CN107103441A - Power Material sorting technique based on Self-Organizing Feature Maps - Google Patents
Power Material sorting technique based on Self-Organizing Feature Maps Download PDFInfo
- Publication number
- CN107103441A CN107103441A CN201710267283.XA CN201710267283A CN107103441A CN 107103441 A CN107103441 A CN 107103441A CN 201710267283 A CN201710267283 A CN 201710267283A CN 107103441 A CN107103441 A CN 107103441A
- Authority
- CN
- China
- Prior art keywords
- mrow
- goods
- materials
- power material
- msub
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Pending
Links
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
- G06Q10/08—Logistics, e.g. warehousing, loading or distribution; Inventory or stock management
- G06Q10/087—Inventory or stock management, e.g. order filling, procurement or balancing against orders
-
- 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
- G06F18/213—Feature extraction, e.g. by transforming the feature space; Summarisation; Mappings, e.g. subspace methods
- G06F18/2135—Feature extraction, e.g. by transforming the feature space; Summarisation; Mappings, e.g. subspace methods based on approximation criteria, e.g. principal component analysis
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/23—Clustering techniques
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/24—Classification techniques
-
- 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
- G06Q10/08—Logistics, e.g. warehousing, loading or distribution; Inventory or stock management
- G06Q10/087—Inventory or stock management, e.g. order filling, procurement or balancing against orders
- G06Q10/0875—Itemisation or classification of parts, supplies or services, e.g. bill of materials
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Business, Economics & Management (AREA)
- Data Mining & Analysis (AREA)
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Economics (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Evolutionary Biology (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Artificial Intelligence (AREA)
- Life Sciences & Earth Sciences (AREA)
- Bioinformatics & Computational Biology (AREA)
- Evolutionary Computation (AREA)
- General Engineering & Computer Science (AREA)
- Entrepreneurship & Innovation (AREA)
- General Business, Economics & Management (AREA)
- Tourism & Hospitality (AREA)
- Strategic Management (AREA)
- Quality & Reliability (AREA)
- Operations Research (AREA)
- Marketing (AREA)
- Human Resources & Organizations (AREA)
- Development Economics (AREA)
- Finance (AREA)
- Accounting & Taxation (AREA)
- Management, Administration, Business Operations System, And Electronic Commerce (AREA)
Abstract
The invention discloses the Power Material sorting technique based on Self-Organizing Feature Maps, comprise the following steps：Goods and materials data acquisition：Choose goods and materials data and the goods and materials data of selection are built into Power Material index system；Goods and materials data analysis：Data in Power Material index system are selected into Power Material principal component analysis index according to Nonlinear Principal Component Analysis method；Materials and equipment classification：Row index subdivision is entered to the Power Material principal component analysis index selected using Self-Organizing Feature Maps (SOFM) algorithm, and clustered according to subdivision index, the classification of Power Material is set up.
Description
Technical field
The present invention relates to Power Material sorting technique field, mainly in grid company handling of goods and materials link based on from group
Knit the Power Material sorting technique of Feature Mapping network.
Background technology
Because power industry is the industry of equipment property, its goods and materials expense is in the cost of power production and infrastructural project investment
All occupy sizable proportion.The selection of power equipment goods and materials needs to consider power network scale and power system operating mode, equipment
Proper mass and operation conditions, the accident of equipment is serious and the extent of injury, the compatibility of equipment, extraneous factor, weather conditions etc.
Factor.Therefore power network stockpile stock control difficulty is big, designs wide, how while power supply reliability is not reduced, protects
Hinder the orderly supply of Spare Parts, efficiently using limited inventory resource, farthest reduction inventory cost, which turns into, works as
Major issue and problem that preceding power network handling of goods and materials is faced.
The content of the invention
It is real it is an object of the invention to provide a kind of Power Material demand classification method based on Self-Organizing Feature Maps
Existing classification of the goods and materials based on demand characteristics, the reasonability of classification managed for Strategy of Inventory Control and optimize etc. have it is extremely important
Meaning；And while power supply reliability is not reduced, ensure the orderly supply of Spare Parts, efficiently utilize limited storehouse
Resource is deposited, farthest reduction inventory cost is realized.
To achieve these goals, the present invention provides following technical scheme：
Power Material sorting technique based on Self-Organizing Feature Maps, comprises the following steps：
Goods and materials data acquisition：Choose goods and materials data and the goods and materials data of selection are built into Power Material index system；
Goods and materials data analysis：Data in Power Material index system are selected into electricity according to Nonlinear Principal Component Analysis method
Power goods and materials principal component analysis index；
Materials and equipment classification：Index is finely divided to the Power Material principal component analysis index selected using SOFM algorithms, and
Clustered according to subdivision index, set up the classification of Power Material.
Further, the foundation for choosing goods and materials data is the herein below according to material requirements feature selecting：
1) the goods and materials amount of money：The goods and materials amount of money is the amount of money that goods and materials are spent within a time cycle；
2) shortage cost：The shortage cost is the cost increase caused by Power Material short supply；
3) order cycle time：The order cycle time is that material requirements carry the average value called time with goods and materials arrival time interval.
4) accounting is rushed to repair：The repairing accounting is that in a time cycle, item description field includes the bar of repairing character
The ratio of mesh number and total entry number；
5) number of times is received：The average value for receiving number of times that number of times is received for goods and materials in a time cycle；
6) quantity is received：The average value for receiving quantity that quantity is received for goods and materials in a time cycle；
7) interim purchase ratio：The interim purchase is than the amount of money and the ratio of total amount for goods and materials interim purchase；
8) amount of money is rushed to repair：The repairing amount of money is the total amount of repairing type of strip mesh number in goods and materials.
Further, the shortage cost is integrated according to goods and materials to construction delay cost, personnel cost and economic loss
Evaluation is divided into ten grades.
Further, the data by Power Material index system described in the goods and materials data analysis step are according to non-thread
Property PCA select the Nonlinear Principal Component Analysis method of Power Material principal component analysis index, be specially：
A. the goods and materials data to selection handle according to its index and sample size progress equalization and obtain equalization data；
B. to obtaining the data after logarithmic transformation after the carry out centralization logarithmic transformation of equalization data；
C. the covariance matrix of the data after logarithmic transformation is calculated；
D. the characteristic value and characteristic vector of the covariance matrix are calculated；
E. the contribution rate and contribution rate of accumulative total of goods and materials are calculated；
F. Power Material principal component analysis index is chosen：Each main composition score is added, then according to the number of addition
According to selecting Power Material principal component analysis index.
Further, the use SOFM algorithms are finely divided index to the Power Material principal component analysis index selected
Specially：
1) neutral net and network topology structure are built, wherein, the neutral net includes the god of input layer and competition layer
Through network, the network topology structure is rectangular lattice structure；
2) to neutral net initialization and the determination of the radius of neighbourhood：By the input node of neutral net and competition layer neuron
Between realize full connection, form n × m and connect, have N number of corresponding weight vector, to all weight vector { WijAssign [0,
1] random number in interval, and all weight vectors are normalized, determines initial neighborhood radius Ng(0), learn
Habit rate η (0) (0 ＜ η (0) ＜ 1)；
3) sample is inputted：Random one input pattern of extraction and it is normalized progress from training set
Input, obtains Xk=(X1, X2... Xn), n is input neuron number；Wherein：
4) triumph neuron is found：The Euclidean distance between input sample and whole competition layer neurons is calculated, it is calculated
Formula is：
Corresponding node when chosen distance is minimumFor competition triumph neuron：
5) winning neighborhood is defined：Centered on triumph neuron, a region of the radius of certain distance is set, in network
Learning process in, with the increase of iterations, the continuous self-organizing of weight vector and adjustment, winning neighborhood are constantly reduced into
Zero, pattern tends towards stability；
6)The neighborhood of triumph neuron is represented, it should meet：
7) weighed value adjusting：After sample is inputted, neuron in triumph neuron and its winning contiguous range can be with from group
Knit form and constantly adjust weight vector, with the passage of iterations, each neuron of competition layer tends to stable state.Weights
Vector adjustment is as follows：
8) check and terminate：Over time and iterations increase, learning rate can constantly reduce so that defeated
Enter being clustered for sample self-organization, judge that the standard terminated is gradually decreased as definite value or the average field of neuron for learning rate
Distance tends to definite value.
Further, the competition layer includes n × m neuron.
Further, the distance of the field radius is the 2/3 of all neighborhood distances of covering.
Further, the weight vector should be met：
By above-mentioned technical proposal, the invention has the advantages that：
In the present invention, the reasonability of classification has very important significance for Strategy of Inventory Control management and optimization etc.,
Sort research in text, the traditional materials and equipment classification method of difference, Power Material demand characteristics of giving overall consideration to is recorded using ERP goods and materials
Price of material, voltage class, the volume of data such as procurement cycle, purchase quantity, procurement value, research is set up special with material requirements
Property (importance, emergency, circulation etc.) characteristic be target NPCA-SOFM materials and equipment classification methods, realize goods and materials be based on demand
The classification of characteristic.
The present invention do not reduce power supply reliability while, ensure Spare Parts orderly supply, efficiently utilize
Limited inventory resource, farthest reduces inventory cost, and its key for solving problem is the rational stock's pipe of Power Material
Control.
Brief description of the drawings
Fig. 1 is the flow chart of the Power Material sorting technique proposed by the present invention based on Self-Organizing Feature Maps；
Fig. 2 is SOFM neutral net neighbour's neuron average distance iteration convergence curves in specific embodiment in the present invention；
Fig. 3 is significance level of each neuron on each attribute in the present invention.
Embodiment
The present invention is further illustrated below in conjunction with the accompanying drawings.
Embodiment
In the present embodiment, using Utilities Electric Co. of Guo Wang Shandong Province handling of goods and materials present situation as research object, company has been extracted
2 months 2013 3201838 full dose material storage data in July, 2016 are analyzed and researched.Target data set is with 1571
Individual goods and materials group observation row, and be made up of eight key indexs chosen for Column Properties, using Merrill Lynch's Tempo Data Analysis Platforms
Data analysis is carried out, checking analysis is carried out to the Power Material Segmentation Model of foundation.
Power Material sorting technique of the present invention based on material requirements characteristic, including：
Goods and materials data acquisition：Specifically, have chosen 1571 kinds of goods and materials data conducts from all goods and materials data 3201838
Study sample.
8 key indexs and its screening for choosing goods and materials data are described as follows：
1) the goods and materials amount of money：Wherein, the goods and materials amount of money is the amount of money that goods and materials are spent within a time cycle.
Specifically, the amount of money of the goods and materials amount of money reflection goods and materials in some cycles, goods and materials species is presented bright according to price height
Aobvious otherness, the goods and materials amount of money reflects the importance of such goods and materials to a certain extent.
2) shortage cost：Shortage cost is the cost increase caused by Power Material short supply.
Wherein, according to the existing shortage cost evaluation method of company, shortage cost is divided into ten grades, according to goods and materials pair
Construction delay cost, personnel cost and other economic losses carry out overall merit marking.
3) order cycle time：Order cycle time is that material requirements carry the average value called time with goods and materials arrival time interval.
Order cycle time reflects cycle goods and materials arrival time and complexity, and statistical is that the material requirements carry report
Time and the average value at goods and materials arrival time interval.
4) accounting is rushed to repair：It is that in a time cycle, item description field includes the entry of repairing character to rush to repair accounting
Number and the ratio of total entry number.
Repairing accounting mainly reflects the urgency level of the material requirements, and statistical is that project is retouched in cycle certain time
State the ratio of entry number of the field comprising " repairing " character and total entry number.Such as can specify that certain time for one month,
One season, 1 year etc..
5) number of times is received：Receive the average value of receiving number of times of the number of times for goods and materials in a time cycle；
The frequency of use that number of times reflects the goods and materials is received, statistical is that receiving for the cycle certain time interior goods and materials is secondary
Several average value.
6) quantity is received：Receive the average value of receiving quantity of the quantity for goods and materials in a time cycle；
Frequency of use and circulation that quantity equally reflects the goods and materials are received, statistical is should in cycle certain time
The average value for receiving quantity of goods and materials.Such as can be to have been received in one month 3 times, it is received quantity and removed for the summation that 3 times are received
With 3.
7) interim purchase ratio：Interim purchase is than the amount of money and the ratio of total amount for goods and materials interim purchase.
Urgency level of the interim purchase than mainly reflecting the goods and materials, index accounting mode is that buying type field is emergent
The amount of money of buying and the ratio of total amount.
8) amount of money is rushed to repair：The repairing amount of money is the total amount of repairing type of strip mesh number in goods and materials.
Goods and materials data analysis：It, which refers to, selects the data in Power Material index system according to Nonlinear Principal Component Analysis method
Go out Power Material principal component analysis index；
In the present embodiment, dimension-reduction treatment is carried out to index using non-linear Principal Component Analysis Method NPCA, dimension is solved and refers to
Mark variation influence, a small number of overall targets, and the characteristic informations for retaining initial data as far as possible are converted to by original index more.
In order to comprehensively be evaluated, selected evaluation index often compares many, so as to cause evaluation to become very multiple
Miscellaneous, Principal Component Analysis Method (Principal Component Analysis) is intended to utilize by the thought of dimension, and multidimensional index is turned
Be changed to the i.e. main composition of a small number of overall targets, each overall target comprehensively reflects the information of original variable as far as possible, and it is each it is main into
It is linear uncorrelated between part, comprising information it is not overlapping, so as to reduce the solution difficulty of problem, keep the effective of data analysis
Property.
But will there are many weak points in dimension in principal component analysis：One is to enter rower in order to eliminate dimension impact
During standardization, difference and information in index degree of variation can be caused to lose, so that characteristics extraction declines；Two be by
It is actually that one kind will linearly tie up technology in main composition, the main composition of gained is the linear combination of original index, it is impossible to which reflection is each
Non-linear relation between index and between main composition and each index.
In order to effectively make up the not enough there is provided the validity of algorithm of principal component analysis, the present invention uses the non-linear masters of NPCA
Analysis of components method is determined on the basis of goods and materials subdivision index, the goods and materials assessment indicator system built above, using non-linear master
Composition method enters to be about to dimension processing to original index, and centralized logarithm processing is carried out to variable, and zero situation is measured for becoming, is taken pair
" plus one less on the occasion of " processing is done during number to variable, the influence of dimension variation is eliminated, the validity that main composition is extracted is improved.
Wherein, the Non-linear Principal Component of Power Material principal component analysis index is selected according to Nonlinear Principal Component Analysis method
Analytic approach is comprised the following steps that：
A. equalization is handled：Provided with n sample, each sample has p indexs, then can obtain the every of every kind of goods and materials data
Initial data X=(the x of individual indexij)n×p, equalization processing is carried out to initial data：
Wherein,Obtain equalization data
B. centralization logarithmic transformation is done to equalization data：
Handled using above-mentioned formula 1-2 digital, obtain the data Z after logarithmic transformationij。
C. the covariance matrix of centralized logarithm data is calculated：
D. the eigenvalue λ of covariance matrix is calculatediWith characteristic vector li。
Wherein, S p characteristic value is designated as in covariance matrix：λ1≥λ2…λp, standardized feature vector is αij=(α1j,
α2j..., αip) (i=1,2 ..., p；J=1,2 ..., p),
Then i-th of m-th of sample index main composition is：
Wherein, i=1,2 ..., p；J=1,2 ..., p；K=1,2 ..., n, it is obvious that FkiIt is ykjNonlinear combination.
E. main composition contribution rate and contribution rate of accumulative total are calculated.
General selection contribution rate of accumulative total corresponding main composition m, this m main compositions of the characteristic value more than 80% are just comprehensive to be embodied
The most informations of all indexs.
Main composition contribution rate：
Main composition contribution rate of accumulative total：
F. each main composition score Z is calculatedij, the overall target segmented as Power Material.
The result walked to b carries out non-linear principal component analysis according to formula (1-3), carries out non-linear principal component analysis, main
The main composition of selection gist of composition adds up variance contribution ratio and is more than 80%, and by calculating, each main composition variance contribution ratio is such as
Shown in table 1 below：
Each main composition variance contribution degree of table 1
Comp is principal component, as seen from the table, main composition one, main composition two, the and of main composition three it can be seen from upper table 1
The accumulative variance contribution ratio of main composition four reaches 81.6% (being more than 80%), thus need to only choose main composition one, main composition two, it is main into
Part three and main composition four, you can represent each main Composition Factor loading matrixs of most information such as following table institute of 8 original indexs
Show, the order of magnitude of numerical value represents the contribution rate height of each factor in table, and absolute value is bigger, and contribution rate is higher, and negative number representation should
There is negative correlation in the factor, positive number represents that the factor has positive correlation with the main composition with the main composition：
Each main Composition Factor loading matrix of table 2
The first principal component is in the goods and materials amount of money, shortage cost and neck it can be seen from each main Composition Factor loading matrix of upper table 2
It is higher with load in three indexs of number of times, then it is assumed that the first principal component represents the importance degree of goods and materials；Second main composition exists
Rush to repair the amount of money and load on repairing quantity accounting two indices is higher, then it is assumed that the second main composition represents the emergency of goods and materials；
3rd main composition load on shortage cost and interim purchase accounting two indices is higher, then it is assumed that the 3rd main composition represents thing
The scarcity of money；4th main composition is in shortage cost and to receive load on quantity two indices higher, then it is assumed that the 4th main composition
Represent circulation.
The calculation formula of each overall target is：
Comp1=0.43x1+0.12x2+0.24x3+0.20x4+0.53x5+0.28x6
+0.34x7+0.47x8
Comp2=-0.35x1+0.33x2-0.41x3+0.55x4-0.13x5+0.32x6
-0.24x7+0.35x8
Comp3=-0.62x2+0.34x3+0.41x4-0.19x6-0.48x7+0.24x8
Comp4=0.63x2+0.43x3+0.21x4-0.12x5-0.6x6
According to above-mentioned formula, score that can be in the hope of all kinds of goods and materials on four main compositions obtains the data after dimensionality reduction such as
Table 3 below, and using this data set as next step SOFM neural network models training set, and then carry out goods and materials clustering.
The main composition score of each goods and materials of table 3
The 4 main compositions selected based on non-linear Principal Component Analysis Method, using SOFM neural network algorithms to Power Material
It is finely divided, by adjusting parameter and interpretation of result, SOFM neural network parameters is set to：Competition layer is empty for 2*2 two dimension
Between plane, be rectangular lattice structure, learning rate linearity variations interval is [0.05,0.01], and the radius of neighbourhood is 0.5, greatest iteration
Number of times is 100 times, now, and model obtains optimal effect, and cluster result is as shown in the table：
The Power Material cluster result of table 5
As seen from the above table, 1571 kinds of Power Materials are according to importance, emergency, four attributes of scarcity and circulation, quilt
The goods and materials attributive character being divided into four classes, each class is closely similar, contribute to for different classes of goods and materials formulate differentiation and
Personalized stock control scheme, foundation is provided for subdivision material requirements forecasting research, so as to improve the handling of goods and materials effect of enterprise
Rate.Group goods and materials during goods and materials class 1 includes 493, goods and materials class 2 includes 111 kinds of group goods and materials, and goods and materials class 3 includes 158 kinds of group things
Money, goods and materials class 4 includes 809 kinds of group goods and materials.
2, SOFM neutral net neighbour's neuron average distance iteration convergence curves referring to the drawings, as shown in Figure 2, from group
Knit in map neural network learning process, each interneuronal average distance of competition layer constantly subtracts with the increase of iterations
It is small, and a definite value is gradually converged to, when iterations reaches or so the 40th generation, in competition layer between neighbouring neuron
Average distance converges to 0.0205, now, and model tends towards stability state.
Referring to the drawings 3, it represents the significance level of each neuron on each attribute, can clearly reflect, the first kind
Goods and materials belong to the stronger goods and materials of circulation；Equations of The Second Kind goods and materials belong to the rare goods and materials of comparison, and shortage cost is higher；3rd class
Goods and materials lay particular emphasis on importance and circulation；4th class goods and materials belong to urgency level height and circulation is than faster goods and materials.In reality
In, enterprise can combine the specific object feature of all kinds of goods and materials, design different material requirements Forecasting Methodologies, improve prediction
Precision, in addition, can also formulate the handling of goods and materials strategy of differentiation for different goods and materials, improves the efficiency of operation of enterprise.
Analyzed by above table and accompanying drawing, the present invention is to the current handling of goods and materials present situation of Shandong provincial electric power company and thing
The deficiency that money subdivision is present, from the demand characteristics of Power Material, with reference to two kinds of algorithm advantages, is devised based on NPCA-
Goods and materials first, roughing index are segmented using non-linear Principal Component Analysis Method NPCA by the Power Material Segmentation Model of SOFM algorithms
Be converted to four overall targets：Importance, emergency, scarcity and circulation, reduce the complexity of problem；Then, base is built
In the goods and materials Segmentation Model of SOFM neural network algorithms, Power Material is subdivided into four major classes, is enterprise material differentiated demand
Prediction and personalized stock control provide reference significance, promote the lifting of enterprise material intensive management and operation benefits.
The foregoing description of the disclosed embodiments, enables those skilled in the art to realize or using the present invention.To this
A variety of modifications of a little embodiments will be apparent for a person skilled in the art, the General Principle defined in the present invention
It can realize in other embodiments without departing from the spirit or scope of the present invention.Therefore, the present invention will not be by
It is limited to the embodiments shown herein, and is to fit to consistent with features of novelty with principles disclosed herein most wide
Scope.
Claims (8)
1. the Power Material sorting technique based on Self-Organizing Feature Maps, it is characterised in that comprise the following steps：
Goods and materials data acquisition：Choose goods and materials data and the goods and materials data of selection are built into Power Material index system；
Goods and materials data analysis：Data in Power Material index system are selected into electric power thing according to Nonlinear Principal Component Analysis method
Provide principal component analysis index；
Materials and equipment classification：Index is finely divided to the Power Material principal component analysis index selected using SOFM algorithms, and according to
Subdivision index is clustered, and sets up the classification of Power Material.
2. the Power Material sorting technique according to claim 1 based on Self-Organizing Feature Maps, it is characterised in that
The foundation for choosing goods and materials data is the herein below according to material requirements feature selecting：
1) the goods and materials amount of money：The goods and materials amount of money is the amount of money that goods and materials are spent within a time cycle；
2) shortage cost：The shortage cost is the cost increase caused by Power Material short supply；
3) order cycle time：The order cycle time is that material requirements carry the average value called time with goods and materials arrival time interval.
4) accounting is rushed to repair：The repairing accounting is that in a time cycle, item description field includes the entry of repairing character
Number and the ratio of total entry number；
5) number of times is received：The average value for receiving number of times that number of times is received for goods and materials in a time cycle；
6) quantity is received：The average value for receiving quantity that quantity is received for goods and materials in a time cycle；
7) interim purchase ratio：The interim purchase is than the amount of money and the ratio of total amount for goods and materials interim purchase；
8) amount of money is rushed to repair：The repairing amount of money is the total amount of repairing type of strip mesh number in goods and materials.
3. the Power Material sorting technique according to claim 2 based on Self-Organizing Feature Maps, it is characterised in that
The shortage cost is divided into ten grades according to goods and materials to construction delay cost, personnel cost and economic loss overall merit.
4. the Power Material sorting technique according to claim 1 based on Self-Organizing Feature Maps, it is characterised in that
The data by Power Material index system described in the goods and materials data analysis step are selected according to Nonlinear Principal Component Analysis method
The Nonlinear Principal Component Analysis method of Power Material principal component analysis index is selected out, is specially：
A. the goods and materials data to selection handle according to its index and sample size progress equalization and obtain equalization data；
B. to obtaining the data after logarithmic transformation after the carry out centralization logarithmic transformation of equalization data；
C. the covariance matrix of the data after logarithmic transformation is calculated；
D. the characteristic value and characteristic vector of the covariance matrix are calculated；
E. the contribution rate and contribution rate of accumulative total of goods and materials are calculated；
F. Power Material principal component analysis index is chosen：Each main composition score is added, then selected according to the data of addition
Select out Power Material principal component analysis index.
5. the Power Material sorting technique according to claim 1 based on Self-Organizing Feature Maps, it is characterised in that
The use SOFM algorithms are finely divided index to the Power Material principal component analysis index selected：
1) neutral net and network topology structure are built, wherein, the neutral net includes input layer and the nerve net of competition layer,
The network topology structure is rectangular lattice structure；
2) to neutral net initialization and the determination of the radius of neighbourhood：By between the input node of neutral net and competition layer neuron
Full connection is realized, n × m connection is formed, has N number of corresponding weight vector, to all weight vector { WijAssign [0,1] area
Interior random number, and all weight vectors are normalized, determine initial neighborhood radius Ng(0), learning rate η
(0) (0 ＜ η (0) ＜ 1)；
3) sample is inputted：From training set random one input pattern of extraction and it is normalized progress it is defeated
Enter, obtain Xk=(X1, X2... Xn), n is input neuron number；Wherein：
<mrow>
<msub>
<mover>
<mi>X</mi>
<mo>&OverBar;</mo>
</mover>
<mi>k</mi>
</msub>
<mo>=</mo>
<mfrac>
<mi>X</mi>
<mrow>
<mo>|</mo>
<mi>X</mi>
<mo>|</mo>
</mrow>
</mfrac>
<mo>=</mo>
<msup>
<mrow>
<mo>&lsqb;</mo>
<mfrac>
<msub>
<mi>x</mi>
<mn>1</mn>
</msub>
<msqrt>
<mrow>
<msubsup>
<mi>&Sigma;</mi>
<mrow>
<mi>j</mi>
<mo>=</mo>
<mn>1</mn>
</mrow>
<mi>n</mi>
</msubsup>
<msubsup>
<mi>x</mi>
<mi>j</mi>
<mn>2</mn>
</msubsup>
</mrow>
</msqrt>
</mfrac>
<mo>,</mo>
<mn>...</mn>
<mo>,</mo>
<mfrac>
<msub>
<mi>x</mi>
<mi>n</mi>
</msub>
<msqrt>
<mrow>
<msubsup>
<mi>&Sigma;</mi>
<mrow>
<mi>j</mi>
<mo>=</mo>
<mn>1</mn>
</mrow>
<mi>n</mi>
</msubsup>
<msubsup>
<mi>x</mi>
<mi>j</mi>
<mn>2</mn>
</msubsup>
</mrow>
</msqrt>
</mfrac>
<mo>&rsqb;</mo>
</mrow>
<mi>T</mi>
</msup>
</mrow>
4) triumph neuron is found：Calculate the Euclidean distance between input sample and whole competition layer neurons, its calculation formula
For：
<mrow>
<msub>
<mi>d</mi>
<mi>j</mi>
</msub>
<mo>=</mo>
<msubsup>
<mi>&Sigma;</mi>
<mrow>
<mi>i</mi>
<mo>=</mo>
<mn>1</mn>
</mrow>
<mi>n</mi>
</msubsup>
<msup>
<mrow>
<mo>(</mo>
<msubsup>
<mi>X</mi>
<mi>i</mi>
<mi>k</mi>
</msubsup>
<mo>-</mo>
<msub>
<mi>W</mi>
<mrow>
<mi>i</mi>
<mi>j</mi>
</mrow>
</msub>
<mo>)</mo>
</mrow>
<mn>2</mn>
</msup>
<mo>,</mo>
<mi>i</mi>
<mo>&Element;</mo>
<mo>{</mo>
<mn>1</mn>
<mo>,</mo>
<mn>2</mn>
<mo>,</mo>
<mo>...</mo>
<mo>,</mo>
<mi>n</mi>
<mo>}</mo>
<mo>,</mo>
<mi>j</mi>
<mo>&Element;</mo>
<mo>{</mo>
<mn>1</mn>
<mo>,</mo>
<mn>2</mn>
<mo>,</mo>
<mo>...</mo>
<mo>,</mo>
<mi>m</mi>
<mo>}</mo>
</mrow>
Corresponding node when chosen distance is minimumFor competition triumph neuron：
<mrow>
<msub>
<mi>d</mi>
<msup>
<mi>j</mi>
<mo>*</mo>
</msup>
</msub>
<mo>=</mo>
<munder>
<mi>min</mi>
<mrow>
<mi>j</mi>
<mo>&Element;</mo>
<mo>{</mo>
<mn>1</mn>
<mo>,</mo>
<mn>2</mn>
<mo>,</mo>
<mn>...</mn>
<mo>,</mo>
<mi>m</mi>
<mo>}</mo>
</mrow>
</munder>
<mo>{</mo>
<msub>
<mi>d</mi>
<mi>j</mi>
</msub>
<mo>}</mo>
<mo>-</mo>
<mo>-</mo>
<mo>-</mo>
<mrow>
<mo>(</mo>
<mn>2</mn>
<mo>-</mo>
<mn>4</mn>
<mo>)</mo>
</mrow>
</mrow>
5) winning neighborhood is defined：Centered on triumph neuron, a region of the radius of certain distance is set, in network
During habit, with the increase of iterations, the continuous self-organizing of weight vector and adjustment, winning neighborhood are constantly reduced into zero, mould
Formula tends towards stability；
6)The neighborhood of triumph neuron is represented, it should meet：
<mrow>
<msub>
<mi>N</mi>
<msup>
<mi>j</mi>
<mo>*</mo>
</msup>
</msub>
<mo>=</mo>
<mo>{</mo>
<mi>m</mi>
<mo>,</mo>
<msub>
<mi>d</mi>
<mi>j</mi>
</msub>
<mo><</mo>
<mi>d</mi>
<mo>}</mo>
</mrow>
7) weighed value adjusting：After sample is inputted, the neuron in triumph neuron and its winning contiguous range can be with self-organizing shape
Formula constantly adjusts weight vector, with the passage of iterations, and each neuron of competition layer tends to stable state.Weight vector
Adjustment is as follows：
8) check and terminate：Over time and iterations increase, learning rate can constantly reduce so that input sample
Being clustered for this self-organization, judges that the standard terminated is gradually decreased as definite value or the average field distance of neuron for learning rate
Tend to definite value.
6. the Power Material sorting technique according to claim 5 based on Self-Organizing Feature Maps, it is characterised in that
The competition layer includes n × m neuron.
7. the Power Material sorting technique according to claim 5 based on Self-Organizing Feature Maps, it is characterised in that
The distance of the field radius is the 2/3 of all neighborhood distances of covering.
8. the Power Material sorting technique according to claim 5 based on Self-Organizing Feature Maps, it is characterised in that
The weight vector should be met：
<mrow>
<msubsup>
<mi>&Sigma;</mi>
<mrow>
<mi>i</mi>
<mo>=</mo>
<mn>1</mn>
</mrow>
<mi>n</mi>
</msubsup>
<msub>
<mi>W</mi>
<mrow>
<mi>i</mi>
<mi>j</mi>
</mrow>
</msub>
<mo>=</mo>
<mn>1.</mn>
</mrow>
2
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201710267283.XA CN107103441A (en) | 2017-04-21 | 2017-04-21 | Power Material sorting technique based on Self-Organizing Feature Maps |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201710267283.XA CN107103441A (en) | 2017-04-21 | 2017-04-21 | Power Material sorting technique based on Self-Organizing Feature Maps |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN107103441A true CN107103441A (en) | 2017-08-29 |
Family
ID=59657058
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201710267283.XA Pending CN107103441A (en) | 2017-04-21 | 2017-04-21 | Power Material sorting technique based on Self-Organizing Feature Maps |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN107103441A (en) |
Cited By (8)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN108154190A (en) * | 2018-01-12 | 2018-06-12 | 上海海事大学 | A kind of gantry crane motor status clustering method based on SOM neural networks |
| CN108985563A (en) * | 2018-06-12 | 2018-12-11 | 西安交通大学 | A kind of dynamically labeled method of Mechatronic Systems military service based on self-organizing feature map |
| CN110053052A (en) * | 2019-05-28 | 2019-07-26 | 北京理工大学 | The task of multilayer SOM isomery welding robot is distributed and paths planning method |
| CN110196911A (en) * | 2019-06-06 | 2019-09-03 | 申林森 | A kind of people's livelihood data automatic classification management system |
| CN110516988A (en) * | 2019-07-08 | 2019-11-29 | 国网浙江省电力有限公司金华供电公司 | One kind being fitted Power Material method based on neural network |
| CN112270977A (en) * | 2020-10-29 | 2021-01-26 | 南通市第一人民医院 | ICU patient material use management method and system |
| CN113592309A (en) * | 2021-08-02 | 2021-11-02 | 上海华能电子商务有限公司 | Multi-level inventory quota making method based on data driving |
| CN115409255A (en) * | 2022-08-24 | 2022-11-29 | 广东电网有限责任公司广州供电局 | Method for managing life cycle of electric power material |
-
2017
- 2017-04-21 CN CN201710267283.XA patent/CN107103441A/en active Pending
Cited By (12)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN108154190A (en) * | 2018-01-12 | 2018-06-12 | 上海海事大学 | A kind of gantry crane motor status clustering method based on SOM neural networks |
| CN108985563A (en) * | 2018-06-12 | 2018-12-11 | 西安交通大学 | A kind of dynamically labeled method of Mechatronic Systems military service based on self-organizing feature map |
| CN110053052A (en) * | 2019-05-28 | 2019-07-26 | 北京理工大学 | The task of multilayer SOM isomery welding robot is distributed and paths planning method |
| CN110053052B (en) * | 2019-05-28 | 2021-02-09 | 北京理工大学 | Task allocation and path planning method for multilayer SOM heterogeneous welding robot |
| CN110196911A (en) * | 2019-06-06 | 2019-09-03 | 申林森 | A kind of people's livelihood data automatic classification management system |
| CN110196911B (en) * | 2019-06-06 | 2022-04-22 | 申林森 | Automatic classification management system for civil data |
| CN110516988A (en) * | 2019-07-08 | 2019-11-29 | 国网浙江省电力有限公司金华供电公司 | One kind being fitted Power Material method based on neural network |
| CN112270977A (en) * | 2020-10-29 | 2021-01-26 | 南通市第一人民医院 | ICU patient material use management method and system |
| CN113592309A (en) * | 2021-08-02 | 2021-11-02 | 上海华能电子商务有限公司 | Multi-level inventory quota making method based on data driving |
| CN113592309B (en) * | 2021-08-02 | 2024-04-30 | 上海华能电子商务有限公司 | Multilevel inventory quota formulation method based on data driving |
| CN115409255A (en) * | 2022-08-24 | 2022-11-29 | 广东电网有限责任公司广州供电局 | Method for managing life cycle of electric power material |
| CN115409255B (en) * | 2022-08-24 | 2023-09-08 | 广东电网有限责任公司广州供电局 | Electric power material life cycle management method |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN107103441A (en) | Power Material sorting technique based on Self-Organizing Feature Maps | |
| US20170308934A1 (en) | Management method of power engineering cost | |
| CN109063976A (en) | Intelligence manufacture capability maturity evaluation method based on Fuzzy AHP | |
| CN106971240A (en) | The short-term load forecasting method that a kind of variables choice is returned with Gaussian process | |
| CN107423769A (en) | Electric load curve adaptive clustering scheme based on morphological feature | |
| CN105117810A (en) | Residential electricity consumption mid-term load prediction method under multistep electricity price mechanism | |
| Halter et al. | Use of simulation in evaluating management policies under uncertainty: application to a large scale ranch | |
| CN101329683A (en) | Recommendation system and method | |
| CN108710905A (en) | One kind being based on the united spare part quantitative forecasting technique of multi-model and system | |
| CN105976054A (en) | Measuring instrument storage system goods location optimization method | |
| CN107508287A (en) | Electricity grid substation load grouping method, device, storage medium and computer equipment | |
| CN104240040A (en) | City smart power grid planning method on basis of land conversion risk | |
| CN110378510A (en) | A kind of distribution material requirements prediction technique being polymerize based on time series and level | |
| CN111428766B (en) | Power consumption mode classification method for high-dimensional mass measurement data | |
| CN111667090A (en) | Load prediction method based on deep belief network and weight sharing | |
| CN107993033A (en) | A kind of Power Material Forecasting Methodology | |
| CN104820942A (en) | Electricity market trade evaluation criterion measuring and calculating method based on hierarchical clustering | |
| CN110796296A (en) | Multi-objective investment optimization model construction method and device and computing equipment | |
| CN112990584B (en) | Automatic production decision system and method based on deep reinforcement learning | |
| CN103136709A (en) | Typical fault tree evaluation method based on grid equipment | |
| CN117455147A (en) | Complete set early warning method for production of multi-model small-batch complex equipment products | |
| CN114004530B (en) | Enterprise electric power credit modeling method and system based on ordering support vector machine | |
| Chen et al. | A decision-making mechanism considering carbon footprint and cost to fulfil orders for multi-site global companies | |
| CN107679753A (en) | A kind of electricity market principle market power evaluation method based on weighted Mahalanobis distance method | |
| Alikhani et al. | Optimal demand response programs selection using CNN‐LSTM algorithm with big data analysis of load curves |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| RJ01 | Rejection of invention patent application after publication | ||
| RJ01 | Rejection of invention patent application after publication |
Application publication date: 20170829 |