CN113239029A - Completion method for missing daily freezing data of electric energy meter - Google Patents
Completion method for missing daily freezing data of electric energy meter Download PDFInfo
- Publication number
- CN113239029A CN113239029A CN202110539537.5A CN202110539537A CN113239029A CN 113239029 A CN113239029 A CN 113239029A CN 202110539537 A CN202110539537 A CN 202110539537A CN 113239029 A CN113239029 A CN 113239029A
- Authority
- CN
- China
- Prior art keywords
- data
- electric energy
- missing
- energy meter
- daily
- 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
- 230000008014 freezing Effects 0.000 title claims abstract description 39
- 238000007710 freezing Methods 0.000 title claims abstract description 39
- 238000000034 method Methods 0.000 title claims abstract description 33
- 238000012549 training Methods 0.000 claims abstract description 41
- 238000012360 testing method Methods 0.000 claims abstract description 31
- 238000010586 diagram Methods 0.000 claims abstract description 24
- 230000005611 electricity Effects 0.000 claims abstract description 23
- 238000012795 verification Methods 0.000 claims abstract description 21
- 239000013598 vector Substances 0.000 claims description 6
- 230000009467 reduction Effects 0.000 claims description 5
- 238000013528 artificial neural network Methods 0.000 claims description 4
- 238000005457 optimization Methods 0.000 claims description 3
- 230000008569 process Effects 0.000 claims description 3
- 230000000306 recurrent effect Effects 0.000 claims description 3
- 238000005096 rolling process Methods 0.000 claims description 3
- 238000004458 analytical method Methods 0.000 abstract description 3
- 238000012544 monitoring process Methods 0.000 abstract description 3
- 238000011084 recovery Methods 0.000 abstract description 3
- 239000011159 matrix material Substances 0.000 description 8
- 230000006870 function Effects 0.000 description 7
- 238000004422 calculation algorithm Methods 0.000 description 3
- 238000011160 research Methods 0.000 description 3
- XEEYBQQBJWHFJM-UHFFFAOYSA-N Iron Chemical compound [Fe] XEEYBQQBJWHFJM-UHFFFAOYSA-N 0.000 description 2
- 230000000295 complement effect Effects 0.000 description 2
- 239000013589 supplement Substances 0.000 description 2
- 206010060904 Freezing phenomenon Diseases 0.000 description 1
- 230000002159 abnormal effect Effects 0.000 description 1
- 230000006399 behavior Effects 0.000 description 1
- 230000009286 beneficial effect Effects 0.000 description 1
- 230000005540 biological transmission Effects 0.000 description 1
- 238000004364 calculation method Methods 0.000 description 1
- 230000008859 change Effects 0.000 description 1
- 238000004891 communication Methods 0.000 description 1
- 230000007547 defect Effects 0.000 description 1
- 230000003111 delayed effect Effects 0.000 description 1
- 238000012217 deletion Methods 0.000 description 1
- 230000037430 deletion Effects 0.000 description 1
- 238000011161 development Methods 0.000 description 1
- 230000000694 effects Effects 0.000 description 1
- 238000005265 energy consumption Methods 0.000 description 1
- 238000005516 engineering process Methods 0.000 description 1
- 238000007689 inspection Methods 0.000 description 1
- 229910052742 iron Inorganic materials 0.000 description 1
- 230000014759 maintenance of location Effects 0.000 description 1
- 238000003062 neural network model Methods 0.000 description 1
- 238000010606 normalization Methods 0.000 description 1
- 238000010248 power generation Methods 0.000 description 1
- 238000012545 processing Methods 0.000 description 1
- 238000007619 statistical method Methods 0.000 description 1
- 230000001502 supplementing effect Effects 0.000 description 1
- 230000001360 synchronised effect Effects 0.000 description 1
- 238000010200 validation analysis Methods 0.000 description 1
Images
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F16/00—Information retrieval; Database structures therefor; File system structures therefor
- G06F16/20—Information retrieval; Database structures therefor; File system structures therefor of structured data, e.g. relational data
- G06F16/21—Design, administration or maintenance of databases
- G06F16/215—Improving data quality; Data cleansing, e.g. de-duplication, removing invalid entries or correcting typographical errors
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/04—Architecture, e.g. interconnection topology
- G06N3/044—Recurrent networks, e.g. Hopfield networks
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- 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—Information and communication technology [ICT] specially adapted for implementation of business processes of specific business sectors, e.g. utilities or tourism
- G06Q50/06—Energy or water supply
Landscapes
- Engineering & Computer Science (AREA)
- Business, Economics & Management (AREA)
- Theoretical Computer Science (AREA)
- Physics & Mathematics (AREA)
- Human Resources & Organizations (AREA)
- Strategic Management (AREA)
- General Physics & Mathematics (AREA)
- Economics (AREA)
- Health & Medical Sciences (AREA)
- General Business, Economics & Management (AREA)
- General Health & Medical Sciences (AREA)
- Entrepreneurship & Innovation (AREA)
- Data Mining & Analysis (AREA)
- General Engineering & Computer Science (AREA)
- Marketing (AREA)
- Quality & Reliability (AREA)
- Tourism & Hospitality (AREA)
- Databases & Information Systems (AREA)
- Evolutionary Computation (AREA)
- Software Systems (AREA)
- Mathematical Physics (AREA)
- Development Economics (AREA)
- Computing Systems (AREA)
- Educational Administration (AREA)
- Molecular Biology (AREA)
- Game Theory and Decision Science (AREA)
- Computational Linguistics (AREA)
- Operations Research (AREA)
- Biophysics (AREA)
- Biomedical Technology (AREA)
- Artificial Intelligence (AREA)
- Life Sciences & Earth Sciences (AREA)
- Public Health (AREA)
- Water Supply & Treatment (AREA)
- Primary Health Care (AREA)
- Management, Administration, Business Operations System, And Electronic Commerce (AREA)
Abstract
The invention discloses a method for complementing missing day freezing data of an electric energy meter, belongs to the technical field of electric power marketing electric quantity monitoring and analysis, relates to a method for complementing missing data of the electric energy meter, and particularly relates to a method for complementing missing data of the electric energy meter, which is used for complementing the missing data of the electric energy meter with missing day freezing data and certain days freezing data. The method comprises the steps of obtaining basic information of a user and relevant characteristic factors of electricity utilization; forming an electric energy data time series characteristic diagram; dividing the training set, the verification set and the test set; transmitting the electric energy data time series characteristic diagram to an input layer of the LSTM network; training and modeling the LSTM network by using the training set data to obtain an LSTM model; inputting the verification set into an LSTM model to determine the optimal parameters of the model; testing the LSTM model; restoring data of the date missing date of the daily freezing data; and outputting a daily freezing data restoration result of the missing date. The method fills data loss, improves the data quality of power supply enterprises, improves the recovery accuracy of the lost data, and has high missing completion accuracy.
Description
Technical Field
The invention discloses a method for complementing missing day freezing data of an electric energy meter, belongs to the technical field of electric power marketing electric quantity monitoring and analysis, relates to a method for complementing missing data of the electric energy meter, and particularly relates to a method for complementing missing data of the electric energy meter, which is used for complementing the missing data of the electric energy meter with missing day freezing data and certain days freezing data.
Background
With the high-speed development and extension of the smart power grid, electric power enterprises widely adopt the power utilization information acquisition system to collect, analyze and process electric energy data acquired by the smart power meters, and the importance of the power consumption data, particularly the daily electric quantity data, to power consumers is gradually improved. The daily electric quantity data is mainly from an electric energy information acquisition system and is obtained by comprehensive multiplying power in an intelligent electric energy meter (daily frozen data of the electric energy meter at 24 points in the day-daily frozen data of the electric energy meter at 24 points in the previous day), but the daily frozen data of the electric energy meter can not be accurately acquired at 24 points often due to the reasons of power grid fluctuation, collapse of the electric energy information acquisition system, meter faults, blockage of a communication channel and the like, so that the problem of daily frozen data loss in a master station system is caused.
At present, a supplementing method adopted by a power enterprise for missing daily frozen data of an electric energy meter is to adopt a master station to manually and repeatedly supplement and recall and manually supplement and copy on site, the method is time-consuming and labor-consuming, and the two methods are difficult to realize especially for a closed place with weak acquired signals; in data processing, the data of the electric energy meter is supplemented by a simple deletion or mean value method at present, and then the electric energy meter data is used as data of relevant statistical analysis, but the result deviation is large, and the reliability of research results is not high. Through technical search, the method for complementing the missing daily freezing data of the electric energy meter has less research at home and abroad, and no corresponding specific solution exists at present.
Disclosure of Invention
The invention aims to overcome the defects, provides a method for complementing the missing daily frozen data of the electric energy meter, is used for solving the problem of missing daily frozen data of the electric energy meter due to signals, systems, equipment and the like, is applied in the scene of the background technology, can improve the data quality of daily electric quantity by restoring the missing value of the daily frozen data of the electric energy meter, truly reflects the real situation of daily electricity consumption of a user, and provides a true and reliable data basis for the research on the electricity consumption behavior of the user.
The invention is realized by adopting the following technical scheme:
a method for complementing missing daily freezing data of an electric energy meter comprises the following steps,
s1, acquiring user information, electric energy meter information, daily electric quantity of historical dates and relevant characteristic factors of the electric energy;
s2, serially connecting the information data obtained in the step S1 and the electricity utilization related characteristic factors into vectors to form an electric energy data time series characteristic diagram;
s3, performing the following steps on the existing electric energy meter data time sequence characteristic diagram data according to the ratio of 4: 1: the proportion of 1 is divided into a training set, a verification set and a test set;
s4, transmitting the electric energy data time series characteristic diagram generated in the step S3 to an input layer of the LSTM network;
s5, training and modeling the LSTM network by using the training set data in the step S3 to obtain an LSTM model; inputting the verification set into an LSTM model, and determining the optimal parameters of the model according to the verification error; testing the LSTM model;
s6, restoring the data of the daily freezing data missing date through the tested LSTM model;
and S7, outputting the date freezing data reduction result of the missing date.
The electricity consumption related characteristic factors described in step S1 include weather data and date factors.
The factors affecting the missing daily freezing data of the electric energy meter are shown in table 1 below.
As shown in table 1 above, the following effects need to be considered in the present invention:
(1) the impact of the contract capacity of the user, the single-phase/three-phase power supply mode and the type of power consumption on the daily power consumption (daily freezing data) is considered.
(2) The influence of the rated working current of the electric energy meter on the daily electricity consumption (daily freezing data) is considered.
(3) Considering the characteristics of the recent electricity consumption of the user, taking the daily electricity consumption of n days before and after as an influence factor; and taking the annual load change trend into consideration, and taking the daily electricity consumption on the same date of the previous year as an influence factor.
(4) Meteorological factors have a crucial influence on daily electricity quantity prediction, and electricity quantity (daily freezing data) used in hot summer and cold winter is changed rapidly; therefore, temperature is used as the influencing factor.
(5) The date type is another important influence factor of the daily electricity consumption (daily freezing data), and the daily electricity consumption of non-working days (holidays, saturdays and sundays) is obviously reduced compared with the working days (monday to friday); date information, i.e. date type, is thus listed as one of the influencing factors.
The electric energy data time series characteristic diagram described in step S2 is arranged mainly according to the sequence of dates corresponding to the electric energy data.
The LSTM (Long-Short Term Memory) model described in step S4 is an effective nonlinear recurrent neural network, and can give consideration to both the time-sequence and nonlinear relations of input data, and the number of network layers is determined by the number of samples of the daily frozen data of the electric energy meter and the nature of the data.
The specific process of step S5 is:
s5-1, training an LSTM model by adopting day freezing data of a certain day in a training set;
s5-2, determining the maximum allowable error and the maximum allowable sub-error according to the accuracy grade of the electric energy meter, judging whether the training error is larger than the maximum allowable sub-error, and if so, entering the step S5-2-1; otherwise, entering the step 5-2-2;
s5-2-1, performing rolling optimization on the LSTM model according to the maximum allowable sub-error, and continuing training the optimized LSTM model according to the step S5-1 to train data of the next date in the training set;
s5-2-2, inputting the verification set in the step S3 into an LSTM model, and determining the optimal parameters of the model according to the verification error; and then, testing the data of the whole test set by using the current LSTM model, comparing the test error with the maximum allowable error, finishing the training of the LSTM model if the test error is within the maximum allowable error range, and returning to the step S5-1 if the test error is not within the maximum allowable error range.
In step S6, the data is restored by the LSTM model by inputting the data time series feature map of the missing date.
In step S7, the trained LSTM model outputs the reduction result of the daily power consumption, and further calculates to obtain the daily freezing data of the missing date, and completes the missing of the daily freezing data of the electric energy meter. If there is more than one day missing, step S6 is repeated in sequence until all missing values are completed.
The invention has the beneficial effects that: aiming at the problem of frozen data loss of the electric energy meter day caused by signals, systems, equipment and the like, under the condition that the data cannot be acquired, the frozen data loss value of the electric energy meter day is complemented for filling the data loss, improving the data quality of a power supply enterprise, ensuring the daily use electricity quantity data of a user, and contributing to improving the synchronous line loss index of the power supply enterprise; meanwhile, the influence of historical data, meteorological data and date information data on daily electricity consumption (daily freezing data) of a user is fully considered, the influence is used as input data, data completion is carried out through an LSTM neural network model, the accuracy rate of missing data restoration is improved, and the missing completion accuracy is high.
Drawings
The invention will be further described with reference to the accompanying drawings:
FIG. 1 is a time series characteristic diagram of daily frozen data of an electric energy meter in the present invention;
FIG. 2 is a logical block diagram of an LSTM network element;
FIG. 3 is a flow chart of an algorithm based on an LSTM neural network;
FIG. 4 is a flow chart of electric energy meter missing daily frozen data completion.
Detailed Description
The technical solution of the present invention is described in detail below with reference to the accompanying drawings and specific embodiments.
Referring to fig. 1, the electric energy data time series characteristic diagram used in the present invention takes the date as the main axis, and is arranged in sequence according to the increasing order of the date, and each date correspondingly records the user information, the electric energy meter information, the daily electric quantity of the historical date and the related characteristic factors of the electric energy consumption, wherein: t is a specific date; t +1 represents the day after T, and so on, and T + n represents the day n after T.
The LSTM (Long-Short Term Memory) model is an effective nonlinear recurrent neural network, and can take account of the time sequence and nonlinear relationship of input data, the number of network layers is determined by the number of samples of the daily frozen data of the electric energy meter and the nature of the data, and the network unit is shown in fig. 2.
In the following variablestTo represent t The time of day, i.e. the current time of day,t-1 represents the last moment.
The basic units of the LSTM network (long-short memory network) comprise a forgetting gate, an input gate and an output gate,x tin order to model the input values at the current time,h tfor intermediate transmission of model at present timeThe value of the obtained value is obtained,S tin the intermediate state at the present moment of time,O tfor the output value of the current time model, the input value of the current time model of the forgetting gatex tAnd intermediate state of last momentS t-1And the intermediate output value of the last time modelh t-1To jointly determine the forgetting part of the state memory unit; wherein, the current time model input value of the forgetting gatex tControl the forgetting of the LSTM network layer to the information and input the intermediate state of the gate at the last momentS t-1Controlling information update of LSTM network layer, outputting intermediate output value of model at last moment of gateh t-1Control information output, input of model input value at present time in gatex t(As can be seen from the network element diagram of FIG. 2, the input values of the current time models of the forgetting gate, the input gate and the output gate are allx tThis can be seen from the formula) respectively through sigmoid function (S-shaped function, also called S-shaped growth curve) and tanh function (hyperbolic tangent function), to jointly determine the retention vector in the state memory unit, and the intermediate output value of the current time modelh tFrom the updated current time intermediate stateS tOutput value of model at current momentO tJointly, the calculation formula is as follows,
f t＝σ(W fx x t+W fh h t-1 +b f )
i t＝σ(W ix x t+W ih h t-1 +b i )
g t＝φ(W gx x t+W gh h t-1 +b g)
o t＝σ(W ox x t+W oh h t-1 +b o )
S t＝g t⊙i t + S t-1⊙f t
h t＝φ(S t)⊙o t
wherein,f t，i t，g t，o t，h tandS trespectively representing the current state of a forgetting gate, the current state of an input node, the current state of an output gate, a model intermediate output value and an intermediate state;W fxinputting values for forgetting gate and current time modelx tA multiplied matrix weight;W fhintermediate output values for forgetting gate and last-time modelh t-1A multiplied matrix weight;W ixinputting the value of the input gate and the current time modelx tA multiplied matrix weight;W ihfor the input gate and intermediate output value at last momenth t-1A multiplied matrix weight;W gxfor the input node and the model input value at the current momentx tA multiplied matrix weight;W ghintermediate output values for input node and last timeh t-1A multiplied matrix weight;W oxfor the output gate and the model input value at the current momentx tA multiplied matrix weight;W ohfor output gate and intermediate output value of last momenth t-1A multiplied matrix weight;b fin order to forget the biased term of the door,b iin order to input the offset term of the gate,b gas a bias term for the input node,b ois the bias term of the output gate; an element in a vector is multiplied by a bit; σ denotes sigmoid function variation, and φ denotes tanh function variation.
The algorithm flow diagram for the LSTM network is shown in fig. 3.
The algorithm of the LSTM network includes the following steps:
(1) acquiring historical data, and carrying out normalization pretreatment on the data;
(2) dividing existing data into a training set, a verification set and a test set according to a proportion;
(3) passing the data to an input layer of the LSTM network;
(4) iterative training of LSTM networks using training set data using time variablest =t+1, circularly inputting data of the next time point until the network converges to obtain an LSTM model;
(5) the validation set data is used to determine the optimal parameters of the model;
(6) and (4) testing the LSTM model, calculating the error (test set error) between the output result of the model and the correct result, finishing the training of the LSTM model when the test set error is optimal, and otherwise, entering the step (4) again.
Fig. 4 is a flowchart of completing missing daily frozen data of an electric energy meter, and a method for completing missing daily frozen data of an electric energy meter includes the following steps:
s1, obtaining electric energy meter information and characteristic factors thereof, wherein the electric energy meter information comprises user information, the electric energy meter information and daily electric quantity of historical dates, and the characteristic factors are related to electricity utilization;
s2, serially connecting the information data obtained in the step S1 and the electricity utilization related characteristic factors into vectors to form an electric energy data time series characteristic diagram, as shown in figure 1;
s3, performing the following steps on the existing electric energy meter data time sequence characteristic diagram data according to the ratio of 4: 1: the proportion of 1 is divided into a training set, a verification set and a test set;
s4, transmitting the electric energy data time series characteristic diagram generated in the step S3 to an input layer of the LSTM network;
s5, training and modeling the LSTM network by using the training set data in the step S3 to obtain an LSTM model; inputting the verification set into an LSTM model, and determining the optimal parameters of the model according to the verification error; testing the LSTM model;
s5-1, training the LSTM model by adopting the electric energy meter data time sequence characteristic diagram of a certain day in the training set, and setting the date variable ast，Order totIs set to an initial value of 1,tthe date training error ise iMaximum allowed sub-error ofε i；
S5-2, maximum allowable errorεMax, ofAllowed sub-errorε iCalculating the accuracy grade of the electric energy meter to obtain and judge the training errore iWhether greater than the maximum allowed sub-errorε iIf yes, go to step S5-2-1; otherwise, entering the step 5-2-2;
s5-2-1, performing rolling optimization on the LSTM model according to the maximum allowable sub-error, and continuing to train the optimized LSTM model according to the step S5-1 for the next date in the training settData of + 1;
s5-2-2, inputting the verification set in the step S3 into an LSTM model, and determining the optimal parameters of the model according to the verification error; the current LSTM model is then used to test the entire test set of data for test errorseThe error of the test is calculated by subtracting the output result of the model training and the correct result and then calculating the squareeAnd maximum allowable errorεBy comparison, if the test error iseAt maximum allowable errorεIf the range is within, the training of the whole LSTM model is completed, otherwise, the step S5-1 is returned,t1 is taken.
S6, restoring the data of the daily freezing data missing date through the tested LSTM model;
and S7, outputting the date freezing data reduction result of the missing date.
Specific example 1:
when the system is used for checking the electric energy meters in abnormal states, a fault of one electric energy meter is found, the frozen data of the previous day cannot be called and measured, the daily electric quantity condition of a user cannot be inquired, the electric energy meter is found to be arranged in a corridor through field inspection, working personnel of a power supply company cannot replace the electric energy meter in time due to the fact that a corridor iron door is locked, and the frozen data of the previous day due to the fault of an old meter is found after one-day replacement is delayed. The method for complementing the missing daily freezing data of the electric energy meter can complement the missing data and accurately and efficiently restore the daily electric quantity information of the user.
Specific example 2:
the data of the day freezing bottom code of the electric energy meter with a certain accuracy grade of 1.0 and a comprehensive multiplying power of 1 is lost due to the chip failure, as shown in table 1,
| date | …… | 1 month and 31 days | 2 month and 1 day | 2 months and 2 days | 2 month and 3 days | 2 month and 4 days |
| Day freezing bottom code | …… | 13344 | 13380 | 13406 | Absence of | 13455 |
The method for complementing the missing daily frozen data of the electric energy meter can complement the missing data, and comprises the following specific steps:
1) acquiring user information, electric energy meter information, daily freezing data of historical dates and electricity utilization related characteristic factors, wherein the characteristic factors are shown in table 2:
table 2 characteristic diagram representation of related information of electric energy meter
2) The factors in the table 2 are connected in series to form a vector to form an electric energy data time series characteristic diagram;
3) and (3) the existing electric energy meter data time sequence characteristic diagram data is processed according to the following steps of 4: 1: the proportion of 1 is divided into a training set (20 days), a verification set (5 days) and a test set (5 days);
4) transmitting the generated electric energy data time series characteristic diagram to an input layer of the LSTM network;
5) training and modeling the LSTM network by using training set data, wherein the maximum allowable error of the electric energy meter with the accuracy grade of 1.0 is 1% (specified by the state), the maximum allowable sub-error is 1%/20=0.05%, and when the network converges and the training error is greater than the sub-error, the LSTM model is obtained;
6) inputting the verification set into an LSTM model, dynamically adjusting model parameters according to verification errors, and finally determining the optimal parameters of the model;
7) testing the LSTM model, comparing the error of the test output result with the maximum allowable error, finishing the training if the test error is smaller than the maximum allowable error range, otherwise, entering the training step 5 again);
8) restoring the data of the daily frozen data missing date through the tested LSTM model to obtain daily electricity consumption 24 kWh of the missing date;
9) and calculating missing daily freezing data, wherein the missing daily freezing data = 24-point previous day electric energy meter daily freezing data + daily electricity consumption/comprehensive multiplying power =13406+24/1=13430, and outputting a recovery result of the missing daily freezing data, and the recovery result is 13430.
The method thoroughly solves the problem of the loss of the daily frozen data of the electric energy meter caused by signals, systems, equipment and the like; the electric energy meter with the data freezing function can be complemented with missing data of the electric energy meter with the data freezing phenomenon of missing a certain day or a certain number of days, the completeness and credibility of electric quantity data of an electric power user are ensured, and the electric energy meter belongs to the technical field of electric power marketing electric quantity monitoring and analysis. Aiming at the problem of frozen data loss of the electric energy meter day caused by signals, systems, equipment and the like, the invention provides a method for complementing the frozen data of the electric energy meter day, which is used for complementing the data loss, improving the data quality of a power supply enterprise and ensuring the daily use electric quantity data of a user, so that the legal benefits of a power generation and supply party are better guaranteed, and meanwhile, the method is helpful for improving the contemporaneous line loss index of the power supply enterprise.
Claims (7)
1. A method for complementing missing daily freezing data of an electric energy meter is characterized by comprising the following steps:
s1, acquiring user information, electric energy meter information, daily electric quantity of historical dates and relevant characteristic factors of the electric energy;
s2, serially connecting the information data obtained in the step S1 and the electricity utilization related characteristic factors into vectors to form an electric energy data time series characteristic diagram;
s3, performing the following steps on the existing electric energy meter data time sequence characteristic diagram data according to the ratio of 4: 1: the proportion of 1 is divided into a training set, a verification set and a test set;
s4, transmitting the electric energy data time series characteristic diagram generated in the step S3 to an input layer of the LSTM network;
s5, training and modeling the LSTM network by using the training set data in the step S3 to obtain an LSTM model; inputting the verification set into an LSTM model, and determining the optimal parameters of the model according to the verification error; testing the LSTM model;
s6, restoring the data of the daily freezing data missing date through the tested LSTM model;
and S7, outputting the date freezing data reduction result of the missing date.
2. The method for complementing missing daily freezing data of an electric energy meter according to claim 1, wherein the electricity consumption related characteristic factors in the step S1 include meteorological data and date factors.
3. The method for complementing missing daily frozen data of an electric energy meter according to claim 1, wherein the time series characteristic diagram of the electric energy data in the step S2 is arranged according to a sequence of dates corresponding to the electric energy data.
4. The method for complementing missing daily frozen data of an electric energy meter according to claim 1, wherein the LSTM model in step S4 is a nonlinear recurrent neural network that can take account of both time-sequence and nonlinear relationships of input data, and the number of network layers is determined by the number of samples of daily frozen data of the electric energy meter and the nature of the data.
5. The method for complementing missing daily frozen data of an electric energy meter according to claim 1, wherein the specific process of the step S5 is as follows:
s5-1, training an LSTM model by adopting day freezing data of a certain day in a training set;
s5-2, determining the maximum allowable error and the maximum allowable sub-error according to the accuracy grade of the electric energy meter, judging whether the training error is larger than the maximum allowable sub-error, and if so, entering the step S5-2-1; otherwise, entering the step 5-2-2;
s5-2-1, performing rolling optimization on the LSTM model according to the maximum allowable sub-error, and continuing training the optimized LSTM model according to the step S5-1 to train data of the next date in the training set;
s5-2-2, inputting the verification set in the step S3 into an LSTM model, and determining the optimal parameters of the model according to the verification error; and then, testing the data of the whole test set by using the current LSTM model, comparing the test error with the maximum allowable error, finishing the training of the LSTM model if the test error is within the maximum allowable error range, and returning to the step S5-1 if the test error is not within the maximum allowable error range.
6. The method for complementing missing daily frozen data of an electric energy meter according to claim 1, wherein in step S6, the data is restored by using an LSTM model by inputting a data time series characteristic diagram of the missing date.
7. The method for complementing missing daily frozen data of an electric energy meter according to claim 1, wherein in step S7, the trained LSTM model outputs a reduction result of daily electricity consumption, daily frozen data of the missing date is further calculated, and the missing daily frozen data of the electric energy meter is complemented; if there is more than one day missing, step S6 is repeated in sequence until all missing values are completed.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202110539537.5A CN113239029A (en) | 2021-05-18 | 2021-05-18 | Completion method for missing daily freezing data of electric energy meter |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202110539537.5A CN113239029A (en) | 2021-05-18 | 2021-05-18 | Completion method for missing daily freezing data of electric energy meter |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN113239029A true CN113239029A (en) | 2021-08-10 |
Family
ID=77134947
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202110539537.5A Pending CN113239029A (en) | 2021-05-18 | 2021-05-18 | Completion method for missing daily freezing data of electric energy meter |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN113239029A (en) |
Cited By (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN113608163A (en) * | 2021-09-10 | 2021-11-05 | 天目数据(福建)科技有限公司 | Ammeter fault diagnosis method and device of stacked cyclic neural network |
Citations (17)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN104598986A (en) * | 2014-12-12 | 2015-05-06 | 国家电网公司 | Big data based power load prediction method |
| CN105260803A (en) * | 2015-11-06 | 2016-01-20 | 国家电网公司 | Power consumption prediction method for system |
| CN109255505A (en) * | 2018-11-20 | 2019-01-22 | 国网辽宁省电力有限公司经济技术研究院 | A kind of short-term load forecasting method of multi-model fused neural network |
| CN109614997A (en) * | 2018-11-29 | 2019-04-12 | 武汉大学 | A kind of stealing Risk Forecast Method and device based on deep learning |
| CN110070229A (en) * | 2019-04-26 | 2019-07-30 | 中国计量大学 | The short term prediction method of home electrical load |
| CN110363360A (en) * | 2019-07-24 | 2019-10-22 | 广东工业大学 | A kind of short-term wind power forecast method, device and equipment |
| CN110767279A (en) * | 2019-10-21 | 2020-02-07 | 山东师范大学 | Electronic health record missing data completion method and system based on LSTM |
| CN110837888A (en) * | 2019-11-13 | 2020-02-25 | 大连理工大学 | Traffic missing data completion method based on bidirectional cyclic neural network |
| CN111178611A (en) * | 2019-12-23 | 2020-05-19 | 广西电网有限责任公司 | Method for predicting daily electric quantity |
| CN111612651A (en) * | 2020-05-27 | 2020-09-01 | 福州大学 | Abnormal electric quantity data detection method based on long-term and short-term memory network |
| CN111783953A (en) * | 2020-06-30 | 2020-10-16 | 重庆大学 | 24-point power load value 7-day prediction method based on optimized LSTM network |
| CN111784057A (en) * | 2020-07-04 | 2020-10-16 | 湘潭大学 | Short-term power load prediction combination method based on snapshot feedback mechanism |
| CN111861039A (en) * | 2020-07-31 | 2020-10-30 | 广东电网有限责任公司广州供电局 | Power load prediction method, system, equipment and storage medium based on LSTM and generalized predictive control algorithm |
| CN112232561A (en) * | 2020-10-13 | 2021-01-15 | 三峡大学 | Power load probability prediction method based on constrained parallel LSTM quantile regression |
| CN112418496A (en) * | 2020-11-10 | 2021-02-26 | 国网四川省电力公司经济技术研究院 | Power distribution station energy storage configuration method based on deep learning |
| CN112633604A (en) * | 2021-01-04 | 2021-04-09 | 重庆邮电大学 | Short-term power consumption prediction method based on I-LSTM |
| CN112651543A (en) * | 2020-11-10 | 2021-04-13 | 沈阳工程学院 | Daily electric quantity prediction method based on VMD decomposition and LSTM network |
-
2021
- 2021-05-18 CN CN202110539537.5A patent/CN113239029A/en active Pending
Patent Citations (17)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN104598986A (en) * | 2014-12-12 | 2015-05-06 | 国家电网公司 | Big data based power load prediction method |
| CN105260803A (en) * | 2015-11-06 | 2016-01-20 | 国家电网公司 | Power consumption prediction method for system |
| CN109255505A (en) * | 2018-11-20 | 2019-01-22 | 国网辽宁省电力有限公司经济技术研究院 | A kind of short-term load forecasting method of multi-model fused neural network |
| CN109614997A (en) * | 2018-11-29 | 2019-04-12 | 武汉大学 | A kind of stealing Risk Forecast Method and device based on deep learning |
| CN110070229A (en) * | 2019-04-26 | 2019-07-30 | 中国计量大学 | The short term prediction method of home electrical load |
| CN110363360A (en) * | 2019-07-24 | 2019-10-22 | 广东工业大学 | A kind of short-term wind power forecast method, device and equipment |
| CN110767279A (en) * | 2019-10-21 | 2020-02-07 | 山东师范大学 | Electronic health record missing data completion method and system based on LSTM |
| CN110837888A (en) * | 2019-11-13 | 2020-02-25 | 大连理工大学 | Traffic missing data completion method based on bidirectional cyclic neural network |
| CN111178611A (en) * | 2019-12-23 | 2020-05-19 | 广西电网有限责任公司 | Method for predicting daily electric quantity |
| CN111612651A (en) * | 2020-05-27 | 2020-09-01 | 福州大学 | Abnormal electric quantity data detection method based on long-term and short-term memory network |
| CN111783953A (en) * | 2020-06-30 | 2020-10-16 | 重庆大学 | 24-point power load value 7-day prediction method based on optimized LSTM network |
| CN111784057A (en) * | 2020-07-04 | 2020-10-16 | 湘潭大学 | Short-term power load prediction combination method based on snapshot feedback mechanism |
| CN111861039A (en) * | 2020-07-31 | 2020-10-30 | 广东电网有限责任公司广州供电局 | Power load prediction method, system, equipment and storage medium based on LSTM and generalized predictive control algorithm |
| CN112232561A (en) * | 2020-10-13 | 2021-01-15 | 三峡大学 | Power load probability prediction method based on constrained parallel LSTM quantile regression |
| CN112418496A (en) * | 2020-11-10 | 2021-02-26 | 国网四川省电力公司经济技术研究院 | Power distribution station energy storage configuration method based on deep learning |
| CN112651543A (en) * | 2020-11-10 | 2021-04-13 | 沈阳工程学院 | Daily electric quantity prediction method based on VMD decomposition and LSTM network |
| CN112633604A (en) * | 2021-01-04 | 2021-04-09 | 重庆邮电大学 | Short-term power consumption prediction method based on I-LSTM |
Non-Patent Citations (2)
| Title |
|---|
| 辜超 等: "LSTM在输变电设备缺失值填补中的应用", 《电测与仪表》 * |
| 陆继翔 等: "基于CNN_LSTM混合神经网络模型的短期负荷预测方法", 《电子系统自动化》 * |
Cited By (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN113608163A (en) * | 2021-09-10 | 2021-11-05 | 天目数据(福建)科技有限公司 | Ammeter fault diagnosis method and device of stacked cyclic neural network |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN110659273B (en) | Data abnormity monitoring and repairing method of distributed big data acquisition platform | |
| CN110263866A (en) | A kind of power consumer load setting prediction technique based on deep learning | |
| CN110222882A (en) | A kind of prediction technique and device of electric system Mid-long Term Load | |
| CN114638425B (en) | Historical data-based large-user monthly electricity consumption prediction method and system | |
| CN114140176B (en) | Adjustable capacity prediction method and device for load aggregation platform | |
| Velasco et al. | Next day electric load forecasting using Artificial Neural Networks | |
| Shaghaghi et al. | Proposing a new optimized forecasting model for the failure rate of power distribution network thermal equipment for educational centers | |
| CN113239029A (en) | Completion method for missing daily freezing data of electric energy meter | |
| Yuan et al. | Improved particle filter for non-gaussian forecasting-aided state estimation | |
| CN117578440A (en) | Power grid load prediction method based on neural network | |
| Jiang et al. | A bottom-up method for probabilistic short-term load forecasting based on medium voltage load patterns | |
| CN112232570A (en) | Forward active total electric quantity prediction method and device and readable storage medium | |
| Mauricio et al. | Cuckoo search algorithm optimization of Holt-Winter Method for distribution transformer load forecasting | |
| Deng et al. | Medium-term rolling load forecasting based on seasonal decomposition and long short-term memory neural network | |
| Huang et al. | Research on Forecasting Method of Industry Electricity Consumption Considering Price Factor | |
| Dung et al. | Handling missing data using standardized load profile (SLP) and support vector regression (SVR) | |
| Chen et al. | Multifeature Short‐Term Power Load Forecasting Based on GCN‐LSTM | |
| Teferra et al. | Fuzzy‐Swarm Intelligence‐Based Short‐Term Load Forecasting Model as a Solution to Power Quality Issues Existing in Microgrid System | |
| Li et al. | Probabilistic load forecasting of adaptive multiple polynomial regression considering temperature scenario and dummy variables | |
| Sabri et al. | Spatial short-term load forecasting using grey dynamic model specific in tropical area | |
| Bhol et al. | Short‐Term Reactive Power Forecasting Based on Real Power Demand Using Holt‐Winters’ Model Ensemble by Global Flower Pollination Algorithm for Microgrid | |
| Wen | Probabilistic wind power forecasting with missing values via adaptive quantile regression | |
| Farhoumandi et al. | Short-term Forecasting of Non-Conforming Net Load Using a Fusion Model with Machine Learning and Deep Learning Methods | |
| Shaghaghi et al. | Original Research Article Proposing a new optimized forecasting model for the failure rate of power distribution network thermal equipment for educational centers | |
| Yan et al. | Power Flow Prediction: A Case in Ningxia Electricity Market |
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 |
Application publication date: 20210810 |
|
| RJ01 | Rejection of invention patent application after publication |