CN110443413B - Electric power material demand prediction system and construction method of electric power material demand model - Google Patents
Electric power material demand prediction system and construction method of electric power material demand model Download PDFInfo
- Publication number
- CN110443413B CN110443413B CN201910664566.7A CN201910664566A CN110443413B CN 110443413 B CN110443413 B CN 110443413B CN 201910664566 A CN201910664566 A CN 201910664566A CN 110443413 B CN110443413 B CN 110443413B
- Authority
- CN
- China
- Prior art keywords
- electric power
- demand
- power material
- model
- prediction
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Active
Links
Images
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/21—Design or setup of recognition systems or techniques; Extraction of features in feature space; Blind source separation
- G06F18/214—Generating training patterns; Bootstrap methods, e.g. bagging or boosting
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/24—Classification techniques
- G06F18/241—Classification techniques relating to the classification model, e.g. parametric or non-parametric approaches
- G06F18/2411—Classification techniques relating to the classification model, e.g. parametric or non-parametric approaches based on the proximity to a decision surface, e.g. support vector machines
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/08—Learning methods
- G06N3/084—Backpropagation, e.g. using gradient descent
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
- G06Q10/00—Administration; Management
- G06Q10/04—Forecasting or optimisation specially adapted for administrative or management purposes, e.g. linear programming or "cutting stock problem"
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
- G06Q10/00—Administration; Management
- G06Q10/06—Resources, workflows, human or project management; Enterprise or organisation planning; Enterprise or organisation modelling
- G06Q10/063—Operations research, analysis or management
- G06Q10/0631—Resource planning, allocation, distributing or scheduling for enterprises or organisations
- G06Q10/06315—Needs-based resource requirements planning or analysis
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
- G06Q10/00—Administration; Management
- G06Q10/06—Resources, workflows, human or project management; Enterprise or organisation planning; Enterprise or organisation modelling
- G06Q10/067—Enterprise or organisation modelling
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
- G06Q50/00—Systems or methods specially adapted for specific business sectors, e.g. utilities or tourism
- G06Q50/06—Electricity, gas or water supply
Abstract
The invention provides a method for constructing a power material demand model, which comprises the following steps: s1, acquiring the whole process data of the electric power material; s2, constructing a prediction model for forming the multi-level comprehensive power material demand by using the whole process data of the power material acquired in S1 and adopting a mode of combining a least square support vector machine, an echo state network and a regularization limit learning machine, wherein the prediction model comprises the following steps: s21, establishing a sample database; s22, predicting the demand of the electric power materials by adopting a least square support vector machine; s23, predicting the demand of the electric power materials by adopting an echo state network; s24, forecasting the demand of the electric power materials by adopting a regularization limit learning machine; and S25, weighting the prediction results to obtain a final prediction result of the demand of the electric power materials. Meanwhile, an electric power material demand forecasting system is provided, and the system comprises an electric power material demand model constructed by the method. The invention effectively improves the prospect of material management of the power grid company, creates favorable conditions for enterprises to extract overall resources, and reduces the operation cost of the power grid enterprise while ensuring the operation reliability support of the power grid.
Description
Technical Field
The invention relates to the technical field of power system material management, in particular to a power material demand prediction system and a construction method of a power material demand model.
Background
The electric power is the basic pillar industry of national economy, the development of each industry and the growth of the national economy are closely related to the electric power, once an accident happens to a transformer substation and a power transmission and distribution line, the stability of a power grid is influenced to form large-area power failure, so that the power supply reliability is influenced, and huge loss is caused to a user, so that the normal construction and operation of the power grid are the basis for ensuring the normal life of people and the normal activities of the country.
The electric power material is the basic guarantee in the power grid construction process, the quality of the material is not only related to the safe operation of the power grid, but also related to the cost management of a company, the reasonable formulation and material purchasing directly influence the safe production, so the quality of the material purchasing is related to the safe and stable power production.
The quality of the power material is not only related to the safe operation of the power grid, but also related to the cost management of the company. According to the material management target of 'centralization, unification and efficiency priority' of network companies, a relatively perfect material management system, a standardized system and an information management system are established. However, the current situations of material data fragmentation, material storage mechanization and subject responsibility fuzzification in the current material management work enable the current related management data of the power grid materials to be fragmented and the operation data to be isolated, so that the information of the materials is mastered by a power grid company as if the information is blindly felt, and the goal of material management work refinement is directly influenced.
In summary, with the development and transformation of power grid enterprises, the connotation and the extension of the responsibility of material companies are expanded, and the problem of data utilization is a main bottleneck limiting the cross-type development of material management work. With the development of technologies such as the internet of things, big data, artificial intelligence and the like, the data acquisition and utilization are more convenient and faster than those in any period in the past, so that the data of electric power materials in the purchasing and production links are fully utilized, the management level of the power grid materials is improved, and the important breakthrough point that the operation cost of enterprises is reduced to improve the material management work is achieved.
At present, no explanation or report of the similar technology of the invention is found, and similar data at home and abroad are not collected.
Disclosure of Invention
Aiming at the defects in the prior art, the invention provides an electric power material demand forecasting system and a construction method of an electric power material demand model, which are characterized in that on the basis of fully researching and analyzing the electric power material demand, the Beidou, the Internet of things, the crawler and other data acquisition technologies are adopted, the material related data dispersed in different systems such as the current materials, the living meters and the like are integrated into the same data platform to form the material overall process data, and the overall process data is utilized to construct and form the multi-level comprehensive electric power material demand forecasting model. The invention effectively improves the prospect of material management of the power grid company, creates favorable conditions for enterprises to extract overall resources, and reduces the operation cost of the power grid enterprise while ensuring the operation reliability support of the power grid.
In order to achieve the purpose, the invention adopts the following technical scheme.
According to one aspect of the invention, a method for constructing a power material demand model is provided, which comprises the following steps:
s1, acquiring the whole process data of the electric power material;
s2, constructing a prediction model for forming the multi-level comprehensive power material demand by using the whole process data of the power material acquired in S1 and adopting a mode of combining a least square support vector machine, an echo state network and a regularization limit learning machine, wherein the prediction model comprises the following steps:
s21, establishing a training sample database;
s22, predicting the demand of the electric power materials by adopting a least square support vector machine to obtain a prediction result V of the demand of the electric power materials1;
S23, predicting the demand of the electric power materials by adopting an echo state network to obtain a prediction result V of the demand of the electric power materials2;
S24, predicting the demand of the electric power materials by adopting a regularization limit learning machine to obtain a prediction result V of the demand of the electric power materials3;
And S25, weighting the prediction results of S22-S24 to obtain a prediction model of the power material demand.
Preferably, the S21, including:
historical information X (N) -x (x) of N data in the whole process data of the power material at N time points1(n),…,xN(n))TAs training sample input vectors; l pieces of electric power material demand information y (n) ═ y at corresponding n time points1(n),…,yL(n))TOutputting a vector as a training sample; the total process data of the electric power material at m time points for predicting the demand of the electric power material is recorded as t (m) ═ t1(m),…,tN(m))TAnd forming a training sample database.
Preferably, in S22, the result of predicting the demand of the electric power material by using the least squares support vector machine is recorded as V1The prediction process comprises the following steps:
s221, establishing a least square support vector machine model: mapping the training sample obtained in the step S21 to a high-dimensional kernel space, taking a radial basis function as a kernel function, and determining an optimal parameter of the least square support vector machine model by using a lagrange multiplier method, thereby obtaining the least square support vector machine model:
κ(x,xi)=φ(x)Tφ(xi)=exp(-||x-xi||2/2σ2)
where φ (x) is a mapping function that maps x to a high-dimensional kernel space; b is a least square vector machine model parameter; alpha is alphaiIs lagrange multiplier; c is a regularization parameter; omega is a weight vector; y isiAn ith output vector of the training sample; epsiloniIs a relaxation variable; kappa (x, x)i) Is a kernel function; σ is the width of the nucleus, taken as
S222, forecasting the demand of the electric power materials: inputting the whole process data T (m) of the power material into the least square support vector machine model1(m),…,tN(m))TObtaining the predicted value V of the demand of the electric power material1The predicted value V of the demand of the electric power material1Comprises the following steps:
preferably, in S23, the result of predicting the demand of the power material by using the echo state network is recorded as V2The prediction process comprises the following steps:
s231, initializing parameters of the echo state network: randomly determining node dimension K of the reserve pool, and randomly generating a reserve pool state matrix W with a spectrum radius smaller than 1K×KInput matrixAnd output feedback weight matrixWherein the pool state matrix WK×KInput matrixAnd output feedback weight matrixOnce determined, remain unchanged during the training and testing phases;
s232, based on the training sample database obtained in S21, inputting the training sample into the echo state network, calculating and recording the reserve pool state u (n) ═ u (n) of the echo state network1(n),…,uK(n))T:
In the formula, f is an activation function of the reserve pool unit and is selected as a hyperbolic tangent function;
f(*)=tanh(*);
s233, calculating an output weight matrix according to the reserve pool state U and the output vector Y
In the formula (I), the compound is shown in the specification,is [ X (n), U (n), Y (n-1)]The pseudo-inverse of (1);
s234, predicting the demand of the electric power materials: according to the trained echo state network and the whole process data of the electric power materials, the demand V for the electric power materials is2The prediction is carried out as follows:
preferably, in S24, the result of predicting the demand of the electric power material by using the regularized limit learning machine is recorded as V3The prediction process comprises the following steps:
s241, randomly determining an input weight matrix A and a bias vector q;
s242, training a regularization extreme learning machine: using the training sample input vector X (X) in the training sample database obtained in S211,…,xN)TAnd training sample output vector Y ═ Y (Y)1,…,yL)TTraining a regularization extreme learning machine:
in the formula, aijIs an element in the input weight matrix A; g (, is an activation function, and is selected as a softplus function; h () is a row vector in matrix H; i is an identity matrix; gamma is a regularization coefficient;is the trained output weight;
g(*)=log(1+e*);
s243, based on the data t (m) of the whole process of the electric power material (t ═ m)1(m),…,tN(m))TForecasting electric power material demand V3Comprises the following steps:
preferably, in S25, the prediction result V of the integrated least squares support vector machine is1Network prediction result V of echo state2And regularizing the extreme learning machine prediction result V3And obtaining a final prediction result V, namely a prediction model:
preferably, the S1, including:
acquiring material purchasing data; aiming at information required by different links in the material purchasing process, establishing a coding rule and/or an identification principle covering necessary data required by the material purchasing process;
acquiring material logistics information; tracking and self-identifying the destination of the electric power materials by adopting Beidou short messages and position services;
acquiring material application information; and aiming at relevant data or information in a test report, a management system and/or an online monitoring system of equipment in the electric power materials, a web crawler method is adopted for information retrieval.
Preferably, the information required by different links in the material purchasing process at least comprises: any one or any plurality of supplier, equipment, time, location, user, quantity, purchase lot, shipping, and acceptance information.
According to another aspect of the invention, an electric power material demand forecasting system is provided, which comprises an electric power material demand model, wherein the electric power material demand model is constructed by any one of the methods.
Compared with the prior art, the invention has the following beneficial effects:
1. the invention improves the service and product level of the power material supplier, improves the power material supply quality and material purchasing efficiency by comprehensively analyzing data in the aspects of equipment management, operation and the like, and provides important support for power grid construction.
2. The information construction of the current power grid materials mainly focuses on the acquisition of state information of the equipment after the equipment is put into operation, the data of the whole process of the power materials basically presents a fracture state, and the data of the whole process is not communicated.
Drawings
Other features, objects and advantages of the invention will become more apparent upon reading of the detailed description of non-limiting embodiments with reference to the following drawings:
FIG. 1 is a schematic diagram illustrating a process of constructing a power material demand model according to an embodiment of the present invention;
FIG. 2 is a flowchart of a web crawler method according to an embodiment of the present invention;
fig. 3 is a schematic diagram of an ESN network structure according to an embodiment of the present invention.
Detailed Description
The following examples illustrate the invention in detail: the embodiment is implemented on the premise of the technical scheme of the invention, and a detailed implementation mode and a specific operation process are given. It should be noted that, for a person skilled in the art, several variations and modifications can be made without departing from the inventive concept, which falls within the scope of the present invention.
The embodiment of the invention provides a method for constructing a demand model of electric power materials, which comprises the following steps of:
s1, acquiring the whole process data of the electric power material;
and S2, constructing and forming a prediction model of the multi-level comprehensive power material demand by using the data of the whole process of the power material acquired in S1 and adopting a mode of combining a least square support vector machine, an echo state network and a regularization limit learning machine.
In S2, a multi-level comprehensive prediction model of a least squares support vector machine, an echo state network, and a regularization extreme learning machine is used to improve the prediction accuracy. The principle is as follows:
(1) least Square Support Vector Machine (LSSVM)
LSSVM is an extension of a support vector machine, and the method replaces inequality constraint in SVM with equality constraint, so that the original quadratic programming problem solution is converted into the solution of a linear equation set, and the optimization problem can be expressed as
Equation (1) can be solved according to the lagrange multiplier method and the KKT condition,
(2) echo State Network (Echo State Network, ESN)
The ESN is a novel recurrent neural network, which can be used for predicting a nonlinear time series, and the network structure is shown in fig. 3. In the figure, u (n), y (n) and x (n) respectively represent input samples, output samples and internal state vectors; win、Wintr、Wout、WbackRespectively representing an input weight matrix, an internal weight matrix, an output weight matrix and an output feedback matrix. As shown in fig. 3.
The ESN adopts a large-scale coefficient network to establish the mapping from the low-dimensional input samples to the high-dimensional state space, adopts a linear regression method to train to obtain the output connection weight, greatly simplifies the network training, and has the state vector and output vector updating formula as
x(n+1)=f(Winu(n+1)+Wintrx(n)+Wbacky(n)) (2)
y(n+1)=fout(Wout[u(n+1),x(n+1)]) (3)
In the formula: f and foutRespectively, activation functions of the reservoir and the output layer.
(3) Regularization Extreme Learning Machine (ELM)
ELM is a new type single hidden layer feedforward neural network, and its input weight and hidden layer bias are randomly set. The method aims at minimizing the training error, solves the output weight by utilizing the Moore-Penrose generalized inverse matrix theory, and obviously improves the network learning speed and generalization performance. However, random initialization of network parameters may degrade its prediction performance and stability. The output weight value is calculated as
In the formula: i is an identity matrix; h is a hidden layer output matrix; and T is a training sample output vector.
Further, based on the above principle, the S2 includes:
and S21, establishing a training sample database. Historical information X (N) -x (x) of N data in the whole process data of the power material at N time points1(n),…,xN(n))TAs training sample input vectors; l pieces of electric power material demand information y (n) ═ y at corresponding n time points1(n),…,yL(n))TOutputting a vector as a training sample; the total process data of the electric power material at m time points for predicting the demand of the electric power material is recorded as t (m) ═ t1(m),…,tN(m))T。
S22, based on the data of the whole process of the electric power material, the demand of the electric power material is predicted by adopting a least square support vector machine, and the result is recorded as V1The electric power material demand prediction process comprises the following steps:
1) establishing a least square support vector machine model, mapping the training sample obtained in the step S21 to a high-dimensional kernel space, taking a radial basis function as a kernel function, and determining the optimal parameters of the least square support vector machine model by adopting a Lagrange multiplier method so as to obtain the least square support vector machine model;
κ(x,xi)=φ(x)Tφ(xi)=exp(-||x-xi||2/2σ2)
where φ (x) is a map that maps x to a high dimensional kernel spaceA function of rays; b is a least square vector machine model parameter; alpha is alphaiIs lagrange multiplier; c is a regularization parameter; omega is a weight vector; y isiAn ith output vector of the training sample; epsiloniIs a relaxation variable; kappa (x, x)i) Is a kernel function; σ is the width of the nucleus, taken as
2) Predicting the demand of electric power materials: inputting power material process data T (m) ═ t into the least square support vector machine model1(m),…,tN(m))TObtaining the predicted value V of the demand of the electric power material1The predicted value V of the demand of the electric power material1The calculation formula of (2) is as follows:
s23, predicting the demand of the electric power materials by adopting an echo state network based on the data of the whole process of the electric power materials, and recording the result as V2The electric power material demand prediction process comprises the following steps:
1) initializing network parameters in an echo state: randomly determining node dimension K of the reserve pool, and randomly generating a reserve pool state matrix W with a spectrum radius smaller than 1K×KInput matrixAnd output feedback weight matrixWherein the pool state matrix WK ×KInput matrixAnd output feedback weight matrixOnce determined, remain unchanged during the training and testing phases;
2) based on the training sample database of S21, inputting training samples to the echo state network to calculate and record the reserve pool state U (n) ═ u (n) of the echo state network1(n),…,uK(n))T(ii) a The calculation formula of the echo state network reserve pool state U is as follows:
in the formula, f is an activation function of the reserve pool unit and is selected as a hyperbolic tangent function;
f(*)=tanh(*);
3) calculating an output weight matrix according to the state U and the output vector Y of the reserve poolThe output weight matrix of the echo state networkThe calculation formula of (2) is as follows:
in the formula (I), the compound is shown in the specification,is [ X (n), U (n), Y (n-1)]The pseudo-inverse of (1);
4) predicting the demand of electric power materials, and according to the trained echo state network and the electric power whole process data, predicting the demand V of the electric power materials2Carrying out prediction; the electric power material demand prediction result V2Comprises the following steps:
V2(i)=Wout[T(i),U(n+i),V2(i-1)]i=1,…,m。
s24, based on the whole process data of the electric power materialPredicting the demand of the electric power materials by adopting a regularization extreme learning machine, and recording the result as V3The electric power material demand prediction process comprises the following steps:
1) randomly determining an input weight matrix A and a bias vector q;
2) training a regularization extreme learning machine: the training sample input vector X in the training sample database obtained in S21 is (X)1,…,xN)TAnd training sample output vector Y ═ Y (Y)1,…,yL)TTraining a regularization extreme learning machine, wherein the training method of the regularization extreme learning machine comprises the following steps:
in the formula, aijIs an element in the input weight matrix A; g (, is an activation function, and is selected as a softplus function; h () is a row vector in matrix H; i is an identity matrix; gamma is a regularization coefficient;is the trained output weight;
g(*)=log(1+e*);
3) based on the whole process data T (m) of the power material1(m),…,tN(m))TForecasting electric power material demand V3The electric power material requirement V3The prediction method comprises the following steps:
s25, synthesizing the least square support vector machine prediction result V1Network prediction result V of echo state2And regularizing the extreme learning machine prediction result V3Obtaining a final prediction result V, namely a prediction model; the prediction model is as follows:
further, the S1 includes:
acquiring material purchasing data; aiming at information required by different links in the material purchasing process, establishing a coding rule and/or an identification principle covering necessary data required by the material purchasing process;
acquiring material logistics information; tracking and self-identifying the destination of the electric power materials by adopting Beidou short messages and position services; the navigation technology of the Beidou satellite navigation system can realize rapid positioning through a mobile communication operation base station, and can realize active reporting and analysis of position information in the whole processes of delivery, transportation, warehouse entry and exit, commissioning and the like according to identity curing information of electric power material equipment and position service of Beidou. The Beidou terminal sends the short message through the information transceiver of the Beidou terminal and the Beidou satellite, and the short message is transmitted through single packet data with several to hundreds of bytes. The Beidou short message transceiver hardware forming module mainly comprises a short message transceiver antenna module, a message coding module, a message decoding module and a message transmission module. The short message receiving and transmitting antenna module is mainly used for receiving and transmitting Beidou wireless short messages; the message coding module mainly realizes the error correction coding function of the message and carries out digital/analog processing on the coded wireless message into an analog message through an intermediate short message amplifier; the message decoding module is mainly used for acquiring data of the analog short message processed by the encoding wireless message, and completing the geographical position positioning decoding of the short message collection, tracking, mode identification, clock synchronization and navigation short message; the main function of the message transmission module is to provide navigation information to the client to fulfill the corresponding requirement.
Acquiring material application information; for relevant data or information in a test report, a management system and/or an online monitoring system of equipment in the electric power material, a web crawler method as shown in fig. 2 is adopted for information retrieval. The web crawler method becomes an important component of a search engine, and related webpage information can be more delicately selected and grabbed according to a set target.
Further, the information required by different links in the material purchasing process at least comprises the following information: any one or any plurality of supplier, equipment, time, location, user, quantity, purchase lot, shipping, and acceptance information.
The embodiment of the invention also provides a power material demand forecasting system which comprises a power material demand model, wherein the power material demand model is constructed by the method provided by the embodiment of the invention.
Specifically, an electric power material demand model in the electric power material demand prediction system mainly predicts the failure probability of equipment and the power grid failure disaster degree, wherein the failure probability of different areas, different times and different equipment is related to statistical analysis of various dimensions of the shipped materials such as variety, time, position, failure defects and the like, historical failure defect data and the like, the power grid material disaster degree is related to weather early warning data, equipment operation and maintenance data combined with historical weather severity, disaster degree and the like, the data size is large, influence factors are more, single prediction methods are limited in prediction, and stable and excellent prediction accuracy cannot be guaranteed for any data sample. In the embodiment of the invention, a multi-level comprehensive prediction model of a least square support vector machine, an echo state network and a regularization extreme learning machine is comprehensively adopted to improve the prediction precision.
The foregoing description of specific embodiments of the present invention has been presented. It is to be understood that the present invention is not limited to the specific embodiments described above, and that various changes and modifications may be made by one skilled in the art within the scope of the appended claims without departing from the spirit of the invention.
Claims (8)
1. A construction method of a power material demand model is characterized by comprising the following steps:
s1, acquiring the whole process data of the electric power material;
s2, constructing a prediction model for forming the multi-level comprehensive power material demand by using the whole process data of the power material acquired in S1 and adopting a mode of combining a least square support vector machine, an echo state network and a regularization limit learning machine, wherein the prediction model comprises the following steps:
s21, establishing a training sample database;
s22, predicting the demand of the electric power materials by adopting a least square support vector machine to obtain a prediction result V of the demand of the electric power materials1The prediction process comprises the following steps:
s221, establishing a least square support vector machine model: mapping the training sample obtained in the step S21 to a high-dimensional kernel space, taking a radial basis function as a kernel function, and determining an optimal parameter of the least square support vector machine model by using a lagrange multiplier method, thereby obtaining the least square support vector machine model:
κ(x,xi)=φ(x)Tφ(xi)=exp(-||x-xi||2/2σ2)
where φ (x) is a mapping function that maps x to a high-dimensional kernel space; b is a least square vector machine model parameter; alpha is alphaiIs lagrange multiplier; c is a regularization parameter; omega is a weight vector; y isiAn ith output vector of the training sample; epsiloniIs a relaxation variable; kappa (x, x)i) Is a kernel function; σ is the width of the nucleus, taken as
S222, forecasting the demand of the electric power materials: inputting the whole process data T (m) of the power material into the least square support vector machine model1(m),…,tN(m))TObtaining the predicted value V of the demand of the electric power material1The predicted value V of the demand of the electric power material1Comprises the following steps:
s23, predicting the demand of the electric power materials by adopting an echo state network to obtain a prediction result V of the demand of the electric power materials2;
S24, predicting the demand of the electric power materials by adopting a regularization limit learning machine to obtain a prediction result V of the demand of the electric power materials3;
And S25, weighting the prediction results of S22-S24 to obtain a prediction model of the power material demand.
2. The method for constructing the electric power material demand model according to claim 1, wherein the S21 includes:
historical information X (N) -x (x) of N data in the whole process data of the power material at N time points1(n),…,xN(n))TAs training sample input vectors; l pieces of electric power material demand information y (n) ═ y at corresponding n time points1(n),…,yL(n))TOutputting a vector as a training sample; the total process data of the electric power material at m time points for predicting the demand of the electric power material is recorded as t (m) ═ t1(m),…,tN(m))TAnd forming a training sample database.
3. The method for constructing the model of demand for electric power supplies according to claim 2, wherein the result of predicting the demand for electric power supplies using the echo state network is denoted as V232The prediction process comprises the following steps:
s231, initializing parameters of the echo state network: randomly determining reserve pool node dimensionDegree K, randomly generating a reserve pool state matrix W with the spectrum radius less than 1K×KInput matrixAnd output feedback weight matrixWherein the pool state matrix WK ×KInput matrixAnd output feedback weight matrixOnce determined, remain unchanged during the training and testing phases;
s232, based on the training sample database obtained in S21, inputting the training sample into the echo state network, calculating and recording the reserve pool state u (n) ═ u (n) of the echo state network1(n),…,uK(n))T:
In the formula, f is an activation function of the reserve pool unit and is selected as a hyperbolic tangent function;
f(*)=tanh(*);
s233, calculating an output weight matrix according to the reserve pool state U and the output vector Y
In the formula (I), the compound is shown in the specification,is [ X (n), U (n), Y (n-1)]The pseudo-inverse of (1);
s234, predicting the demand of the electric power materials: according to the trained echo state network and the whole process data of the electric power materials, the demand V for the electric power materials is2The prediction is carried out as follows:
4. the method for constructing the electric power material demand model according to claim 2, wherein in step S24, the result of predicting the electric power material demand by using a regularized limit learning machine is recorded as V3The prediction process comprises the following steps:
s241, randomly determining an input weight matrix A and a bias vector q;
s242, training a regularization extreme learning machine: using the training sample input vector X (X) in the training sample database obtained in S211,…,xN)TAnd training sample output vector Y ═ Y (Y)1,…,yL)TTraining a regularization extreme learning machine:
in the formula, aijIs an element in the input weight matrix A; g (, is an activation function, and is selected as a softplus function; h () is a row vector in matrix H; i is an identity matrix; gamma is a regularization coefficient;is the trained output weight;
g(*)=log(1+e*);
s243, based on the data t (m) of the whole process of the electric power material (t ═ m)1(m),…,tN(m))TForecasting electric power material demand V3Comprises the following steps:
5. the method for constructing the power material demand model according to claim 2, wherein the step S25 is performed by using a prediction result V of an integrated least squares support vector machine1Network prediction result V of echo state2And regularizing the extreme learning machine prediction result V3And obtaining a final prediction result V, namely a prediction model:
6. the method for constructing the electric power material demand model according to claim 1, wherein the S1 includes:
acquiring material purchasing data; aiming at information required by different links in the material purchasing process, establishing a coding rule and/or an identification principle covering necessary data required by the material purchasing process;
acquiring material logistics information; tracking and self-identifying the destination of the electric power materials by adopting Beidou short messages and position services;
acquiring material application information; and aiming at relevant data or information in a test report, a management system and/or an online monitoring system of equipment in the electric power materials, a web crawler method is adopted for information retrieval.
7. The method for constructing the electric power material demand model according to claim 6, wherein the information required by different links in the material purchasing process at least comprises: any one or any plurality of supplier, equipment, time, location, user, quantity, purchase lot, shipping, and acceptance information.
8. An electric power material demand forecasting system, characterized by comprising an electric power material demand model, wherein the electric power material demand model is constructed by the method of any one of claims 1 to 7.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201910664566.7A CN110443413B (en) | 2019-07-23 | 2019-07-23 | Electric power material demand prediction system and construction method of electric power material demand model |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201910664566.7A CN110443413B (en) | 2019-07-23 | 2019-07-23 | Electric power material demand prediction system and construction method of electric power material demand model |
Publications (2)
Publication Number | Publication Date |
---|---|
CN110443413A CN110443413A (en) | 2019-11-12 |
CN110443413B true CN110443413B (en) | 2021-08-27 |
Family
ID=68431116
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN201910664566.7A Active CN110443413B (en) | 2019-07-23 | 2019-07-23 | Electric power material demand prediction system and construction method of electric power material demand model |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN110443413B (en) |
Families Citing this family (10)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN108764584B (en) * | 2018-06-05 | 2021-04-09 | 国网浙江省电力有限公司 | Enterprise electric energy substitution potential evaluation method |
CN111598288A (en) * | 2020-02-29 | 2020-08-28 | 贵州电网有限责任公司 | Material demand prediction management method based on material overall process data |
CN111369058B (en) * | 2020-03-05 | 2022-06-14 | 中国民用航空飞行学院 | Forest fire fighting helicopter demand prediction method and system |
CN111639815B (en) * | 2020-06-02 | 2023-09-05 | 贵州电网有限责任公司 | Method and system for predicting power grid defect materials through multi-model fusion |
CN111639817A (en) * | 2020-06-05 | 2020-09-08 | 山东大学 | Emergency material demand prediction method and system for power grid meteorological disasters |
CN112613642B (en) * | 2020-12-07 | 2024-03-15 | 国网北京市电力公司 | Emergency material demand prediction method and device, storage medium and electronic equipment |
CN112465258A (en) * | 2020-12-11 | 2021-03-09 | 国网北京市电力公司 | Discontinuous power material demand prediction method and system |
CN113283657B (en) * | 2021-06-01 | 2023-05-26 | 国网山东省电力公司嘉祥县供电公司 | Prediction method for electric power emergency material demand |
CN114152622A (en) * | 2021-11-30 | 2022-03-08 | 南方电网物资有限公司 | Electric power material warehouse entry detection system |
CN115619098A (en) * | 2022-10-26 | 2023-01-17 | 国网浙江省电力有限公司物资分公司 | Intelligent electric power material data processing method based on grading monitoring and early warning |
Citations (3)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN107292428A (en) * | 2017-06-07 | 2017-10-24 | 国网浙江省电力公司物资分公司 | A kind of distribution Power Material procurement demand forecasting system |
CN107784397A (en) * | 2017-11-09 | 2018-03-09 | 贵州电网有限责任公司 | A kind of power network material requirements forecasting system and its Forecasting Methodology |
CN209085657U (en) * | 2017-08-02 | 2019-07-09 | 强力物联网投资组合2016有限公司 | For data gathering system related or industrial environment with chemical production technology |
Family Cites Families (3)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN102831489B (en) * | 2012-08-29 | 2016-03-09 | 广东电网公司 | Power matching network builds material requirements Forecasting Methodology and device |
CN104881707B (en) * | 2015-05-04 | 2018-07-03 | 同济大学 | A kind of sintering energy consumption Forecasting Methodology based on integrated model |
CN106203701A (en) * | 2016-07-06 | 2016-12-07 | 吴本刚 | A kind of power matching network builds material requirements prognoses system |
-
2019
- 2019-07-23 CN CN201910664566.7A patent/CN110443413B/en active Active
Patent Citations (3)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN107292428A (en) * | 2017-06-07 | 2017-10-24 | 国网浙江省电力公司物资分公司 | A kind of distribution Power Material procurement demand forecasting system |
CN209085657U (en) * | 2017-08-02 | 2019-07-09 | 强力物联网投资组合2016有限公司 | For data gathering system related or industrial environment with chemical production technology |
CN107784397A (en) * | 2017-11-09 | 2018-03-09 | 贵州电网有限责任公司 | A kind of power network material requirements forecasting system and its Forecasting Methodology |
Non-Patent Citations (3)
Title |
---|
"基于ESN的光伏发电功率预测模型研究";彭光虎 等;《青岛大学学报(工程技术版)》;20150930;第30卷(第3期);全文 * |
"基于最小二乘的支持向量机在电力需求预测中的应用";黄元生 等;《中国管理科学》;20051031;第13卷;全文 * |
"基于极限学习机的电力配网物资需求预测";杨晶晶 等;《2012年中国电机工程学会年会论文集》;20140313;全文 * |
Also Published As
Publication number | Publication date |
---|---|
CN110443413A (en) | 2019-11-12 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN110443413B (en) | Electric power material demand prediction system and construction method of electric power material demand model | |
Hong et al. | Energy forecasting: A review and outlook | |
Hammad et al. | Methods and models for electric load forecasting: a comprehensive review | |
Ahmad et al. | Quantum GIS based descriptive and predictive data analysis for effective planning of waste management | |
EP3822880A1 (en) | Load prediction method and apparatus based on neural network | |
Zhang et al. | Spatial-temporal solar power forecast through use of Gaussian conditional random fields | |
CN115049171B (en) | Photovoltaic power prediction method and system based on feature migration | |
Afrasiabi et al. | Deep learning architecture for direct probability density prediction of small‐scale solar generation | |
Zhou et al. | Accurate prediction of photovoltaic power output based on long short‐term memory network | |
Carrera et al. | PVHybNet: A hybrid framework for predicting photovoltaic power generation using both weather forecast and observation data | |
Zhao et al. | Spatio‐temporal Markov chain model for very‐short‐term wind power forecasting | |
Zhang et al. | All‐factor short‐term photovoltaic output power forecast | |
Zhao et al. | Station-level short-term demand forecast of carsharing system via station-embedding-based hybrid neural network | |
CN112101612A (en) | Prediction method of town gas usage, storage medium and terminal equipment | |
Liu et al. | Boosted multi-task learning for inter-district collaborative load forecasting | |
Gallina et al. | Work in progress level prediction with long short-term memory recurrent neural network | |
ȚIBOACĂ et al. | Design of short-term wind production forecasting model using machine learning algorithms | |
Lu et al. | Residential Energy Consumption Forecasting Based on Federated Reinforcement Learning with Data Privacy Protection. | |
Brenner et al. | Learning spatio-temporal aggregations for large-scale capacity expansion problems | |
Shi et al. | Study on quick judgement of small signal stability using CNN | |
Huang et al. | Application of an extension method on optimal site selection for PV power systems: A case study in Taiwan | |
Sun et al. | Optimization of a multi-objective location model of manufacturing base considering cooperative manufacturing capabilities and service benefits. | |
Guo et al. | Short‐term photovoltaic power forecasting with adaptive stochastic configuration network ensemble | |
Ghiassi et al. | On the use of AI as a requirement for improved insolation forecasting accuracy to achieve optimized PV utilization | |
Yu et al. | A multiple spatio‐temporal features fusion approach for short‐term passenger flow forecasting in urban rail transit |
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 |