CN111461377A - Prediction method/system, storage medium and equipment for total energy demand - Google Patents
Prediction method/system, storage medium and equipment for total energy demand Download PDFInfo
- Publication number
- CN111461377A CN111461377A CN201910046186.7A CN201910046186A CN111461377A CN 111461377 A CN111461377 A CN 111461377A CN 201910046186 A CN201910046186 A CN 201910046186A CN 111461377 A CN111461377 A CN 111461377A
- Authority
- CN
- China
- Prior art keywords
- sequence
- matrix
- neural network
- user
- historical
- 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
Links
- 238000000034 method Methods 0.000 title claims abstract description 82
- 238000003860 storage Methods 0.000 title claims abstract description 12
- 239000011159 matrix material Substances 0.000 claims abstract description 120
- 238000013528 artificial neural network Methods 0.000 claims abstract description 68
- 238000005265 energy consumption Methods 0.000 claims abstract description 11
- 238000012549 training Methods 0.000 claims description 30
- 230000015654 memory Effects 0.000 claims description 18
- 238000013527 convolutional neural network Methods 0.000 claims description 17
- 230000000306 recurrent effect Effects 0.000 claims description 16
- 238000012216 screening Methods 0.000 claims description 15
- 238000004364 calculation method Methods 0.000 claims description 14
- 238000004590 computer program Methods 0.000 claims description 13
- 238000004422 calculation algorithm Methods 0.000 claims description 12
- 238000009826 distribution Methods 0.000 claims description 10
- 125000004122 cyclic group Chemical group 0.000 claims description 7
- 230000009466 transformation Effects 0.000 claims description 6
- 230000006870 function Effects 0.000 claims description 4
- 238000005457 optimization Methods 0.000 claims description 3
- 230000006403 short-term memory Effects 0.000 claims description 3
- 230000001537 neural effect Effects 0.000 claims description 2
- 210000002569 neuron Anatomy 0.000 claims description 2
- 230000007787 long-term memory Effects 0.000 claims 1
- 230000008569 process Effects 0.000 description 8
- 238000004891 communication Methods 0.000 description 5
- 238000012545 processing Methods 0.000 description 4
- 238000004519 manufacturing process Methods 0.000 description 3
- 238000010586 diagram Methods 0.000 description 2
- 230000000717 retained effect Effects 0.000 description 2
- 239000002699 waste material Substances 0.000 description 2
- 230000009286 beneficial effect Effects 0.000 description 1
- 230000008859 change Effects 0.000 description 1
- 230000002860 competitive effect Effects 0.000 description 1
- 238000012423 maintenance Methods 0.000 description 1
- 238000012986 modification Methods 0.000 description 1
- 230000004048 modification Effects 0.000 description 1
- 230000003287 optical effect Effects 0.000 description 1
- 238000011144 upstream manufacturing Methods 0.000 description 1
Images
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
- G06Q10/00—Administration; Management
- G06Q10/04—Forecasting or optimisation specially adapted for administrative or management purposes, e.g. linear programming or "cutting stock problem"
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- 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
- 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
Landscapes
- Engineering & Computer Science (AREA)
- Business, Economics & Management (AREA)
- Physics & Mathematics (AREA)
- Economics (AREA)
- Theoretical Computer Science (AREA)
- Health & Medical Sciences (AREA)
- Human Resources & Organizations (AREA)
- Strategic Management (AREA)
- General Physics & Mathematics (AREA)
- Marketing (AREA)
- General Business, Economics & Management (AREA)
- Tourism & Hospitality (AREA)
- General Health & Medical Sciences (AREA)
- Artificial Intelligence (AREA)
- Biophysics (AREA)
- Game Theory and Decision Science (AREA)
- Development Economics (AREA)
- Public Health (AREA)
- Water Supply & Treatment (AREA)
- Operations Research (AREA)
- Primary Health Care (AREA)
- Life Sciences & Earth Sciences (AREA)
- Quality & Reliability (AREA)
- Biomedical Technology (AREA)
- Entrepreneurship & Innovation (AREA)
- Computational Linguistics (AREA)
- Data Mining & Analysis (AREA)
- Evolutionary Computation (AREA)
- Molecular Biology (AREA)
- Computing Systems (AREA)
- General Engineering & Computer Science (AREA)
- Mathematical Physics (AREA)
- Software Systems (AREA)
- Management, Administration, Business Operations System, And Electronic Commerce (AREA)
Abstract
The invention discloses a method/system, a storage medium and equipment for predicting total energy demand, which are used for arranging user energy consumption data according to the degree of association by utilizing a maximization matrix with maximized correlation coefficients and extracting the characteristics of the relevance and the like in and among energy consumption sequences of users by utilizing the maximization matrix to train a neural network, so that the constructed prediction model can more accurately predict the total energy demand of the users.
Description
Technical Field
The invention relates to the field of data processing, in particular to a prediction method/calculation system, a storage medium and equipment for total energy demand.
Background
Energy producers generally obtain total energy demand by counting energy demand of each energy using user, in the counting process, each energy using user is prepared for occasional needs, the declared energy demand is often larger than the actual demand, when the energy producer supplies energy for more energy using users, the declared total amount of each user can far exceed the final actual total usage, so that great resource waste is caused, and if the energy producer does not reduce the production amount according to the requirement, the energy producing user needs to bear the responsibility risk of energy outage.
Disclosure of Invention
In view of the above-mentioned shortcomings of the prior art, it is a primary object of the present invention to provide a method/system, storage medium and device for predicting total energy demand, so as to improve the accuracy of the prediction result.
In order to achieve the above objects and other related objects, the technical solution of the present invention is as follows:
a method for forecasting total energy demand, comprising:
acquiring existing data and outputting a maximized matrix;
inputting the maximization matrix into a pre-constructed neural network, and outputting the energy total demand prediction result of each user;
wherein the existing data includes historical usage amounts of users, and the method of outputting the maximization matrix includes: and arranging the historical use amount of each user according to a time sequence to form a historical use amount sequence, randomly arranging and configuring the historical use amount sequence of each user according to a direction different from the single historical use amount sequence to form a plurality of sequence matrixes, and calculating and screening the maximized matrix with the largest correlation coefficient in the sequence matrixes.
Optionally, the method for calculating and screening out the maximized matrix with the largest number of relationships in the sequence matrix includes: and calculating a correlation coefficient between the historical use sequences of any two users, and then calculating the correlation coefficient of each sequence matrix according to the correlation coefficient between the historical use sequences of any two users.
Optionally, the calculation formula of the correlation coefficient between the historical usage sequences of any two users is:
wherein i represents the current number of days, n represents the total number of days, x represents any one of the users,
represents the average usage of user x for 1 to n days, y represents any user other than x,
represents the average usage of user y in 1-n days, rx,yRepresenting the correlation coefficient for user x and user y. If the value of i is any one of 1 to n, the historical usage sequence of the user x in the period of 1 to n days is (x)1,x2,····,xn) User ofyThe historical usage sequence during 1-n days is (y)1,y2,····,yn)。
Optionally, the method for calculating the correlation coefficient of the sequence matrix includes: ignoring correlation coefficients between said historically used sequences in said sequence matrix having a distance greater than 3.
Optionally, the correlation coefficient calculation formula of the sequence matrix is:
wherein, N denotes the arrangement order of the single users in the sequence matrix.
Optionally, if the correlation coefficients between two or more sequence matrices are equal after ignoring the correlation coefficients between the historical usage sequences with the distance greater than 3 in the sequence matrices, the correlation coefficients of each sequence matrix are recalculated, and the correlation coefficients between the historical usage sequences with the distance greater than 3 in the sequence matrices are sequentially incremented during recalculation.
Optionally, the method for computationally screening the maximization matrix comprises a brute force exhaustive algorithm.
Optionally, the method for computing and screening the maximized matrix includes a greedy algorithm.
Optionally, before configuring the sequence matrix, it is verified whether the historical usage sequence of each user meets normal distribution, and if not, data adjustment is performed to make the historical usage sequence meet normal distribution.
Optionally, the method for adjusting data to make the historical usage sequence satisfy the normal distribution includes a logarithmic transformation and/or a square root transformation.
Optionally, the sequence matrix includes a two-dimensional matrix or a multi-dimensional matrix.
Optionally, the sequence matrix is a multidimensional matrix, the existing data further includes one or more of historical energy reservation amount data of each user, overhaul date data of each user, and holiday data of each user, the method for configuring and forming the sequence matrix further includes adding one or more of a predetermined amount sequence, an overhaul date sequence, and a holiday sequence of each user, one of the reservation amount sequences is formed by arranging historical energy reservation amounts of a single user in a time sequence, one of the overhaul date sequences is formed by arranging overhaul schedules of a single user in a time sequence, and further includes adding holiday sequences of each user, one of the holiday sequences is formed by arranging holiday arrangements of a single user in a time sequence.
Optionally, the number of data in the historical usage sequence is greater than 50.
Optionally, the method for predicting total energy demand further includes training the neural network, and the training method includes:
setting parameters of the neural network,
in the maximization matrix, randomly intercepting the historical energy consumption of each user in a continuous time period as a training sample;
and sequentially inputting the training samples into the neural network, and adjusting parameters according to an output result of the neural network based on the training samples until the error degree of the output result is lower than an expected threshold value.
Optionally, the neural network includes a convolutional neural network and a cyclic neural network, and the method of sequentially inputting the training samples into the neural network for training includes:
inputting the training sample into the convolutional neural network to extract local features;
and inputting the local features into the recurrent neural network to extract time sequence features, and outputting results.
Optionally, the parameters of the neural network include:
the number of days of the continuous period;
one or more of the number of layers, the convolution size and the network architecture of the convolutional neural network; and
one or more of the number of hidden layers, the number of neurons and the network structure in the recurrent neural network.
Optionally, the recurrent neural network includes one of a long-short term memory network, a gated recurrent neural network, and a neural turing machine.
Optionally, the method for inputting the maximized matrix into the pre-constructed neural network for training further includes: optimizing the neural network using a loss function and/or a gradient optimization algorithm.
The invention also provides a prediction system for the total energy demand, which comprises the following components:
a maximization matrix calculation unit for acquiring the existing data and outputting a maximization matrix,
the neural network is used for acquiring the maximization matrix and outputting the energy total demand prediction result of each user;
wherein the existing data includes historical usage amounts of the users, and the operation of the maximization matrix calculation unit outputting the maximization matrix includes: and arranging the historical use amount of each user according to a time sequence to form a historical use amount sequence, randomly arranging and configuring the historical use amount sequence of each user according to a direction different from the single historical use amount sequence to form a plurality of sequence matrixes, and calculating and screening the maximized matrix with the largest correlation coefficient in the sequence matrixes.
Optionally, the neural network includes:
the input matrix generation unit is used for intercepting historical energy consumption data of each user in a continuous time period in the maximization matrix to form an input matrix;
a convolutional neural network for extracting local features of the input matrix; and
and the cyclic neural network is used for extracting time sequence characteristics from the output result of the convolutional neural network and outputting a prediction result.
A storage medium having stored thereon a computer program which, when executed by a processor, implements the method for predicting total energy demand according to any one of claims 3 to 20.
An apparatus comprising a processor and a memory, the memory for storing a computer program, the processor for executing the computer program stored by the memory to cause the apparatus to perform the method for predicting total energy demand as claimed in any one of claims 3 to 20.
According to the method/system, the storage medium and the equipment for predicting the total energy demand, the correlation of the energy consumption of the users is identified by arranging the energy consumption data of the users according to the correlation degree by utilizing the maximization matrix with the maximized correlation coefficient, and the characteristics such as the correlation and the like in the energy consumption sequence of each user are extracted by utilizing the maximization matrix to train the neural network, so that the constructed prediction model can more accurately predict the total energy demand of each user.
Drawings
FIG. 1 is a block diagram of the total energy demand prediction system according to the present invention;
FIG. 2 is a network structure diagram of a neural network according to the present invention;
FIG. 3 is a flow chart illustrating the total energy demand prediction of the present invention;
FIG. 4 is a flow chart of the neural network training performed in the present invention.
Detailed Description
In order to make the objects, technical solutions and advantages of the present invention more apparent, embodiments of the present invention are described in further detail below with reference to the accompanying drawings. It is to be understood that the specific embodiments described herein are merely illustrative of the invention and are not limiting of the invention.
It should be further noted that, for the convenience of description, only some but not all of the relevant aspects of the present invention are shown in the drawings. Before discussing exemplary embodiments in more detail, it should be noted that some exemplary embodiments are described as processes or methods depicted as flowcharts. Although a flowchart may describe the operations (or steps) as a sequential process, many of the operations can be performed in parallel, concurrently or simultaneously. In addition, the order of the operations may be re-arranged. The process may be terminated when its operations are completed, but may have additional steps not included in the figure. The processes may correspond to methods, functions, procedures, subroutines, and the like.
For the purpose of clearly and clearly describing the contents of the embodiments of the present invention, the inventive concept of the present invention will be described first with emphasis on:
energy production users generally obtain total energy demand by adding energy demand reported by each energy use user, and use the total energy demand as a production plan, but most of the energy use users have a multi-report phenomenon in the reporting process, so that the reported total demand far exceeds the final actual use amount, and energy waste is caused. If the traditional algorithm is used for predicting the demand, part of the algorithms are too complex, and the prediction result of the part of algorithms is not accurate enough, the method mainly obtains the maximum matrix with the maximum correlation coefficient by using the historical use data of each user, and uses the maximum matrix as the input item for predicting the total energy demand, so that the correlation of the energy consumption among the users can be considered in the prediction result, and the prediction result is more accurate. The relevance between the users may be the relevance of the energy usage between the upstream and downstream enterprise users, or the relevance of the energy usage between the users having a competitive relationship, and the relevance is not limited to these two relevance, and may be other forms of relevance.
Referring to fig. 3, the present invention provides a prediction method for total energy demand, the method comprising:
210. obtaining, by one or more computing devices, existing data and outputting a maximized matrix;
220. inputting, by one or more computing devices, the maximization matrix into a pre-constructed neural network, and outputting a total energy demand prediction result of each user (e.g., a business or a family);
wherein the existing data includes historical usage amounts of users, and in step 210, the method for outputting the maximization matrix includes: and arranging the historical use amount of each user according to a time sequence to form a historical use amount sequence, randomly arranging and configuring the historical use amount sequence of each user according to a direction different from the single historical use amount sequence to form a plurality of sequence matrixes, and calculating and screening the maximized matrix with the largest correlation coefficient in the sequence matrixes.
The present embodiment also provides a storage medium having stored thereon a computer program corresponding to the method described above and below, which when executed by a processor, implements any of the methods described above and below for predicting total energy demand.
The storage medium in this embodiment can be understood by those skilled in the art as follows: all or a portion of the steps for implementing the method embodiments of the present description may be performed by computer program related hardware. The aforementioned computer program may be stored in a computer readable storage medium. When executed, performs steps comprising method embodiments of the present specification; and the aforementioned storage medium includes: various media that can store program codes, such as ROM, RAM, magnetic or optical disks.
The historical usage amount sequence may be a row vector or a column vector, and if the historical usage amount vector is a row vector, the historical usage amounts of the users may be sorted up and down when the sequence matrix is created, and if the historical usage amount vector is a column vector, the historical usage amounts of the users may be sorted left and right when the sequence matrix is created.
For example, if the energy usage users include user a, user b, user c … …, user p, and n is the total number of days, the historical usage sequence for user a may be (a)1,a2····ai····an) The created sequence matrix includes the first matrix:
(a1,a2····ai····an)
(b1,b2····bi····bn)
(c1,c2····ci····cn)
....
(p1,p2····pi····pn)
the created sequence matrix also comprises other sequences after the historical usage sequences of the users are randomly ordered.
In this embodiment, before configuring the sequence matrix, it is necessary to verify whether the historical usage sequence of each user satisfies the normal distribution, and if not, data adjustment is performed so that the historical usage sequence satisfies the normal distribution. Specifically, the method for adjusting the data so that the historical usage sequence satisfies the normal distribution may include a logarithmic transformation and/or a square root transformation.
In some embodiments, the method for calculating and screening out the maximized matrix with the largest number of relations in the sequence matrix comprises: and calculating a correlation coefficient between the historical use sequences of any two users, and then calculating the correlation coefficient of each sequence matrix according to the correlation coefficient between the historical use sequences of any two users.
In some embodiments, the correlation coefficient between the historical usage sequences of any two users may be calculated by a pearson correlation coefficient calculation formula:
wherein i represents the current number of days, n represents the total number of days, x represents any one of the users,
represents the average usage of user x for 1 to n days, y represents any user other than x,
represents the average usage of user y in 1-n days, rx,yRepresenting the correlation coefficient for user x and user y. If the value of i is any one of 1 to n, the historical usage sequence of the user x in the period of 1 to n days is (x)1,x2,····,xn) The historical usage sequence of user y during 1-n days is (y)1,y2,····,yn)。
In some embodiments, the method for calculating the correlation coefficient of the sequence matrix includes: ignoring correlation coefficients between historically used sequences in the sequence matrix that are more than 3 apart. The data processing amount can be greatly reduced, and the influence on the accuracy of the final prediction result is small.
Specifically, after ignoring the correlation coefficient between the historical use sequences with a distance greater than 3 in the sequence matrix, the correlation coefficient calculation formula of the sequence matrix is as follows:
wherein, N refers to the arrangement order of the single user in the sequence matrix.
In some embodiments, if the correlation coefficients between two or more sequence matrices are equal after ignoring the correlation coefficients between the historical use sequences with a distance greater than 3 in the sequence matrices, the correlation coefficients of the sequence matrices are recalculated, and the correlation coefficients between the historical use sequences with a distance greater than 3 in the sequence matrices are sequentially incremented when the correlation coefficients are recalculated. If the distance between two history use sequences is defined as d, the correlation coefficient calculation formula between the history use sequences with the distance d is as follows:
for example,when the correlation coefficients of two sequence matrixes are equal after ignoring the correlation coefficient between the historical use sequences with the distance larger than 3 in the sequence matrixes, the correlation coefficient between the historical use sequences with the distance of 4 can be increased
And recalculating the correlation coefficient of each matrix, wherein the obtained correlation coefficient calculation company is as follows:
if there are still sequence matrices with equal correlation coefficients after increasing the correlation coefficient between the historical use sequences with the distance of 4, the correlation coefficient of each matrix is calculated again, and the correlation coefficient between the historical use sequences with the distance of 5 is increased.
In some embodiments, the method for computationally screening the maximized matrix includes a brute-force exhaustive algorithm, that is, a correlation matrix of all sequence matrices is computed, and the sequence matrix with the largest correlation coefficient is selected as the maximized matrix.
In some embodiments, the method for calculating and screening the maximized matrix includes a greedy algorithm, that is, the correlation coefficient of the current sequence matrix is calculated, then the sequences of any two users are randomly exchanged, the correlation coefficient of the exchanged sequence matrix is calculated, if the correlation coefficient of the exchanged sequence matrix is greater than the correlation coefficient of the sequence matrix before the exchange, the exchanged sequence is retained, otherwise, the sequence before the exchange is retained, and then iteration is performed based on the principle until the value of the correlation coefficient after the exchange and the number of times of the exchange reaches a preset threshold value, and the iteration is stopped.
In some embodiments, the sequence matrix may be a two-dimensional matrix, i.e., containing only the users and the chronological sequence of historical usage by the users. In still other embodiments, the sequence matrix may be a multi-dimensional matrix, the existing data further includes one or more of historical energy reserve data for each user, service day data for each user, and holiday data for each user, the method configured to form the sequence matrix further includes adding one or more of a predetermined number sequence, a service day sequence, and a holiday sequence for each user, one of the reservation number sequences being formed by chronologically arranging historical energy reserves for individual users, one of the service day sequences being formed by chronologically arranging service schedules for individual users, and further including adding a holiday sequence for each user, one of the holiday sequences being formed by chronologically arranging vacation schedules for individual users. . During prediction, factors which may influence the use amount, such as the reservation amount, the maintenance day, the holidays and the like, are increased, and the prediction result is more accurate.
In some embodiments, the number of data in the historical usage sequence is greater than 50, that is, n is greater than 50, in an actual implementation process, the number of data in one historical usage sequence may be less than or equal to 50, or may be any number greater than 50, such as 60, 70, 80, 82, and the like, and when the number of data in a single historical usage sequence is greater than 50, it can be ensured that data for constructing a prediction model is sufficient, so that the historical usage sequence can meet the requirement of normal distribution after data processing.
Referring to fig. 4, in some embodiments, the prediction method for total energy demand of the present invention further includes training the neural network, and the training method includes:
221. setting parameters of the neural network,
222. in the maximization matrix, randomly intercepting the historical energy consumption of each user in a continuous time period as a training sample;
223. the training samples are sequentially input into the neural network, parameters are adjusted according to the output result of the neural network based on the training samples,
224. and circularly executing the steps 221-223 until the error degree of the output result is lower than an expected threshold value.
Specifically, when the historical usage sequence is a row vector, a column vector with the length of c is randomly intercepted as a training sample, and if an output result is defined as an output result of the t day, the training sample is the historical usage sequence of each user on the t-c day and the t-1 day.
In some embodiments, referring to fig. 4, the neural network includes a convolutional neural network and a cyclic neural network, and the method of inputting the training samples into the neural network in sequence for training includes:
223a, inputting the maximized matrix into the convolutional neural network to extract local features;
223b, inputting the local features into the recurrent neural network to extract time sequence features, and outputting results.
The convolutional neural network is beneficial to keeping the correlation among data in the maximization matrix when the characteristics are extracted, the cyclic neural network can extract time sequence characteristics, and after the neural network formed by the convolutional neural network and the cyclic neural network is subjected to combined training, a prediction model which is more excellent than that of other types of neural networks can be obtained, namely a more accurate prediction result is obtained.
In some embodiments, the parameters of the neural network include:
the number of days of the continuous period;
one or more of the number of layers, the convolution size and the network architecture of the convolutional neural network; and
in some embodiments, the Recurrent Neural network may include any one of a long Short Term Memory network (L ong-Term Memory, L STM), a Gated Recurrent Neural network (GRU), and a Neural Networks (NTM).
In some embodiments, the method of training by inputting the maximization matrix into the neural network constructed in advance further comprises: optimizing the neural network using a loss function and/or a gradient optimization algorithm. The prediction accuracy of the prediction model can be improved.
Referring to fig. 1, a predictive system for total energy demand is shown, comprising:
a maximization matrix calculation unit for outputting a maximization matrix according to the existing data;
the neural network is used for acquiring the maximization matrix and outputting the energy total demand prediction result of each user;
wherein the existing data includes historical usage amounts of the users, and the operation of the maximization matrix calculation unit outputting the maximization matrix includes: and arranging the historical use amount of each user according to a time sequence to form a historical use amount sequence, randomly arranging and configuring the historical use amount sequence of each user according to a direction different from the single historical use amount sequence to form a plurality of sequence matrixes, and calculating and screening the maximized matrix with the largest correlation coefficient in the sequence matrixes.
In some embodiments, referring to fig. 2, the neural network comprises:
an input matrix generating unit 21, configured to intercept historical energy usage data of each user in a continuous time period in the maximization matrix to form an input matrix;
a convolutional neural network 22 for extracting local features of the input matrix; and
and a recurrent neural network 23 for extracting a time series feature from an output result of the convolutional neural network and outputting a prediction result.
In practical implementation, the composition of the neural network is not limited to the above configuration, and a type of configuration may be adopted as long as the total energy demand can be predicted by using the maximization matrix.
In this embodiment, the Recurrent neural network is a Gated Recurrent neural network (GRU), and in an actual implementation process, the Recurrent neural network may be any one of a long-Short term memory network (L ong Short-term memory, <tttranslation & >t/t >tstm), a Gated Recurrent neural network (GRU), and a neural Network (NTM).
The invention also provides equipment comprising a processor and a memory, wherein the memory is used for storing a computer program, and the processor is used for executing the computer program stored by the memory so as to enable the equipment to execute any one of the prediction methods for the total energy demand.
The device provided by the embodiment comprises a processor, a memory, a transceiver and a communication interface, wherein the memory and the communication interface are connected with the processor and the transceiver and are used for realizing mutual communication, the memory is used for storing a computer program, the communication interface is used for carrying out communication, and the processor and the transceiver are used for operating the computer program to enable the device to execute the steps of the total energy demand prediction method.
In this embodiment, the Memory may include a Random Access Memory (RAM), and may also include a non-volatile Memory (non-volatile Memory), such as at least one disk Memory.
The Processor may be a general-purpose Processor, and includes a Central Processing Unit (CPU), a Network Processor (NP), and the like; the device can also be a Digital Signal Processor (DSP), an Application Specific Integrated Circuit (ASIC), a Field Programmable Gate Array (FPGA) or other Programmable logic device, a discrete Gate or transistor logic device, or a discrete hardware component.
The foregoing embodiments are merely illustrative of the principles and utilities of the present invention and are not intended to limit the invention. Any person skilled in the art can modify or change the above-mentioned embodiments without departing from the spirit and scope of the present invention. Accordingly, it is intended that all equivalent modifications or changes which can be made by those skilled in the art without departing from the spirit and technical spirit of the present invention be covered by the claims of the present invention.
Claims (22)
1. A prediction method for total energy demand is characterized by comprising
Acquiring existing data and outputting a maximized matrix;
inputting the maximization matrix into a pre-constructed neural network, and outputting the energy total demand prediction result of each user;
wherein the existing data includes historical usage amounts of users, and the method of outputting the maximization matrix includes: and arranging the historical use amount of each user according to a time sequence to form a historical use amount sequence, randomly arranging and configuring the historical use amount sequence of each user according to a direction different from the single historical use amount sequence to form a plurality of sequence matrixes, and calculating and screening the maximized matrix with the largest correlation coefficient in the sequence matrixes.
2. The method according to claim 1, wherein the step of calculating and screening the maximized matrix with the largest number of relationships in the sequence matrix comprises: and calculating a correlation coefficient between the historical use sequences of any two users, and then calculating the correlation coefficient of each sequence matrix according to the correlation coefficient between the historical use sequences of any two users.
3. The prediction method for the total energy demand according to claim 2, wherein the correlation coefficient between the historical usage sequences of any two users is calculated by the formula:
wherein i represents the current number of days, n represents the total number of days, x represents any one of the users,
represents the average usage of user x for 1 to n days, y represents any user other than x,
represents the average usage of user y in 1-n days, rx,yRepresenting a userx and the correlation coefficient of user y. If the value of i is any one of 1 to n, the historical usage sequence of the user x in the period of 1 to n days is (x)1,x2,····,xn) The historical usage sequence of user y during 1-n days is (y)1,y2,····,yn)。
4. The prediction method for total energy demand according to claim 3, wherein the method of calculating the correlation coefficient of the sequence matrix comprises: ignoring correlation coefficients between said historically used sequences in said sequence matrix having a distance greater than 3.
5. The prediction method for total demand for energy according to claim 4, wherein: the correlation coefficient calculation formula of the sequence matrix is as follows:
wherein, N denotes the arrangement order of the single users in the sequence matrix.
6. The prediction method for total demand for energy according to claim 4, wherein: and if the correlation coefficients between two or more than two sequence matrixes are equal after ignoring the correlation coefficients between the historical use sequences with the distance larger than 3 in the sequence matrixes, recalculating the correlation coefficients of the sequence matrixes, and sequentially increasing the correlation coefficients between the historical use sequences with the distance larger than 3 in the sequence matrixes during recalculation.
7. The prediction method for the total demand for energy according to claim 2, characterized in that: the method for calculating and screening the maximization matrix comprises a brute force exhaustion algorithm.
8. The prediction method for the total demand for energy according to claim 2, characterized in that: the method for computing and screening the maximization matrix comprises a greedy algorithm.
9. The prediction method for total demand for energy according to claim 1, wherein: before configuring a sequence matrix, verifying whether the historical use amount sequence of each user meets normal distribution, and if not, adjusting data to enable the historical use amount sequence to meet the normal distribution.
10. The prediction method for total demand for energy according to claim 9, wherein: the method for adjusting data to make the historical usage sequence satisfy the normal distribution comprises a logarithmic transformation and/or a square root transformation.
11. The prediction method for total demand for energy according to claim 1, wherein: the sequence matrix comprises a two-dimensional matrix or a multi-dimensional matrix.
12. The prediction method for total demand for energy according to claim 1, wherein: the sequence matrix is a multi-dimensional matrix, the existing data further comprises one or more of historical energy reservation amount data of each user, overhaul day data of each user and holiday data of each user, the method for forming the sequence matrix is configured, the method further comprises adding one or more of a predetermined amount sequence, an overhaul day sequence and a holiday sequence of each user, one reservation amount sequence is formed by arranging the historical energy reservation amount of each user in time sequence, one overhaul day sequence is formed by arranging overhaul arrangement of each user in time sequence, the method further comprises adding holiday sequence of each user, and one holiday sequence is formed by arranging holiday arrangement of each user in time sequence.
13. The prediction method for total demand for energy according to claim 1, wherein: the number of data in the historical usage sequence is greater than 50.
14. The prediction method for total energy demand according to claim 1, further comprising training the neural network, the training method comprising:
setting parameters of the neural network,
in the maximization matrix, randomly intercepting the historical energy consumption of each user in a continuous time period as a training sample;
and sequentially inputting the training samples into the neural network, and adjusting parameters according to an output result of the neural network based on the training samples until the error degree of the output result is lower than an expected threshold value.
15. The method according to claim 14, wherein the neural network comprises a convolutional neural network and a cyclic neural network, and the method of inputting the training samples into the neural network in sequence for training comprises: inputting the training sample into the convolutional neural network to extract local features;
and inputting the local features into the recurrent neural network to extract time sequence features, and outputting results.
16. The prediction method for total energy demand according to claim 15, wherein the parameters of the neural network include:
the number of days of the continuous period;
one or more of the number of layers, the convolution size and the network architecture of the convolutional neural network; and
one or more of the number of hidden layers, the number of neurons and the network structure in the recurrent neural network.
17. The prediction method for total demand for energy according to claim 15, wherein: the recurrent neural network comprises any one of a long-term and short-term memory network, a gated recurrent neural network and a neural turing machine.
18. The method of predicting total energy demand according to claim 1, wherein the method of training the maximization matrix input to the neural network constructed in advance further comprises: optimizing the neural network using a loss function and/or a gradient optimization algorithm.
19. A predictive system for total energy demand, comprising:
a maximization matrix calculation unit for acquiring the existing data and outputting a maximization matrix,
the neural network is used for acquiring the maximization matrix and outputting the energy total demand prediction result of each user;
wherein the existing data includes historical usage amounts of the users, and the operation of the maximization matrix calculation unit outputting the maximization matrix includes: and arranging the historical use amount of each user according to a time sequence to form a historical use amount sequence, randomly arranging and configuring the historical use amount sequence of each user according to a direction different from the single historical use amount sequence to form a plurality of sequence matrixes, and calculating and screening the maximized matrix with the largest correlation coefficient in the sequence matrixes.
20. The predictive system for total energy demand of claim 19, wherein the neural network comprises: the input matrix generation unit is used for intercepting historical energy consumption data of each user in a continuous time period in the maximization matrix to form an input matrix;
a convolutional neural network for extracting local features of the input matrix; and
and the cyclic neural network is used for extracting time sequence characteristics from the output result of the convolutional neural network and outputting a prediction result.
21. A storage medium having a computer program stored thereon, characterized in that: the program is executed by a processor to implement the method for predicting total energy demand according to any one of claims 1 to 18.
22. An apparatus, characterized by: comprising a processor and a memory for storing a computer program, the processor being configured to execute the computer program stored by the memory to cause the apparatus to perform the method for predicting total energy demand according to any one of claims 1 to 18.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201910046186.7A CN111461377B (en) | 2019-01-18 | 2019-01-18 | Prediction method/system, storage medium and apparatus for total energy demand |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201910046186.7A CN111461377B (en) | 2019-01-18 | 2019-01-18 | Prediction method/system, storage medium and apparatus for total energy demand |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN111461377A true CN111461377A (en) | 2020-07-28 |
| CN111461377B CN111461377B (en) | 2024-02-02 |
Family
ID=71682115
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201910046186.7A Active CN111461377B (en) | 2019-01-18 | 2019-01-18 | Prediction method/system, storage medium and apparatus for total energy demand |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN111461377B (en) |
Citations (9)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| JP2007199862A (en) * | 2006-01-24 | 2007-08-09 | Nippon Telegr & Teleph Corp <Ntt> | Energy demand predicting method, predicting device, program and recording medium |
| US20120109393A1 (en) * | 2010-07-28 | 2012-05-03 | Yoko Kosaka | Electric power demand/supply planning apparatus and method for the same |
| JP2012247998A (en) * | 2011-05-27 | 2012-12-13 | Toshiba Corp | Time series data prediction apparatus, prediction method, prediction program and memory medium |
| CN102938742A (en) * | 2012-11-08 | 2013-02-20 | 广东电网公司电力调度控制中心 | Communication bandwidth forecasting method and device based on power business requirements |
| CN104134096A (en) * | 2014-06-12 | 2014-11-05 | 国网安徽省电力公司 | Intelligent load prediction method based on province and prefecture integration |
| CN105069519A (en) * | 2015-07-16 | 2015-11-18 | 国网天津市电力公司 | Intelligent power grid park terminal user energy demand condition dynamic prediction system and method |
| CN106779129A (en) * | 2015-11-19 | 2017-05-31 | 华北电力大学（保定） | A kind of Short-Term Load Forecasting Method for considering meteorologic factor |
| CN107730054A (en) * | 2017-11-15 | 2018-02-23 | 西南石油大学 | A kind of Gas Load combination forecasting method based on support vector regression |
| CN108573323A (en) * | 2018-03-19 | 2018-09-25 | 国网浙江省电力有限公司嘉兴供电公司 | A kind of energy Internet user electricity demand forecasting method and system |
-
2019
- 2019-01-18 CN CN201910046186.7A patent/CN111461377B/en active Active
Patent Citations (9)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| JP2007199862A (en) * | 2006-01-24 | 2007-08-09 | Nippon Telegr & Teleph Corp <Ntt> | Energy demand predicting method, predicting device, program and recording medium |
| US20120109393A1 (en) * | 2010-07-28 | 2012-05-03 | Yoko Kosaka | Electric power demand/supply planning apparatus and method for the same |
| JP2012247998A (en) * | 2011-05-27 | 2012-12-13 | Toshiba Corp | Time series data prediction apparatus, prediction method, prediction program and memory medium |
| CN102938742A (en) * | 2012-11-08 | 2013-02-20 | 广东电网公司电力调度控制中心 | Communication bandwidth forecasting method and device based on power business requirements |
| CN104134096A (en) * | 2014-06-12 | 2014-11-05 | 国网安徽省电力公司 | Intelligent load prediction method based on province and prefecture integration |
| CN105069519A (en) * | 2015-07-16 | 2015-11-18 | 国网天津市电力公司 | Intelligent power grid park terminal user energy demand condition dynamic prediction system and method |
| CN106779129A (en) * | 2015-11-19 | 2017-05-31 | 华北电力大学（保定） | A kind of Short-Term Load Forecasting Method for considering meteorologic factor |
| CN107730054A (en) * | 2017-11-15 | 2018-02-23 | 西南石油大学 | A kind of Gas Load combination forecasting method based on support vector regression |
| CN108573323A (en) * | 2018-03-19 | 2018-09-25 | 国网浙江省电力有限公司嘉兴供电公司 | A kind of energy Internet user electricity demand forecasting method and system |
Non-Patent Citations (2)
| Title |
|---|
| D.J. NING: "A Data Oriented Analysis and Design Method for Smart Complex Software Systems of IoT", 《2018 INTERNATIONAL SYMPOSIUM IN SENSING AND INSTRUMENTATION IN IOT ERA (ISSI)》 * |
| 杨江;刘鹏飞;: "基于指标数据的供电可靠性预测方法", 云南电力技术, no. 04 * |
Also Published As
| Publication number | Publication date |
|---|---|
| CN111461377B (en) | 2024-02-02 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| Chang et al. | An adaptive sampling algorithm for solving Markov decision processes | |
| CN111027772B (en) | Multi-factor short-term load prediction method based on PCA-DBILSTM | |
| CN109886388A (en) | A kind of training sample data extending method and device based on variation self-encoding encoder | |
| CN110163429B (en) | Short-term load prediction method based on similarity day optimization screening | |
| CN104468413B (en) | A kind of network service method and system | |
| CN107544251B (en) | Single machine scheduling method for minimizing total stall period based on distributed robust model | |
| Ge et al. | Parameter estimation of a Gaussian mixture model for wind power forecast error by Riemann L-BFGS optimization | |
| CN109063939A (en) | A kind of wind speed forecasting method and system based on neighborhood door shot and long term memory network | |
| CN109787821B (en) | Intelligent prediction method for large-scale mobile client traffic consumption | |
| CN112884236B (en) | Short-term load prediction method and system based on VDM decomposition and LSTM improvement | |
| CN112949913A (en) | Photovoltaic output prediction method and terminal equipment | |
| Huh et al. | A general strategic capacity planning model under demand uncertainty | |
| CN115358437A (en) | Power supply load prediction method based on convolutional neural network | |
| Nielsen et al. | Inference on the dimension of the nonstationary subspace in functional time series | |
| Souhe et al. | Forecasting of electrical energy consumption of households in a smart grid | |
| CN110007371A (en) | Wind speed forecasting method and device | |
| CN112232570A (en) | Forward active total electric quantity prediction method and device and readable storage medium | |
| CN111461377A (en) | Prediction method/system, storage medium and equipment for total energy demand | |
| CN115099514A (en) | Data processing method and device, electronic equipment and storage medium | |
| CN114722490A (en) | Agent model global optimization method based on mixed increase and interval reduction | |
| Li et al. | Financial time series prediction based on GGM-GAN | |
| Lardin-Puech et al. | Analysing large datasets of functional data: a survey sampling point of view | |
| CN112667394A (en) | Computer resource utilization rate optimization method | |
| CN110348623A (en) | Complex Product Development time prediction and optimization method based on Design Structure Model | |
| Sun et al. | Asynchronous parallel surrogate optimization algorithm based on ensemble surrogating model and stochastic response surface method |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |