CN115481788A - Load prediction method and system for phase change energy storage system - Google Patents

Load prediction method and system for phase change energy storage system Download PDF

Info

Publication number
CN115481788A
CN115481788A CN202211062224.6A CN202211062224A CN115481788A CN 115481788 A CN115481788 A CN 115481788A CN 202211062224 A CN202211062224 A CN 202211062224A CN 115481788 A CN115481788 A CN 115481788A
Authority
CN
China
Prior art keywords
neural network
energy storage
load prediction
phase change
data
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.)
Granted
Application number
CN202211062224.6A
Other languages
Chinese (zh)
Other versions
CN115481788B (en
Inventor
李壮举
魏贞祥
李壮辉
史子棋
陈石毓
Current Assignee (The listed assignees may be inaccurate. Google has not performed a legal analysis and makes no representation or warranty as to the accuracy of the list.)
Beijing University of Civil Engineering and Architecture
Original Assignee
Beijing University of Civil Engineering and Architecture
Priority date (The priority date 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 date listed.)
Filing date
Publication date
Application filed by Beijing University of Civil Engineering and Architecture filed Critical Beijing University of Civil Engineering and Architecture
Priority to CN202211062224.6A priority Critical patent/CN115481788B/en
Publication of CN115481788A publication Critical patent/CN115481788A/en
Application granted granted Critical
Publication of CN115481788B publication Critical patent/CN115481788B/en
Active legal-status Critical Current
Anticipated expiration legal-status Critical

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
    • G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
    • G06N3/00—Computing arrangements based on biological models
    • G06N3/02—Neural networks
    • G06N3/08—Learning methods
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
    • G06N3/00—Computing arrangements based on biological models
    • G06N3/12—Computing arrangements based on biological models using genetic models
    • G06N3/126—Evolutionary algorithms, e.g. genetic algorithms or genetic programming
    • 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
    • H—ELECTRICITY
    • H02—GENERATION; CONVERSION OR DISTRIBUTION OF ELECTRIC POWER
    • H02J—CIRCUIT ARRANGEMENTS OR SYSTEMS FOR SUPPLYING OR DISTRIBUTING ELECTRIC POWER; SYSTEMS FOR STORING ELECTRIC ENERGY
    • H02J3/00—Circuit arrangements for ac mains or ac distribution networks
    • H02J3/003—Load forecast, e.g. methods or systems for forecasting future load demand
    • H—ELECTRICITY
    • H02—GENERATION; CONVERSION OR DISTRIBUTION OF ELECTRIC POWER
    • H02J—CIRCUIT ARRANGEMENTS OR SYSTEMS FOR SUPPLYING OR DISTRIBUTING ELECTRIC POWER; SYSTEMS FOR STORING ELECTRIC ENERGY
    • H02J2300/00—Systems for supplying or distributing electric power characterised by decentralized, dispersed, or local generation
    • H02J2300/20—The dispersed energy generation being of renewable origin
    • H02J2300/22—The renewable source being solar energy
    • Y—GENERAL TAGGING OF NEW TECHNOLOGICAL DEVELOPMENTS; GENERAL TAGGING OF CROSS-SECTIONAL TECHNOLOGIES SPANNING OVER SEVERAL SECTIONS OF THE IPC; TECHNICAL SUBJECTS COVERED BY FORMER USPC CROSS-REFERENCE ART COLLECTIONS [XRACs] AND DIGESTS
    • Y04—INFORMATION OR COMMUNICATION TECHNOLOGIES HAVING AN IMPACT ON OTHER TECHNOLOGY AREAS
    • Y04S—SYSTEMS INTEGRATING TECHNOLOGIES RELATED TO POWER NETWORK OPERATION, COMMUNICATION OR INFORMATION TECHNOLOGIES FOR IMPROVING THE ELECTRICAL POWER GENERATION, TRANSMISSION, DISTRIBUTION, MANAGEMENT OR USAGE, i.e. SMART GRIDS
    • Y04S10/00—Systems supporting electrical power generation, transmission or distribution
    • Y04S10/50—Systems or methods supporting the power network operation or management, involving a certain degree of interaction with the load-side end user applications

Landscapes

  • Engineering & Computer Science (AREA)
  • Physics & Mathematics (AREA)
  • Business, Economics & Management (AREA)
  • Health & Medical Sciences (AREA)
  • Theoretical Computer Science (AREA)
  • Life Sciences & Earth Sciences (AREA)
  • General Physics & Mathematics (AREA)
  • Biophysics (AREA)
  • Economics (AREA)
  • General Health & Medical Sciences (AREA)
  • Strategic Management (AREA)
  • Human Resources & Organizations (AREA)
  • Software Systems (AREA)
  • Marketing (AREA)
  • Mathematical Physics (AREA)
  • Tourism & Hospitality (AREA)
  • Computing Systems (AREA)
  • Molecular Biology (AREA)
  • Evolutionary Computation (AREA)
  • General Business, Economics & Management (AREA)
  • Data Mining & Analysis (AREA)
  • Artificial Intelligence (AREA)
  • Bioinformatics & Cheminformatics (AREA)
  • Bioinformatics & Computational Biology (AREA)
  • General Engineering & Computer Science (AREA)
  • Evolutionary Biology (AREA)
  • Computational Linguistics (AREA)
  • Biomedical Technology (AREA)
  • Quality & Reliability (AREA)
  • Genetics & Genomics (AREA)
  • Physiology (AREA)
  • Primary Health Care (AREA)
  • Water Supply & Treatment (AREA)
  • Public Health (AREA)
  • Operations Research (AREA)
  • Entrepreneurship & Innovation (AREA)
  • Game Theory and Decision Science (AREA)
  • Development Economics (AREA)
  • Power Engineering (AREA)
  • Management, Administration, Business Operations System, And Electronic Commerce (AREA)

Abstract

The invention provides a load prediction method and system for a phase change energy storage system, which belong to the technical field of operation control of the phase change energy storage system and are used for acquiring weather environment parameters of a period to be predicted; processing the acquired weather environment parameters of the period to be predicted by using a pre-trained load prediction model to obtain the power consumption of the period to be predicted; the pre-trained load prediction model is characterized in that a basic network model of the pre-trained load prediction model is a combination of a convolutional neural network and a long-term and short-term memory neural network, and a genetic algorithm and a self-attention mechanism are added into the long-term and short-term memory neural network. The invention provides a hybrid neural network prediction model SA-GA-CNN-LSTM, which reduces the error of a prediction result on an extreme point and improves the prediction accuracy rate when the load fluctuates; based on the predicted load result, the daily energy storage demand can be calculated, the planning design of the phase change energy storage system is served, reliable data reference is provided for the deployment of the phase change energy storage system, and the power supply pressure can be relieved.

Description

Load prediction method and system for phase change energy storage system
Technical Field
The invention relates to the technical field of operation control of phase change energy storage systems, in particular to a phase change energy storage system load prediction method and system capable of improving prediction accuracy rate during load fluctuation.
Background
The existing power grid system still has two general problems: firstly, the power generation side is influenced by clean energy such as photovoltaic power generation, wind power generation and the like, the phenomenon of real-time fluctuation of the total power generation amount occurs, and the problem of unstable power supply occurs when the power consumption approaches the upper limit of power generation; and secondly, the power utilization side has the problems of overlarge power supply pressure in the peak period of power utilization and overlarge difference between the power consumption and the upper limit of power supply in the valley period. The energy storage technology is a key means for solving the problems of the power grid system.
For cooling and heating systems, the development of phase change energy storage (PCM) is one of the important means for reducing the power supply pressure, and PCM is a novel environment-friendly energy-saving technology for energy throughput by using the phase change heat of phase change materials. The existing novel low-temperature-difference phase change energy storage material has the phase change temperature of about 25 ℃, small difference with the natural temperature and phase change heat of 243J/g.
As shown in fig. 1, the heating principle of the heating system of the novel phase-change energy storage material is as follows: the phase change energy storage material is stored with heat by solar energy in the daytime, and the heating unit is used for supplementing heat storage in the valley period of electricity price. In addition, a low-temperature phase-change material is arranged to collect low-quality air energy in low-temperature air and convert the low-quality air energy into energy above zero degree to supply the energy to the heating unit for secondary conversion, so that the heating unit can ensure the energy efficiency ratio of more than 3.5 times at extremely low temperature. The refrigeration principle is as follows: the refrigerating unit works at night, the characteristic that the refrigerating unit is high in efficiency due to small temperature difference at night is utilized, the refrigerating unit belongs to valley electricity time at night, energy is released by the phase change energy storage material for refrigeration in other time periods, and the energy efficiency ratio can reach 3-5 times. Because the phase change energy storage system stores energy in the valley electricity period, and the peak flat section releases energy, therefore the cold and hot system of having equipped with the energy storage plate in theory can realize the peak clipping and fill the valley in the aspect of the power consumption, reduces municipal administration power supply pressure, reduces the effect of cost. In order to provide accurate reference in the actual planning design of the phase change energy storage system, the load needs to be accurately predicted, and the error is reduced, so that the condition that the resource waste or the energy storage is insufficient due to excessive energy storage can be effectively avoided.
In the existing data-driven prediction method, in the aspect of inputting a data set, most of the existing researches input experimental data as a whole, actual measurement load data of a phase change energy storage system and load curves have obvious difference on working days, rest days, holidays and other dates, if data with overlarge characteristic difference are trained together, the prediction precision of a model is influenced to a certain extent, in addition, in the prediction method, researchers put forward that the conditions of gradient disappearance or explosion still exist when training is carried out by using a BP (back propagation) neural network, an RBF (radial basis function) neural network, a long-time and short-time memory neural network (LSTM) and a mixed neural network model, the learning capability of the model on a long-distance characteristic relation is insufficient, and input characteristics which have great influence on a prediction result are difficult to learn.
Disclosure of Invention
The invention aims to provide a phase change energy storage system load prediction method and system with improved prediction accuracy, so as to solve at least one technical problem in the background technology.
In order to achieve the purpose, the invention adopts the following technical scheme:
in one aspect, the present invention provides a method for predicting a load of a phase change energy storage system, including:
acquiring weather environment parameters of a period to be predicted;
processing the acquired weather environment parameters of the period to be predicted by using a pre-trained load prediction model to obtain the power consumption of the period to be predicted; the pre-trained load prediction model is characterized in that a basic network model of the pre-trained load prediction model is a combination of a convolutional neural network and a long-term and short-term memory neural network, and a genetic algorithm and a self-attention mechanism are added into the long-term and short-term memory neural network.
Preferably, the pre-trained load prediction model is trained by a training set, the training set includes a plurality of sets of historical data, and each set of historical data includes a weather environment parameter of a certain historical period and a tag marking power consumption data of the historical period.
Preferably, a genetic algorithm added into the long and short term memory neural network is used for optimizing the number of layers of the long and short term memory neural network and the number of neurons of a hidden layer; the self-attention mechanism is added to be used independently in encoding or decoding, and intrinsic correlation of data or characteristics is better concerned.
Preferably, training the load prediction model comprises: transmitting the preprocessed historical data into a convolutional neural network, extracting spatial features of the data, and extracting time features of the data through a long-time short-time memory neural network; and according to the set updated optimization parameters, stopping the training of the model until the iteration times are finished or the loss function tends to be converged, and obtaining the finally trained load prediction model.
Preferably, the preprocessing of the historical data comprises: firstly, normalization processing is carried out, and data of various variables are uniformly scaled to a range; and then carrying out fuzzy C-means clustering on the normalized data, and obtaining the membership degrees of all samples to all class centers by optimizing a target function so as to determine the classes of sample points and achieve the aim of automatically classifying the sample data.
Preferably, the power consumption and the weather environment parameters of the nth-1 date in each class after the fuzzy C-means clustering and the predicted weather environment parameters of the nth date are taken as input variable characteristics, and the power consumption of the data of the nth date is output.
In a second aspect, the present invention provides a load prediction system for a phase change energy storage system, including:
the acquisition module is used for acquiring weather environment parameters of a period to be predicted;
the forecasting module is used for processing the acquired weather environment parameters of the period to be forecasted by utilizing a pre-trained load forecasting model to obtain the power consumption of the period to be forecasted; the pre-trained load prediction model is characterized in that a basic network model of the pre-trained load prediction model is a combination of a convolutional neural network and a long-term and short-term memory neural network, and a genetic algorithm and a self-attention mechanism are added into the long-term and short-term memory neural network.
In a third aspect, the invention provides a non-transitory computer readable storage medium for storing computer instructions which, when executed by a processor, implement a phase change energy storage system load prediction method as described above.
In a fourth aspect, the invention provides a computer program product comprising a computer program for implementing a method for load prediction for a phase change energy storage system as described above when the computer program is run on one or more processors.
In a fifth aspect, the present invention provides an electronic device, comprising: a processor, a memory, and a computer program; wherein a processor is connected with a memory, a computer program is stored in the memory, and when the electronic device runs, the processor executes the computer program stored in the memory to make the electronic device execute the instructions for implementing the phase change energy storage system load prediction method.
The invention has the beneficial effects that: a hybrid neural network prediction model SA-GA-CNN-LSTM is provided, so that the error of a prediction result on an extreme point is reduced, and the prediction accuracy rate in the load fluctuation is improved; based on the predicted load result, the daily energy storage demand can be calculated, the planning design of the phase change energy storage system is served, reliable data reference is provided for the deployment of the phase change energy storage system, and the power supply pressure can be relieved.
Advantages of additional aspects of the invention will be set forth in part in the description which follows, and in part will be obvious from the description, or may be learned by practice of the invention.
Drawings
In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the drawings needed to be used in the description of the embodiments are briefly introduced below, and it is obvious that the drawings in the following description are only some embodiments of the present invention, and it is obvious for those skilled in the art to obtain other drawings based on these drawings without creative efforts.
Fig. 1 is a structural diagram of a heating system of a conventional novel phase change energy storage material.
Fig. 2 is a flowchart of the FCM algorithm according to the embodiment of the present invention.
Fig. 3 is a schematic diagram of FCM cluster index scores according to an embodiment of the present invention.
Fig. 4 is a schematic diagram of a model structure of the CNN-LSTM algorithm according to the embodiment of the present invention.
Fig. 5 is a schematic structural diagram of a one-dimensional convolutional neural network according to an embodiment of the present invention.
Fig. 6 is a schematic structural diagram of an LSTM neural network according to an embodiment of the present invention.
FIG. 7 is a schematic diagram of a GA-LSTM flow according to an embodiment of the present invention.
Fig. 8 is a schematic flow chart illustrating a principle of a self-attention mechanism according to an embodiment of the present invention.
FIG. 9 is a flowchart of load prediction based on the SA-GA-CNN-LSTM model according to an embodiment of the present invention.
Fig. 10 is a diagram of a CNN model prediction result according to an embodiment of the present invention.
Fig. 11 is a diagram of the prediction result of the LSTM model according to the embodiment of the present invention.
FIG. 12 is a graph of the predicted results of the CNN-LSTM model according to the embodiment of the present invention.
FIG. 13 is a graph of the prediction results of the GA-CNN-LSTM model according to the embodiment of the present invention.
FIG. 14 is a diagram of the prediction results of the SA-GA-CNN-LSTM model according to the embodiment of the present invention.
Detailed Description
Reference will now be made in detail to embodiments of the present invention, examples of which are illustrated in the accompanying drawings, wherein like reference numerals refer to the same or similar elements or elements having the same or similar functions throughout. The embodiments described below with reference to the drawings are illustrative only and should not be construed as limiting the invention.
It will be understood by those skilled in the art that, unless otherwise defined, all terms (including technical and scientific terms) used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this invention belongs.
It will be further understood that terms, such as those defined in commonly used dictionaries, should be interpreted as having a meaning that is consistent with their meaning in the context of the prior art and will not be interpreted in an idealized or overly formal sense unless expressly so defined herein.
As used herein, the singular forms "a", "an", "the" and "the" are intended to include the plural forms as well, unless the context clearly indicates otherwise. It will be further understood that the terms "comprises" and/or "comprising," when used in this specification, specify the presence of stated features, integers, steps, operations, elements, and/or components, but do not preclude the presence or addition of one or more other features, integers, steps, operations, elements, components, and/or groups thereof.
In the description herein, references to the description of the term "one embodiment," "some embodiments," "an example," "a specific example," or "some examples," etc., mean that a particular feature, structure, material, or characteristic described in connection with the embodiment or example is included in at least one embodiment or example of the invention. Furthermore, the particular features, structures, materials, or characteristics described may be combined in any suitable manner in any one or more embodiments or examples. Furthermore, various embodiments or examples and features of different embodiments or examples described in this specification can be combined and combined by one skilled in the art without contradiction.
For the purpose of facilitating an understanding of the present invention, the present invention will be further explained by way of specific embodiments with reference to the accompanying drawings, which are not intended to limit the present invention.
It should be understood by those skilled in the art that the drawings are merely schematic representations of embodiments and that the elements shown in the drawings are not necessarily required to practice the invention.
Example 1
The present embodiment 1 provides a load prediction system for a phase change energy storage system, including:
the acquisition module is used for acquiring weather environment parameters of a period to be predicted;
the forecasting module is used for processing the acquired weather environment parameters of the period to be forecasted by utilizing a pre-trained load forecasting model to obtain the power consumption of the period to be forecasted; the pre-trained load prediction model is characterized in that a basic network model of the pre-trained load prediction model is a combination of a convolutional neural network and a long-term and short-term memory neural network, and a genetic algorithm and a self-attention mechanism are added into the long-term and short-term memory neural network.
In this embodiment 1, the method for predicting the load of the phase change energy storage system is implemented by using the system described above, and includes:
acquiring weather environment parameters of a period to be predicted by using an acquisition module;
using a prediction module to process the acquired weather environment parameters of the period to be predicted based on a pre-trained load prediction model to obtain the power consumption of the period to be predicted; the pre-trained load prediction model is characterized in that a basic network model of the pre-trained load prediction model is a combination of a convolutional neural network and a long-term and short-term memory neural network, and a genetic algorithm and a self-attention mechanism are added into the long-term and short-term memory neural network.
The pre-trained load prediction model is obtained by training a training set, the training set comprises a plurality of groups of historical data, and each group of historical data comprises weather environment parameters of a certain historical period and a label marking power consumption data of the historical period.
The number of layers of the long-short time memory neural network and the number of neurons of a hidden layer are optimized by a genetic algorithm added into the long-short time memory neural network; the mechanism of adding self-attention is used in encoding or decoding independently, and the intrinsic correlation of data or characteristics is better concerned.
Training the load prediction model comprises: transmitting the preprocessed historical data into a convolutional neural network, extracting spatial features of the data, and extracting time features of the data through a long-time short-time memory neural network; and according to the set updated optimization parameters, stopping training until the iteration times are finished or the loss function tends to be converged, and obtaining the finally trained load prediction model.
The preprocessing of the historical data comprises the following steps: firstly, normalization processing is carried out, and data of various variables are uniformly scaled to a range; and then carrying out fuzzy C-means clustering on the normalized data, and obtaining the membership of all samples to all class centers by optimizing a target function so as to determine the class of sample points and achieve the aim of automatically classifying the sample data.
And (4) taking the power consumption and weather environment parameters of the (n-1) th date in each class after the fuzzy C mean value clustering and the weather environment parameters with the predicted nth date as input variable characteristics, and outputting the power consumption which is data of the nth date.
Example 2
In this embodiment 2, a load prediction method for a phase change energy storage system is provided, in which input data is subjected to FCM clustering, and then respectively transmitted into two basic models, namely a convolutional neural network and a long-term and short-term memory neural network (CNN-LSTM), and then an SA-GA-CNN-LSTM neural network model optimized by a Genetic Algorithm (GA) and a self-attention mechanism (SA) is added, so that an error of a prediction result on an extreme point is reduced, and prediction accuracy is improved. The specific process is as follows: firstly, clustering data with different characteristics by using an FCM algorithm, extracting the data with different characteristics, constructing an SA-GA-CNN-LSTM model by classification, realizing load prediction of the characteristics with different classes, and finally combining various prediction results to obtain a phase change material energy storage prediction model, thereby enabling the prediction results to be more accurate.
In this embodiment 2, the actually measured historical data of the phase change energy storage cooling and heating system of a certain substation in a certain urban area is used as a data set. Firstly, preprocessing data, and modifying and filling abnormal values and missing values of the sensor, wherein the modified and filled values are average values of nearest adjacent acquisition moments. All input data features are subjected to data normalization processing to facilitate consistency analysis, then a fuzzy C-means clustering method is adopted to cluster data samples, and data with overlarge differences are respectively predicted to avoid mutual influence.
The purpose of the data normalization processing is to uniformly scale the data of various variables to a range, and because the influence degree of the data with a larger range is larger than that of the data with a smaller range, the method accelerates the training speed of the model and ensures the accuracy and stability of the model. Assuming that the input feature data set is an m × n matrix X, X is expressed as:
Figure BDA0003826726430000081
the raw data is normalized according to the following formula:
Figure BDA0003826726430000082
wherein x is ij As the original data, it is the original data,
Figure BDA0003826726430000083
respectively representing the maximum and minimum values of each column of the matrix X, X i ′ j The normalized data has a value range of [0,1]。
As shown in fig. 2, a fuzzy C-means algorithm (FCM) obtains membership of all samples to all class centers by optimizing an objective function, so as to determine classes of sample points, thereby achieving the purpose of automatically classifying sample data. And assigning a membership function belonging to each class to each sample, and clustering the samples according to the membership value. FCM is a soft clustering algorithm that is less prone to errors than hard clustering algorithms. Because some data are not easily distinguished into obvious classes, and the clustering result calculated according to the algorithm is not accurate, in this embodiment 2, the FCM is adopted to assign a weight to each class for each object, and mark the degree of the class to which the object belongs for fuzzy clustering, which is a data processing method with good practicability.
Parameters in the target function and the formula in the FCM are set as:
Figure BDA0003826726430000091
Figure BDA0003826726430000092
wherein x is i The ith column of X, n the number of samples, c the number of clusters, and U by U ij A constituent membership matrix (i =1,2.. N, j =1,2.. C), V = (p =) 1 ,p 2 ,...,p c )，u ij Indicating degree of membership, p, of the jth sample to the ith class j Represents the jth cluster center, m represents a weighting index, and generally m ≧ 1.
U in membership matrix derived by Lagrange multiplier method ij The expression of (a) is:
Figure BDA0003826726430000093
cluster center p j The expression of (c) is:
Figure BDA0003826726430000094
the evaluation index is intra-class distance (intra-class distance), and the index represents the mean square distance from each sample point in the same class to the cluster center. Assuming that clustering results in K classes of different load characteristics,
Figure BDA0003826726430000095
Figure BDA0003826726430000096
is C i Sample in class N i Is C i The number of samples in the category of the sample,
Figure BDA0003826726430000097
indicating the cluster centers in the respective categories,
Figure BDA0003826726430000098
representing all elements of different classes in the sample. The expression for the distance within the class is:
Figure BDA0003826726430000099
in this embodiment 2, the fuzzy C-means clustering algorithm is used as a clustering method to cluster the normalized historical data, and a FCM clustering index score histogram is drawn, where the evaluation index is the intra-class distance, d 2 (C i ) The smaller the value is, the better the clustering effect is, and the clustering result is as shown in fig. 3, when the clustering center C =8, the intra-class distance value is the lowest, and the clustering effect is the best. These 8 results are Monday, respectively; tuesday, wednesday, thursday, friday; saturday; the weekday; legal festivals and holidays; the day before the statutory holiday; the first day after the statutory holiday; the day after the statutory holiday.
As shown in fig. 4, in the embodiment 2, the CNN-LSTM neural network model firstly transmits the preprocessed data into the convolutional neural network CNN to extract the spatial features of the data, and then extracts the temporal features of the data through the long-short term memory neural network LSTM. And outputting the electricity power on the nth date and the temperature, the humidity, the wind direction and the wind speed of the nth date in each clustered class as input variable characteristics. The whole model combines the advantages of a Convolutional Neural Network (CNN) and a long-term memory network (LSTM) respectively.
In addition, in this example 2, a Genetic Algorithm (GA) and a self-attention mechanism (SA) were added to the training of LSTM.
The CNN is a special linear operation and mainly used for extracting data features, and the CNN has the basic structure of an input layer, a convolution layer, a pooling layer, a full-connection layer and an output layer. Because the time sequence data prediction problem is one-dimensional, a one-dimensional convolution layer is adopted in the text, and the convolution kernel moves in a single direction by taking a time step as a characteristic, so that the time sequence data prediction problem can be the same as the dimension transmitted to the LSTM in the next stage, and dimension conversion is avoided. The structure of the one-dimensional convolutional neural network is shown in fig. 5.
The LSTM is a special RNNs, which can effectively solve the long-term dependence problem, and the network structure is shown in fig. 6. The LSTM mainly contains the cell state and three gates, respectively a forgetting gate, an input gate, and an output gate. LSTM first determines what information can pass through the cell state by a forgetting gate, and this determination is controlled by a sigmoid function, which will depend on the output h from the previous moment t-1 And the current input x t To produce a f of 0 to 1 t Thereby determining the information C learned at the previous moment t-1 Whether passing or partially passing. Followed by generating information to be updated, and a second step of generating new candidate values from the tanh function
Figure BDA0003826726430000101
Added to the cell state as a candidate generated for the current layer. The values generated by the first two parts are combined to update, and the process is a process of losing unnecessary information and adding new information. In the last step, an initial output is obtained through a sigmoid function, and then C is processed through a tanh function t Scaling to-1 to 1, and multiplying the output by sigmoid function pair by pair to obtain the output h of the model t . Where w and b represent the weight matrix and offset vector of the gate. The specific calculation formula is as follows:
f t ＝σ(ω f [h t-1 ，x t ]+b f )
i t ＝σ(ω i [h t-1 ，x t ]+b i )
Figure BDA0003826726430000111
Figure BDA0003826726430000112
o t ＝σ(w o [h t-1 ，x t ]+b o )
h t ＝o t ×tanh(C t )
in this embodiment 2, a Genetic Algorithm (GA) is used to optimize the number of layers of the long-term and short-term memory neural network and the number of neurons in the hidden layer, and a flow chart of GA-optimized LSTM is shown in fig. 7. Genetic algorithm is originated from computer mode research on biological system, and is one algorithm simulating selection, crossing and variation in biological evolution process. Wherein the selection has the functions of overcoming the preference and eliminating the preference, survival of the fittest and searching the optimal individual; the cross function is to ensure the stability of the population and evolve towards the optimal solution; the function of the variation is to ensure the diversity of the population and avoid the local convergence possibly generated by the cross. The selection of the fitness function directly influences the convergence speed of the genetic algorithm and whether the optimal solution can be found. The selection function uses a roulette function, i.e., the higher the fitness function score is, the greater the probability of being selected.
As shown in fig. 8, in this embodiment 2, the attention mechanism is embedded into the LSTM neural network model to solve the problem that the memory capacity is insufficient and the extraction of important features is weak in the face of a long input sequence, so as to improve the accuracy of the model in predicting the power consumption load. The self-Attention Mechanism (self-Attention Mechanism) is a variant of the Attention Mechanism (Attention Mechanism), and compared with the Attention Mechanism, the self-Attention Mechanism can be independently used in encoding or decoding, and the intrinsic relation of input data is concerned more, so that the stability and the prediction accuracy of a model are improved.
As in FIG. 8, wherein D x For the input sequence, Q, K, V is the input sequence D, respectively x The Value (Value), key (Key) and Query (Query) matrix of (A) are input from the same input D x Subject to different weights (w) q 、w k 、w v ) Respectively combined into a matrix, q i For features D in the input sequence i Query vector of, K T Being a transposed matrix of K, softmax () is column-wiseA normalization activation function for normalizing the attention score,
Figure BDA0003826726430000113
to scale the factor, let softmax () compute easily, ai be feature D i Attention weight of, V i Is characterized by i The corresponding value vector, attention () represents the attention score function.
The energy consumption of the cold and hot system is a group of time sequence data which changes along with the energy demand and is influenced by environmental factors such as temperature and humidity and the like and shows nonlinear change. Therefore, in the load prediction task, not only the connection between the input parameters but also the change of the power consumption in the time dimension need to be considered. Since the long-term memory neural network is well represented in the prediction of time-series data, it is used as a reference model. However, when the input parameters are more, the feature extraction capability and stability of the single long-term and short-term memory neural network cannot meet the requirements, so that the convolutional neural network is introduced in the embodiment to solve the problem, and the convolutional neural network can be used for processing time series data, so that a result obtained by extracting the features of the convolutional neural network is mapped into a sequence vector to be used as the input of the long-term and short-term memory neural network. And combining the convolutional neural network with the long-term and short-term memory neural network to construct a CNN-LSTM prediction model.
Although the CNN-LSTM model is significantly more predictive than CNN and LSTM alone, two problems remain. Firstly: although LSTM provides some relief from gradient disappearance or explosions, there are still cases where gradients disappear or explode on the cell units. Secondly, the method comprises the following steps: when the input sequence is too long, the transmitted information passes through the forgetting gate for many times, so that the learning capacity of the model for the long-distance feature relationship is reduced, the LSTM cannot effectively find the internal relation of the input sequence, namely the influence of different input features on the prediction result is inconsistent, the LSTM cannot treat the input sequence differently, and the LSTM focuses on learning data with large influence on the prediction result. For the above problems, in embodiment 2, the hyper-parameter weight and the bias value of the Genetic Algorithm (GA) training network are added, so that overfitting of model training can be effectively alleviated, the situation of gradient disappearance or explosion is avoided, the optimal window size and the number of neurons are obtained, and the step of manually adjusting parameters is omitted; and adding a self-attention mechanism (SA), acquiring input data information at each moment during model training, obtaining weights of different data, weighting a hidden layer in the LSTM, capturing long-distance dependency relationship better, effectively distinguishing the influence degree of different data characteristics on a prediction result, and learning data characteristics with high influence degree on the prediction result in a key way. In summary, an SA-GA-CNN-LSTM neural network model is constructed, and an electrical load prediction flow chart of the model is shown in fig. 9, and the specific steps are as follows:
and (4) sorting the data obtained by the sensor, modifying the abnormal value and the missing value, and carrying out normalization processing. And then, carrying out fuzzy C-means clustering operation on the power consumption data with different characteristics, and removing the influence of overlarge difference on model prediction. Finally, dividing all classes of data according to the proportion of the training set to the test set of 8:2;
initializing model hyper-parameters, transmitting a training set into the model for training, extracting features by a network, and updating optimized parameters according to settings. Stopping training the model until the iteration times are completed or the loss function tends to converge;
transmitting the test set data into a trained network model for prediction, obtaining different evaluation scores through four different evaluation indexes, and comparing and analyzing the scores with other network models to obtain an experimental result;
the method predicts the future load by establishing a neural network model, and belongs to the scope of regression problems. A single evaluation index has a certain limitation, and therefore four evaluation indexes, namely Mean Square Error (MSE), root Mean Square Error (RMSE), mean Absolute Error (MAE) and Mean Absolute Percentage Error (MAPE), are selected to compare the prediction accuracy and stability of different models. The calculation formula is as follows:
Figure BDA0003826726430000131
Figure BDA0003826726430000132
Figure BDA0003826726430000133
Figure BDA0003826726430000134
wherein, the value ranges of the four evaluation indexes are all [0, + ∞ ], n is the total number of the samples of the test set,
Figure BDA0003826726430000141
and y i And respectively representing the predicted value and the real value of the test set, wherein the smaller the MSE and the RMSE are, the closer the predicted value and the real value are, and the higher the model precision is. The smaller the MAE and MAPE, the better the stability of the model.
In this embodiment, the data sampling interval of the actually measured data of the phase change energy storage cooling and heating system of a certain substation in a certain city is 10min, and the actually measured data is preprocessed to be used as a data set, wherein 80% of the actually measured data is used as a training set, and 20% of the actually measured data is used as a test set. All experiments in this embodiment are performed on polar chain AI clouds, an experiment machine adopts an Ampere a100 graphics card, a CPU model is AMD EPYC 730216-Core Processor,251G memory, 64 cores, 39G memory, an example frame is tensorflow2.4.1, a program code is python3.8, a CUDA version is 11.1.1, and the number of iterations in the experiment training phase in this embodiment is 10000.
In order to verify the prediction capability of the SA-GA-CNN-LSTM neural network model, the prediction results of the SA-GA-CNN-LSTM neural network model, the non-optimized CNN-LSTM hybrid neural network model and the GA-CNN-LSTM hybrid neural network model without adding an attention mechanism are compared, and the evaluation indexes are Mean Square Error (MSE), root Mean Square Error (RMSE), mean Absolute Error (MAE), mean Absolute Percentage Error (MAPE) and model running time.
In this embodiment, the single neural network models CNN and LSTM, and the mixed neural network models CNN-LSTM, GA-CNN-LSTM, and SA-GA-CNN-LSTM are tested, respectively. And after the model prediction results are subjected to inverse normalization, intercepting comparison graphs of the power consumption prediction results of the corresponding time points of one week, as shown in fig. 10 to 14. It can be obviously seen from the image that the single CNN and LSTM neural network models are poor in prediction effect, the superposition degree of the predicted value and the true value of the SA-GA-CNN-LSTM neural network model is maximum, and compared with the GA-CNN-LSTM neural network model without the self-attention mechanism, the prediction model is better in performance in the time period when the electric power is changed, the error of the prediction result on the extreme point is reduced, and the prediction model is obviously superior to other three neural network models.
Further comparing and analyzing the experimental results, as listed in table 1, in the MSE and RMSE evaluation indexes, the error of the single neural network model CNN is the largest, and the error is LSTM, and the error of the SA-GA-CNN-LSTM in the mixed neural network model is the smallest. Wherein the SA-GA-CNN-LSTM is reduced by 0.477kW and 1.185kW compared with the errors of GA-CNN-LSTM and CNN-LSTM respectively under the RMSE index. Under the evaluation indexes of MAE and MAPE, the error of the CNN of the single neural network model is the largest, and the error of the LSTM is the second, and the error of the SA-GA-CNN-LSTM in the mixed neural network model is still the smallest. Wherein, under the MAPE evaluation index, the error of SA-GA-CNN-LSTM is reduced by 0.39 percent and 1.49 percent respectively compared with the error of GA-CNN-LSTM and CNN-LSTM. The result fully embodies the advantages of high key learning influence degree data after the self-attention mechanism acquires comprehensive data and combination with genetic algorithm optimization hyper-parameters, and by combining the table 2 and the graphs 10 to 14, compared with the SA-GA-CNN-LSTM neural network model provided by the embodiment, four error indexes of the SA-GA-CNN-LSTM neural network model are minimum, an optimal prediction effect is presented, and the prediction model provided by the embodiment is proved to be higher in accuracy and stability and better in prediction performance.
TABLE 1
Figure BDA0003826726430000151
In this embodiment 2, the running time is divided into two parts, namely training time and testing time, and the running times of different neural network models are listed in table 2. In the training phase: the time spent by the single CNN and LSTM neural networks is 208.23s and 68.32s, while the time spent by the hybrid neural networks CNN-LSTM, GA-CNN-LSTM and SA-GA-CNN-LSTM is 456.76s, 327.52s and 341.67s, respectively. In the testing stage: the time consumed by the CNN, LSTM, CNN-LSTM, GA-CNN-LSTM and SA-GA-CNN-LSTM neural network models is 0.29s, 0.20s, 0.74s, 0.55s and 0.61s respectively. It can be seen that the single neural network model has better operation speed performance in both training and testing stages, but needs comprehensive error performance indexes in the actual prediction task, so that the hybrid neural network model has more advantages after the comprehensive error performance.
TABLE 2
Figure BDA0003826726430000161
Example 3
Embodiment 3 of the present invention provides a non-transitory computer-readable storage medium, where the non-transitory computer-readable storage medium is used to store computer instructions, and when the computer instructions are executed by a processor, the non-transitory computer-readable storage medium implements a method for load prediction of a phase change energy storage system, where the method includes:
acquiring weather environment parameters of a period to be predicted;
processing the acquired weather environment parameters of the period to be predicted by using a pre-trained load prediction model to obtain the power consumption of the period to be predicted; the pre-trained load prediction model is characterized in that a basic network model of the pre-trained load prediction model is a combination of a convolutional neural network and a long-term and short-term memory neural network, and a genetic algorithm and a self-attention mechanism are added into the long-term and short-term memory neural network.
Example 4
An embodiment 4 of the present invention provides a computer program (product) comprising a computer program, which when run on one or more processors, is configured to implement a method for load prediction for a phase change energy storage system, the method comprising:
acquiring weather environment parameters of a period to be predicted;
processing the acquired weather environment parameters of the period to be predicted by using a pre-trained load prediction model to obtain the power consumption of the period to be predicted; the basic network model of the pre-trained load prediction model is a combination of a convolutional neural network and a long-term and short-term memory neural network, and a genetic algorithm and a self-attention mechanism are added into the long-term and short-term memory neural network.
Example 5
An embodiment 5 of the present invention provides an electronic device, including: a processor, a memory, and a computer program; wherein a processor is connected to a memory, a computer program is stored in the memory, and when the electronic device is running, the processor executes the computer program stored in the memory to make the electronic device execute instructions for implementing a phase change energy storage system load prediction method, the method comprising:
acquiring weather environment parameters of a period to be predicted;
processing the acquired weather environment parameters of the period to be predicted by using a pre-trained load prediction model to obtain the power consumption of the period to be predicted; the pre-trained load prediction model is characterized in that a basic network model of the pre-trained load prediction model is a combination of a convolutional neural network and a long-term and short-term memory neural network, and a genetic algorithm and a self-attention mechanism are added into the long-term and short-term memory neural network.
In conclusion, the phase change energy storage system load prediction method and system provided by the embodiment of the invention provide the hybrid neural network prediction model SA-GA-CNN-LSTM, and the prediction result shows better performance after integrating four error indexes of MSE, RMSE, MAE and MAPE and the operation speed, so that the error of the prediction result on an extreme point is reduced, and the prediction accuracy rate in the load fluctuation is improved. Based on the load result predicted by the model, the daily energy storage demand can be calculated, the method is used for planning and designing the phase change energy storage system, a valuable reference is provided for the deployment of the phase change energy storage system, the energy storage development is assisted, the commercial power supply pressure is relieved, and the method has certain practical value.
As will be appreciated by one skilled in the art, embodiments of the present invention may be provided as a method, system, or computer program product. Accordingly, the present invention may take the form of an entirely hardware embodiment, an entirely software embodiment or an embodiment combining software and hardware aspects. Furthermore, the present invention may take the form of a computer program product embodied on one or more computer-usable storage media (including, but not limited to, disk storage, CD-ROM, optical storage, and the like) having computer-usable program code embodied therein.
The present invention has been described with reference to flowchart illustrations and/or block diagrams of methods, apparatus (systems), and computer program products according to embodiments of the invention. It will be understood that each flow and/or block of the flow diagrams and/or block diagrams, and combinations of flows and/or blocks in the flow diagrams and/or block diagrams, can be implemented by computer program instructions. These computer program instructions may be provided to a processor of a general purpose computer, special purpose computer, embedded processor, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, create means for implementing the functions specified in the flowchart flow or flows and/or block diagram block or blocks.
These computer program instructions may also be stored in a computer-readable memory that can direct a computer or other programmable data processing apparatus to function in a particular manner, such that the instructions stored in the computer-readable memory produce an article of manufacture including instruction means which implement the function specified in the flowchart flow or flows and/or block diagram block or blocks.
These computer program instructions may also be loaded onto a computer or other programmable data processing apparatus to cause a series of operational steps to be performed on the computer or other programmable apparatus to produce a computer implemented process such that the instructions which execute on the computer or other programmable apparatus provide steps for implementing the functions specified in the flowchart flow or flows and/or block diagram block or blocks.
Although the embodiments of the present invention have been described with reference to the accompanying drawings, it is not intended to limit the scope of the present invention, and it should be understood by those skilled in the art that various modifications and variations can be made without inventive efforts based on the technical solutions disclosed in the present invention.

Claims (10)

1. A method for predicting load of a phase change energy storage system is characterized by comprising the following steps:
acquiring weather environment parameters of a period to be predicted;
processing the acquired weather environment parameters of the period to be predicted by using a pre-trained load prediction model to obtain the power consumption of the period to be predicted; the pre-trained load prediction model is characterized in that a basic network model of the pre-trained load prediction model is a combination of a convolutional neural network and a long-term and short-term memory neural network, and a genetic algorithm and a self-attention mechanism are added into the long-term and short-term memory neural network.
2. The method for predicting the load of the phase change energy storage system according to claim 1, wherein the pre-trained load prediction model is trained by a training set, the training set comprises a plurality of sets of historical data, and each set of historical data comprises a weather environment parameter of a certain historical period and a label marking the power consumption data of the historical period.
3. The phase change energy storage system load prediction method according to claim 1, characterized in that a genetic algorithm added to the long and short term memory neural network optimizes the number of layers of the long and short term memory neural network and the number of neurons in a hidden layer; the self-attention mechanism is added to be used independently in encoding or decoding, and intrinsic correlation of data or characteristics is better concerned.
4. The phase change energy storage system load prediction method of claim 2, wherein training the load prediction model comprises: transmitting the preprocessed historical data into a convolutional neural network, extracting spatial features of the data, and extracting time features of the data through a long-time short-time memory neural network; and according to the set updated optimization parameters, stopping training until the iteration times are finished or the loss function tends to be converged, and obtaining the finally trained load prediction model.
5. The phase change energy storage system load prediction method according to claim 4, wherein the preprocessing of the historical data comprises: firstly, normalization processing is carried out, and data of various variables are uniformly scaled to a range; and then carrying out fuzzy C-means clustering on the normalized data, and obtaining the membership of all samples to all class centers by optimizing a target function so as to determine the class of sample points and achieve the aim of automatically classifying the sample data.
6. The phase change energy storage system load prediction method according to claim 5, wherein the power consumption and the weather environment parameters at the n-1 th date in each class after fuzzy C-means clustering and the predicted weather environment parameters at the n-th date are used as input variable characteristics, and the power consumption of the data at the n-th date is output.
7. A phase change energy storage system load prediction system, comprising:
the acquisition module is used for acquiring weather environment parameters of a period to be predicted;
the prediction module is used for processing the acquired weather environment parameters of the period to be predicted by utilizing a pre-trained load prediction model to obtain the power consumption of the period to be predicted; the basic network model of the pre-trained load prediction model is a combination of a convolutional neural network and a long-term and short-term memory neural network, and a genetic algorithm and a self-attention mechanism are added into the long-term and short-term memory neural network.
8. A non-transitory computer readable storage medium storing computer instructions which, when executed by a processor, implement the phase change energy storage system load prediction method according to any one of claims 1 to 6.
9. A computer program product, comprising a computer program for implementing a phase change energy storage system load prediction method according to any one of claims 1 to 6 when run on one or more processors.
10. An electronic device, comprising: a processor, a memory, and a computer program; wherein a processor is connected to the memory, the computer program being stored in the memory, the processor executing the computer program stored in the memory when the electronic device is running, to cause the electronic device to execute instructions implementing the method of phase change energy storage system load prediction according to any of claims 1-6.
CN202211062224.6A 2022-08-31 2022-08-31 Phase change energy storage system load prediction method and system Active CN115481788B (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN202211062224.6A CN115481788B (en) 2022-08-31 2022-08-31 Phase change energy storage system load prediction method and system

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN202211062224.6A CN115481788B (en) 2022-08-31 2022-08-31 Phase change energy storage system load prediction method and system

Publications (2)

Publication Number Publication Date
CN115481788A true CN115481788A (en) 2022-12-16
CN115481788B CN115481788B (en) 2023-08-25

Family

ID=84423074

Family Applications (1)

Application Number Title Priority Date Filing Date
CN202211062224.6A Active CN115481788B (en) 2022-08-31 2022-08-31 Phase change energy storage system load prediction method and system

Country Status (1)

Country Link
CN (1) CN115481788B (en)

Cited By (2)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN116028838A (en) * 2023-01-09 2023-04-28 广东电网有限责任公司 Clustering algorithm-based energy data processing method and device and terminal equipment
CN118365183A (en) * 2024-06-20 2024-07-19 西南石油大学 Oil well yield prediction method and system based on hybrid neural network

Citations (18)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
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
CN110619389A (en) * 2019-09-23 2019-12-27 山东大学 Load prediction method and system of combined cooling heating and power system based on LSTM-RNN
CN111105104A (en) * 2020-02-20 2020-05-05 河南理工大学 Short-term power load prediction method based on similar day and RBF neural network
JP2020088944A (en) * 2018-11-19 2020-06-04 富士電機株式会社 Power demand prediction device, power demand prediction method, and program therefor
CN111563610A (en) * 2020-03-30 2020-08-21 山东电力工程咨询院有限公司 LSTM neural network-based building electrical load comprehensive prediction method and system
US20200333767A1 (en) * 2018-02-17 2020-10-22 Electro Industries/Gauge Tech Devices, systems and methods for predicting future consumption values of load(s) in power distribution systems
AU2020104000A4 (en) * 2020-12-10 2021-02-18 Guangxi University Short-term Load Forecasting Method Based on TCN and IPSO-LSSVM Combined Model
CN112507793A (en) * 2020-11-05 2021-03-16 上海电力大学 Ultra-short-term photovoltaic power prediction method
CN112529283A (en) * 2020-12-04 2021-03-19 天津天大求实电力新技术股份有限公司 Comprehensive energy system short-term load prediction method based on attention mechanism
US20210133536A1 (en) * 2018-10-22 2021-05-06 Ennew Digital Technology Co., Ltd. Load prediction method and apparatus based on neural network
CN113094822A (en) * 2021-03-12 2021-07-09 华中科技大学 Method and system for predicting residual life of mechanical equipment
WO2021138925A1 (en) * 2020-01-08 2021-07-15 重庆邮电大学 Lithium battery capacity estimation method based on improved convolution-long short term memory neural network
CN113191069A (en) * 2021-03-19 2021-07-30 国网江苏省电力有限公司 Air conditioner load estimation method and system based on double-branch deep learning model
CN113743650A (en) * 2021-08-04 2021-12-03 南方电网科学研究院有限责任公司 Power load prediction method, device, equipment and storage medium
CN113962364A (en) * 2021-10-22 2022-01-21 四川大学 Multi-factor power load prediction method based on deep learning
CN113988426A (en) * 2021-10-29 2022-01-28 国网江苏省电力有限公司苏州供电分公司 Electric vehicle charging load prediction method and system based on FCM clustering and LSTM
CN114510870A (en) * 2022-01-07 2022-05-17 华东交通大学 Method and device for predicting residual life of underground structure of urban rail transit
CN114580707A (en) * 2022-01-26 2022-06-03 安徽农业大学 Emotional tendency prediction model, building method and prediction method of multi-feature fusion product

Patent Citations (18)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US20200333767A1 (en) * 2018-02-17 2020-10-22 Electro Industries/Gauge Tech Devices, systems and methods for predicting future consumption values of load(s) in power distribution systems
US20210133536A1 (en) * 2018-10-22 2021-05-06 Ennew Digital Technology Co., Ltd. Load prediction method and apparatus based on neural network
JP2020088944A (en) * 2018-11-19 2020-06-04 富士電機株式会社 Power demand prediction device, power demand prediction method, and program therefor
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
CN110619389A (en) * 2019-09-23 2019-12-27 山东大学 Load prediction method and system of combined cooling heating and power system based on LSTM-RNN
WO2021138925A1 (en) * 2020-01-08 2021-07-15 重庆邮电大学 Lithium battery capacity estimation method based on improved convolution-long short term memory neural network
CN111105104A (en) * 2020-02-20 2020-05-05 河南理工大学 Short-term power load prediction method based on similar day and RBF neural network
CN111563610A (en) * 2020-03-30 2020-08-21 山东电力工程咨询院有限公司 LSTM neural network-based building electrical load comprehensive prediction method and system
CN112507793A (en) * 2020-11-05 2021-03-16 上海电力大学 Ultra-short-term photovoltaic power prediction method
CN112529283A (en) * 2020-12-04 2021-03-19 天津天大求实电力新技术股份有限公司 Comprehensive energy system short-term load prediction method based on attention mechanism
AU2020104000A4 (en) * 2020-12-10 2021-02-18 Guangxi University Short-term Load Forecasting Method Based on TCN and IPSO-LSSVM Combined Model
CN113094822A (en) * 2021-03-12 2021-07-09 华中科技大学 Method and system for predicting residual life of mechanical equipment
CN113191069A (en) * 2021-03-19 2021-07-30 国网江苏省电力有限公司 Air conditioner load estimation method and system based on double-branch deep learning model
CN113743650A (en) * 2021-08-04 2021-12-03 南方电网科学研究院有限责任公司 Power load prediction method, device, equipment and storage medium
CN113962364A (en) * 2021-10-22 2022-01-21 四川大学 Multi-factor power load prediction method based on deep learning
CN113988426A (en) * 2021-10-29 2022-01-28 国网江苏省电力有限公司苏州供电分公司 Electric vehicle charging load prediction method and system based on FCM clustering and LSTM
CN114510870A (en) * 2022-01-07 2022-05-17 华东交通大学 Method and device for predicting residual life of underground structure of urban rail transit
CN114580707A (en) * 2022-01-26 2022-06-03 安徽农业大学 Emotional tendency prediction model, building method and prediction method of multi-feature fusion product

Non-Patent Citations (5)

* Cited by examiner, † Cited by third party
Title
KUIHUA WU ET AL.: "《An attention-based CNN-LSTM-BiLSTM model for short-term electric load forecasting in integrated energy system》", <ELECTIRCAL ENERGY SYSTEM>, pages 1 - 15 *
刘新志: "《环境因素影响下的形似日分析与短期负荷预测》", 《中国优秀硕士学位论文全文数据库》, pages 1 - 59 *
袁洪跃 等: "建筑用分布式相变储能系统研究", 智能建筑电气技术, pages 46 - 48 *
郭虹，李壮举: "基于中央空调的几种负荷预测方法对比分析", 建筑热能通风空调, pages 1 - 5 *
陈国涛;滕欢;: "基于混合神经网络深度学习的短期负荷预测", 水电能源科学, no. 04, pages 199 - 202 *

Cited By (3)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN116028838A (en) * 2023-01-09 2023-04-28 广东电网有限责任公司 Clustering algorithm-based energy data processing method and device and terminal equipment
CN116028838B (en) * 2023-01-09 2023-09-19 广东电网有限责任公司 Clustering algorithm-based energy data processing method and device and terminal equipment
CN118365183A (en) * 2024-06-20 2024-07-19 西南石油大学 Oil well yield prediction method and system based on hybrid neural network

Also Published As

Publication number Publication date
CN115481788B (en) 2023-08-25

Similar Documents

Publication Publication Date Title
CN109754113B (en) Load prediction method based on dynamic time warping and long-and-short time memory
CN113962364B (en) Multi-factor power load prediction method based on deep learning
Ke et al. Short-term electrical load forecasting method based on stacked auto-encoding and GRU neural network
CN108280551B (en) Photovoltaic power generation power prediction method utilizing long-term and short-term memory network
CN110619360A (en) Ultra-short-term wind power prediction method considering historical sample similarity
CN115481788B (en) Phase change energy storage system load prediction method and system
CN106251001A (en) A kind of based on the photovoltaic power Forecasting Methodology improving fuzzy clustering algorithm
Liu et al. Heating load forecasting for combined heat and power plants via strand-based LSTM
CN112116144A (en) Regional power distribution network short-term load prediction method
CN113554466A (en) Short-term power consumption prediction model construction method, prediction method and device
CN111861013A (en) Power load prediction method and device
CN112149890A (en) Comprehensive energy load prediction method and system based on user energy label
CN113762387B (en) Multi-element load prediction method for data center station based on hybrid model prediction
CN114462718A (en) CNN-GRU wind power prediction method based on time sliding window
He et al. A power forecasting approach for PV plant based on irradiance index and LSTM
CN115907122A (en) Regional electric vehicle charging load prediction method
CN113762591B (en) Short-term electric quantity prediction method and system based on GRU and multi-core SVM countermeasure learning
CN116826710A (en) Peak clipping strategy recommendation method and device based on load prediction and storage medium
CN117132132A (en) Photovoltaic power generation power prediction method based on meteorological data
CN114611757A (en) Electric power system short-term load prediction method based on genetic algorithm and improved depth residual error network
CN117390550A (en) Low-carbon park carbon emission dynamic prediction method and system considering emission training set
Yong et al. Short-term building load forecasting based on similar day selection and LSTM network
Lv et al. Research on short-term load forecasting approach for smart grid
CN115759343A (en) E-LSTM-based user electric quantity prediction method and device
Jiang et al. Short-term power load forecasting based on PSO-GRU

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
CB03 Change of inventor or designer information

Inventor after: Li Zhuangju

Inventor after: Shi Ziqi

Inventor after: Wei Zhenxiang

Inventor after: Li Zhuanghui

Inventor after: Chen Shiyu

Inventor before: Li Zhuangju

Inventor before: Wei Zhenxiang

Inventor before: Li Zhuanghui

Inventor before: Shi Ziqi

Inventor before: Chen Shiyu

CB03 Change of inventor or designer information