types of users with good power utilization stability, poor power utilization stability, low temperature sensitivity and high temperature sensitivity, an improved support vector machine prediction model, a multi-model fuzzy comprehensive model, a clustered LSTM prediction model and a load partition prediction model are respectively adopted to carry out short-term load prediction.
Further, the multi-model fuzzy comprehensive prediction model comprises:
the grouping unit is used for obtaining typical load patterns reflecting the characteristics of the user load patterns by adopting a load pattern analysis algorithm based on clustering, and grouping the typical load patterns according to the distance from the sample to the center of mass;
the establishing unit is used for establishing a unit sub-prediction model corresponding to each typical load mode group of the neural network based on a neural network learning algorithm;
and the analysis unit is used for analyzing and determining the weight of the unit sub-prediction model according to the similarity between the input at the moment to be predicted and the corresponding input of the typical load mode class corresponding to each unit sub-prediction model.
And the integration unit is used for adding the products of the weight and the predicted value of each unit sub-prediction model to obtain a multi-model fuzzy comprehensive prediction result.
Further, the method also comprises the following steps:
and the storage or display unit is used for storing the prediction result into a database and/or carrying out graphical display.
It is clear to those skilled in the art that, for convenience and brevity of description, the specific working processes of the above-described systems, apparatuses and units may refer to the corresponding processes in the foregoing method embodiments, and are not described herein again.
The terms "first," "second," "third," "fourth," and the like in the description of the application and the above-described figures, if any, are used for distinguishing between similar elements and not necessarily for describing a particular sequential or chronological order. It is to be understood that the data so used is interchangeable under appropriate circumstances such that the embodiments of the application described herein are, for example, capable of operation in sequences other than those illustrated or otherwise described herein. Furthermore, the terms "comprises," "comprising," and "having," and any variations thereof, are intended to cover a non-exclusive inclusion, such that a process, method, system, article, or apparatus that comprises a list of steps or elements is not necessarily limited to those steps or elements expressly listed, but may include other steps or elements not expressly listed or inherent to such process, method, article, or apparatus.
It should be understood that in the present application, "at least one" means one or more, "a plurality" means two or more. "and/or" for describing an association relationship of associated objects, indicating that there may be three relationships, e.g., "a and/or B" may indicate: only A, only B and both A and B are present, wherein A and B may be singular or plural. The character "/" generally indicates that the former and latter associated objects are in an "or" relationship. "at least one of the following" or similar expressions refer to any combination of these items, including any combination of single item(s) or plural items. For example, at least one (one) of a, b, or c, may represent: a, b, c, "a and b", "a and c", "b and c", or "a and b and c", wherein a, b, c may be single or plural.
In the several embodiments provided in the present application, it should be understood that the disclosed system, apparatus and method may be implemented in other manners. For example, the above-described apparatus embodiments are merely illustrative, and for example, the division of the units is only one logical division, and other divisions may be realized in practice, for example, a plurality of units or components may be combined or integrated into another system, or some features may be omitted, or not executed. In addition, the shown or discussed mutual coupling or direct coupling or communication connection may be an indirect coupling or communication connection through some interfaces, devices or units, and may be in an electrical, mechanical or other form.
The units described as separate parts may or may not be physically separate, and parts displayed as units may or may not be physical units, may be located in one place, or may be distributed on a plurality of network units. Some or all of the units can be selected according to actual needs to achieve the purpose of the solution of the embodiment.
In addition, functional units in the embodiments of the present application may be integrated into one processing unit, or each unit may exist alone physically, or two or more units are integrated into one unit. The integrated unit can be realized in a form of hardware, and can also be realized in a form of a software functional unit.
The integrated unit, if implemented in the form of a software functional unit and sold or used as a stand-alone product, may be stored in a computer readable storage medium. Based on such understanding, the technical solution of the present application may be substantially implemented or contributed to by the prior art, or all or part of the technical solution may be embodied in a software product, which is stored in a storage medium and includes instructions for causing a computer device (which may be a personal computer, a server, or a network device) to execute all or part of the steps of the method according to the embodiments of the present application. And the aforementioned storage medium includes: a U-disk, a removable hard disk, a Read-only Memory (ROM), a Random Access Memory (RAM), a magnetic disk or an optical disk, and other various media capable of storing program codes.
The above embodiments are only used for illustrating the technical solutions of the present application, and not for limiting the same; although the present application has been described in detail with reference to the foregoing embodiments, it should be understood by those of ordinary skill in the art that: the technical solutions described in the foregoing embodiments may still be modified, or some technical features may be equivalently replaced; and such modifications or substitutions do not depart from the spirit and scope of the corresponding technical solutions in the embodiments of the present application.
Claims (10)
1. A user load short-term prediction method is characterized by comprising the following steps:
classifying users, and classifying the users into four categories of good power utilization stability, poor power utilization stability, low temperature sensitivity and high temperature sensitivity according to the power utilization stability and the temperature sensitivity of the users;
predicting four types of users by adopting a plurality of prediction models, and selecting the prediction model with the highest efficiency as the main prediction model of the current type of users;
and respectively adopting corresponding main prediction models to carry out short-term load prediction on the four types of users.
2. The method of claim 1, further comprising, before classifying the users:
collecting historical load data of a user and temperature data of a corresponding date;
and identifying the user by adopting a clustering analysis method based on the historical load data and the temperature data, and determining the power utilization stability and the temperature sensitivity of the user.
3. The user load short-term prediction method according to claim 1, wherein the performing of the load short-term prediction on the four types of users respectively by using the corresponding main prediction models specifically comprises:
aiming at four types of users with good power utilization stability, poor power utilization stability, low temperature sensitivity and high temperature sensitivity, an improved support vector machine prediction model, a multi-model fuzzy comprehensive model, a clustered LSTM prediction model and a load partition prediction model are respectively adopted to carry out short-term load prediction.
4. The short-term user load prediction method according to claim 3, wherein the multi-model fuzzy comprehensive prediction model comprises:
obtaining typical load patterns reflecting the characteristics of the user load patterns by adopting a load pattern analysis algorithm based on clustering, and grouping the typical load patterns according to the distance from the sample to the centroid;
establishing a unit sub-prediction model corresponding to each typical load mode group by the neural network based on a neural network learning algorithm;
analyzing and determining the weight of the unit sub-prediction model according to the similarity between the input at the moment to be predicted and the corresponding input of the typical load mode class corresponding to each unit sub-prediction model;
and adding the products of the weight and the predicted value of each unit sub-prediction model to obtain a multi-model fuzzy comprehensive prediction result.
5. The method according to claim 1, wherein after the short-term load prediction is performed on the four types of users by using the corresponding main prediction models, the method further comprises:
and storing the prediction result into a database and/or carrying out graphical display.
6. A user load short-term prediction apparatus, comprising:
the classification unit is used for classifying the users, and classifying the users into four categories of good power utilization stability, poor power utilization stability, low temperature sensitivity and high temperature sensitivity according to the power utilization stability and the temperature sensitivity of the users;
the selection unit is used for predicting four types of users by adopting a plurality of prediction models, and selecting the prediction model with the highest efficiency as the main prediction model of the current type of users;
and the prediction unit is used for performing short-term load prediction on the four types of users by adopting corresponding main prediction models.
7. The apparatus for short-term prediction of user load according to claim 6, further comprising:
the acquisition unit is used for acquiring historical load data of a user and temperature data of a corresponding date;
and the identification unit is used for identifying the user by adopting a clustering analysis method based on the historical load data and the temperature data and determining the power utilization stability and the temperature sensitivity of the user.
8. The device according to claim 6, wherein the prediction unit is specifically configured to: aiming at four types of users with good power utilization stability, poor power utilization stability, low temperature sensitivity and high temperature sensitivity, an improved support vector machine prediction model, a multi-model fuzzy comprehensive model, a clustered LSTM prediction model and a load partition prediction model are respectively adopted to carry out short-term load prediction.
9. The apparatus according to claim 8, wherein the multi-model fuzzy comprehensive predictive model comprises:
the grouping unit is used for obtaining typical load patterns reflecting the characteristics of the user load patterns by adopting a load pattern analysis algorithm based on clustering, and grouping the typical load patterns according to the distance from the sample to the center of mass;
the establishing unit is used for establishing a unit sub-prediction model corresponding to each typical load mode group of the neural network based on a neural network learning algorithm;
the analysis unit is used for analyzing and determining the weight of the unit sub-prediction models according to the similarity between the input at the moment to be predicted and the corresponding input of the typical load mode class corresponding to each unit sub-prediction model;
and the integration unit is used for adding the products of the weight and the predicted value of each unit sub-prediction model to obtain a multi-model fuzzy comprehensive prediction result.
10. The apparatus according to claim 6, further comprising:
and the storage or display unit is used for storing the prediction result into a database and/or carrying out graphical display.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201910916662.6A CN110674993A (en) | 2019-09-26 | 2019-09-26 | User load short-term prediction method and device |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201910916662.6A CN110674993A (en) | 2019-09-26 | 2019-09-26 | User load short-term prediction method and device |
Publications (1)
Publication Number | Publication Date |
---|---|
CN110674993A true CN110674993A (en) | 2020-01-10 |
Family
ID=69079300
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN201910916662.6A Pending CN110674993A (en) | 2019-09-26 | 2019-09-26 | User load short-term prediction method and device |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN110674993A (en) |
Cited By (10)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN111582548A (en) * | 2020-04-14 | 2020-08-25 | 广东卓维网络有限公司 | Power load prediction method based on multivariate user behavior portrait |
CN111582911A (en) * | 2020-04-14 | 2020-08-25 | 广东卓维网络有限公司 | Multi-user and power grid friendly interactive power utilization system |
CN111600304A (en) * | 2020-06-17 | 2020-08-28 | 广东工业大学 | Building power scheduling method, device and equipment |
CN111612275A (en) * | 2020-05-29 | 2020-09-01 | 云南电网有限责任公司 | Method and device for predicting load of regional user |
CN112183634A (en) * | 2020-09-29 | 2021-01-05 | 国网山东省电力公司冠县供电公司 | User demand response method, device, terminal and storage medium |
CN112330077A (en) * | 2021-01-04 | 2021-02-05 | 南方电网数字电网研究院有限公司 | Power load prediction method, power load prediction device, computer equipment and storage medium |
CN112348070A (en) * | 2020-10-30 | 2021-02-09 | 广东电网有限责任公司电力调度控制中心 | Method and system for forecasting medium and short term loads of smart power grid |
WO2021259540A1 (en) * | 2020-06-22 | 2021-12-30 | Siemens Aktiengesellschaft | Method for power prediction of an energy system |
CN113988436A (en) * | 2021-11-01 | 2022-01-28 | 广西电网有限责任公司 | Power consumption prediction method based on LSTM neural network and hierarchical relation correction |
CN114358213A (en) * | 2022-03-08 | 2022-04-15 | 湖南大学 | Error ablation processing method, system and medium for nonlinear time series data prediction |
Citations (10)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN101145225A (en) * | 2007-10-16 | 2008-03-19 | 刘吉成 | Middle and long-term power load forecasting and pre-alarming system |
CN103218675A (en) * | 2013-05-06 | 2013-07-24 | 国家电网公司 | Short-term load prediction method based on clustering and sliding window |
CN105654196A (en) * | 2015-12-29 | 2016-06-08 | 中国电力科学研究院 | Adaptive load prediction selection method based on electric power big data |
CN105678404A (en) * | 2015-12-30 | 2016-06-15 | 东北大学 | Micro-grid load prediction system and method based on electricity purchased on-line and dynamic correlation factor |
CN105989420A (en) * | 2015-02-12 | 2016-10-05 | 西门子公司 | Method of determining user electricity consumption behavior features, method of predicting user electricity consumption load and device |
CN106127360A (en) * | 2016-06-06 | 2016-11-16 | 国网天津市电力公司 | A kind of multi-model load forecasting method analyzed based on user personality |
CN107609712A (en) * | 2017-09-30 | 2018-01-19 | 新智能源系统控制有限责任公司 | A kind of steer method and apparatus based on classification prediction |
CN108932671A (en) * | 2018-06-06 | 2018-12-04 | 上海电力学院 | A kind of LSTM wind-powered electricity generation load forecasting method joined using depth Q neural network tune |
CN109754113A (en) * | 2018-11-29 | 2019-05-14 | 南京邮电大学 | Load forecasting method based on dynamic time warping Yu length time memory |
CN110263866A (en) * | 2019-06-24 | 2019-09-20 | 苏州智睿新能信息科技有限公司 | A kind of power consumer load setting prediction technique based on deep learning |
-
2019
- 2019-09-26 CN CN201910916662.6A patent/CN110674993A/en active Pending
Patent Citations (10)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN101145225A (en) * | 2007-10-16 | 2008-03-19 | 刘吉成 | Middle and long-term power load forecasting and pre-alarming system |
CN103218675A (en) * | 2013-05-06 | 2013-07-24 | 国家电网公司 | Short-term load prediction method based on clustering and sliding window |
CN105989420A (en) * | 2015-02-12 | 2016-10-05 | 西门子公司 | Method of determining user electricity consumption behavior features, method of predicting user electricity consumption load and device |
CN105654196A (en) * | 2015-12-29 | 2016-06-08 | 中国电力科学研究院 | Adaptive load prediction selection method based on electric power big data |
CN105678404A (en) * | 2015-12-30 | 2016-06-15 | 东北大学 | Micro-grid load prediction system and method based on electricity purchased on-line and dynamic correlation factor |
CN106127360A (en) * | 2016-06-06 | 2016-11-16 | 国网天津市电力公司 | A kind of multi-model load forecasting method analyzed based on user personality |
CN107609712A (en) * | 2017-09-30 | 2018-01-19 | 新智能源系统控制有限责任公司 | A kind of steer method and apparatus based on classification prediction |
CN108932671A (en) * | 2018-06-06 | 2018-12-04 | 上海电力学院 | A kind of LSTM wind-powered electricity generation load forecasting method joined using depth Q neural network tune |
CN109754113A (en) * | 2018-11-29 | 2019-05-14 | 南京邮电大学 | Load forecasting method based on dynamic time warping Yu length time memory |
CN110263866A (en) * | 2019-06-24 | 2019-09-20 | 苏州智睿新能信息科技有限公司 | A kind of power consumer load setting prediction technique based on deep learning |
Non-Patent Citations (1)
Title |
---|
谷云东等: "大用户电力负荷的多模型模糊综合预测", 《电工技术学报》 * |
Cited By (15)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN111582911A (en) * | 2020-04-14 | 2020-08-25 | 广东卓维网络有限公司 | Multi-user and power grid friendly interactive power utilization system |
CN111582911B (en) * | 2020-04-14 | 2023-06-30 | 广东卓维网络有限公司 | Friendly interactive power utilization system for multiple users and power grid |
CN111582548A (en) * | 2020-04-14 | 2020-08-25 | 广东卓维网络有限公司 | Power load prediction method based on multivariate user behavior portrait |
CN111612275B (en) * | 2020-05-29 | 2022-04-01 | 云南电网有限责任公司 | Method and device for predicting load of regional user |
CN111612275A (en) * | 2020-05-29 | 2020-09-01 | 云南电网有限责任公司 | Method and device for predicting load of regional user |
CN111600304A (en) * | 2020-06-17 | 2020-08-28 | 广东工业大学 | Building power scheduling method, device and equipment |
CN111600304B (en) * | 2020-06-17 | 2021-11-26 | 广东工业大学 | Building power scheduling method, device and equipment |
WO2021259540A1 (en) * | 2020-06-22 | 2021-12-30 | Siemens Aktiengesellschaft | Method for power prediction of an energy system |
CN112183634A (en) * | 2020-09-29 | 2021-01-05 | 国网山东省电力公司冠县供电公司 | User demand response method, device, terminal and storage medium |
CN112348070A (en) * | 2020-10-30 | 2021-02-09 | 广东电网有限责任公司电力调度控制中心 | Method and system for forecasting medium and short term loads of smart power grid |
CN112330077A (en) * | 2021-01-04 | 2021-02-05 | 南方电网数字电网研究院有限公司 | Power load prediction method, power load prediction device, computer equipment and storage medium |
CN113988436A (en) * | 2021-11-01 | 2022-01-28 | 广西电网有限责任公司 | Power consumption prediction method based on LSTM neural network and hierarchical relation correction |
CN113988436B (en) * | 2021-11-01 | 2023-04-28 | 广西电网有限责任公司 | Power consumption prediction method based on LSTM neural network and hierarchical relation correction |
CN114358213A (en) * | 2022-03-08 | 2022-04-15 | 湖南大学 | Error ablation processing method, system and medium for nonlinear time series data prediction |
CN114358213B (en) * | 2022-03-08 | 2022-06-10 | 湖南大学 | Error ablation processing method, system and medium for nonlinear time series data prediction |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN110674993A (en) | User load short-term prediction method and device | |
US11043808B2 (en) | Method for identifying pattern of load cycle | |
Cai et al. | Predicting the energy consumption of residential buildings for regional electricity supply-side and demand-side management | |
CN111324642A (en) | Model algorithm type selection and evaluation method for power grid big data analysis | |
Li et al. | Midterm load forecasting: A multistep approach based on phase space reconstruction and support vector machine | |
CN111178611A (en) | Method for predicting daily electric quantity | |
Zeng et al. | A learning framework based on weighted knowledge transfer for holiday load forecasting | |
US20210326696A1 (en) | Method and apparatus for forecasting power demand | |
CN111028100A (en) | Refined short-term load prediction method, device and medium considering meteorological factors | |
CN108345908A (en) | Sorting technique, sorting device and the storage medium of electric network data | |
CN113255900A (en) | Impulse load prediction method considering improved spectral clustering and Bi-LSTM neural network | |
CN111582548A (en) | Power load prediction method based on multivariate user behavior portrait | |
CN113361785A (en) | Power distribution network short-term load prediction method and device, terminal and storage medium | |
CN113393028A (en) | Load prediction method based on data mining technology | |
CN109858667A (en) | It is a kind of based on thunder and lightning weather to the short term clustering method of loading effects | |
CN110852370A (en) | Clustering algorithm-based large-industry user segmentation method | |
Mao et al. | Naive Bayesian algorithm classification model with local attribute weighted based on KNN | |
CN114372835A (en) | Comprehensive energy service potential customer identification method, system and computer equipment | |
CN114004654A (en) | Internet card user portrait construction method and system based on big data analysis | |
CN113988919A (en) | Electricity purchasing structure adjusting method, system and medium | |
Asghari et al. | Selecting and prioritizing the electricity customers for participating in demand response programs | |
Liu et al. | Short-term Load Forecasting Approach with SVM and Similar Days Based on United Data Mining Technology | |
Ramos et al. | Data Mining Techniques Applied to Power Systems | |
CN112365280B (en) | Electric power demand prediction method and device | |
Zhao et al. | Analysis the Consumption Behavior Based on Weekly Load Correlation and K-means Clustering Algorithm |
Legal Events
Date | Code | Title | Description |
---|---|---|---|
PB01 | Publication | ||
PB01 | Publication | ||
SE01 | Entry into force of request for substantive examination | ||
SE01 | Entry into force of request for substantive examination | ||
RJ01 | Rejection of invention patent application after publication | ||
RJ01 | Rejection of invention patent application after publication |
Application publication date: 20200110 |