CN104636822B - A kind of resident load prediction technique based on elman neural networks - Google Patents
A kind of resident load prediction technique based on elman neural networks Download PDFInfo
- Publication number
- CN104636822B CN104636822B CN201510031651.1A CN201510031651A CN104636822B CN 104636822 B CN104636822 B CN 104636822B CN 201510031651 A CN201510031651 A CN 201510031651A CN 104636822 B CN104636822 B CN 104636822B
- Authority
- CN
- China
- Prior art keywords
- data
- resident load
- neural network
- day
- resident
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Active
Links
- 238000013528 artificial neural network Methods 0.000 title claims abstract description 102
- 238000000034 method Methods 0.000 title claims abstract description 25
- 230000001932 seasonal effect Effects 0.000 claims abstract description 36
- 230000000153 supplemental effect Effects 0.000 claims abstract description 28
- 238000012549 training Methods 0.000 claims abstract description 12
- 210000002569 neuron Anatomy 0.000 claims abstract description 11
- 239000013598 vector Substances 0.000 claims description 36
- 230000005540 biological transmission Effects 0.000 claims description 8
- 238000004364 calculation method Methods 0.000 claims description 4
- 210000004205 output neuron Anatomy 0.000 claims description 4
- 240000002853 Nelumbo nucifera Species 0.000 description 7
- 235000006508 Nelumbo nucifera Nutrition 0.000 description 7
- 235000006510 Nelumbo pentapetala Nutrition 0.000 description 7
- 230000005611 electricity Effects 0.000 description 7
- 238000013507 mapping Methods 0.000 description 3
- 230000009286 beneficial effect Effects 0.000 description 2
- 238000010586 diagram Methods 0.000 description 2
- 244000131316 Panax pseudoginseng Species 0.000 description 1
- 235000005035 Panax pseudoginseng ssp. pseudoginseng Nutrition 0.000 description 1
- 235000003140 Panax quinquefolius Nutrition 0.000 description 1
- 230000006978 adaptation Effects 0.000 description 1
- 238000004458 analytical method Methods 0.000 description 1
- 238000013459 approach Methods 0.000 description 1
- 239000012141 concentrate Substances 0.000 description 1
- 230000001186 cumulative effect Effects 0.000 description 1
- 238000011161 development Methods 0.000 description 1
- 239000006185 dispersion Substances 0.000 description 1
- 230000000694 effects Effects 0.000 description 1
- 238000005516 engineering process Methods 0.000 description 1
- 238000013277 forecasting method Methods 0.000 description 1
- 235000008434 ginseng Nutrition 0.000 description 1
- 238000007726 management method Methods 0.000 description 1
- 238000004519 manufacturing process Methods 0.000 description 1
- 238000012986 modification Methods 0.000 description 1
- 230000004048 modification Effects 0.000 description 1
- 238000010606 normalization Methods 0.000 description 1
- 238000012545 processing Methods 0.000 description 1
- 230000000306 recurrent effect Effects 0.000 description 1
- 239000002699 waste material Substances 0.000 description 1
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
- G06Q10/00—Administration; Management
- G06Q10/04—Forecasting or optimisation specially adapted for administrative or management purposes, e.g. linear programming or "cutting stock problem"
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- 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
- Business, Economics & Management (AREA)
- Engineering & Computer Science (AREA)
- Economics (AREA)
- Human Resources & Organizations (AREA)
- Strategic Management (AREA)
- Theoretical Computer Science (AREA)
- Physics & Mathematics (AREA)
- Health & Medical Sciences (AREA)
- Marketing (AREA)
- General Physics & Mathematics (AREA)
- General Business, Economics & Management (AREA)
- Tourism & Hospitality (AREA)
- Public Health (AREA)
- General Health & Medical Sciences (AREA)
- Primary Health Care (AREA)
- Water Supply & Treatment (AREA)
- Development Economics (AREA)
- Game Theory and Decision Science (AREA)
- Entrepreneurship & Innovation (AREA)
- Operations Research (AREA)
- Quality & Reliability (AREA)
- Management, Administration, Business Operations System, And Electronic Commerce (AREA)
Abstract
The invention discloses a kind of resident load prediction techniques based on elman neural networks, including：Obtain the resident load historical data of previous year and corresponding weather history supplemental characteristic；Calculate the seasonal index number of the resident load in each month；Resident load historical data is modified using seasonal index number；It determines the data that output and input of neural network, and determines optimal hidden layer neuron number, to establish the neural network based on elman；Revised resident load historical data and corresponding weather history supplemental characteristic are normalized, and then the neural network of foundation is trained, within a preset range by prediction control errors；Resident load is predicted using the neural network after training.The present invention has the ability for the seasonal fluctuations for adapting to time-varying characteristics and resident load, can directly predict and reflect the dynamic characteristic of resident load, precision of prediction is higher, can be widely applied in the load prediction field of electric system.
Description
Technical field
The present invention relates to Load Prediction In Power Systems technical field, more particularly to a kind of based on elman neural networks
Resident load prediction technique.
Background technology
Electric power all plays very important effect, with state as a kind of important energy source in daily life and work
The fast development of people's economy, Analyzing Total Electricity Consumption and each industry electricity consumption amount also steady growth, therefore the use of electricity consumption becomes
Gesture not only influences the production and management decision-making and economic benefit of enterprises of managing electric wire netting, also affects the trend analysis of social economy.
It is the precondition that electric system is scheduled electric power resource, plans reasonably to carry out load forecast.
Electric load can be generally divided into industrial load, Commercial Load, resident load etc., and wherein industrial load and business is negative
Proportion of the lotus in electric load is higher, and power grid enterprises always compare attention to the load prediction of this block, and has built up successively negative
Lotus control system and power information acquisition system are to complete data acquisition and load prediction to industry and commerce load；Resident is negative
Lotus is due to distribution dispersion, scale feature less than normal, and what is taken always is all the method for concentrating prediction, i.e., with taiwan area or feeder load
Predicted for unit, the shortcomings that this prediction technique is exactly that precision is not high, especially as the increasing year by year of resident's household electrical appliance,
Steady-state growth trend and apparent is presented in the gradually popularization of the universal and electric vehicle of electric bicycle, the power load of resident
Seasonal fluctuation, by concentrate prediction method the disadvantage of resident's load prediction is more shown.
Invention content
In order to solve the above technical problems, the object of the present invention is to provide a kind of residents based on elman neural networks
Load forecasting method.
The technical solution adopted by the present invention to solve the technical problems is：
A kind of resident load prediction technique based on elman neural networks, including：
S1, the resident load historical data for obtaining previous year and corresponding weather history supplemental characteristic, while to this
Effective number of days in year carries out date type division；
S2, the resident load historical data according to acquisition, calculate the same period average of the resident load in each month, in turn
After the overall average for calculating all same period averages, each same period average and overall average are divided by and obtain seasonal index number；
S3, resident load historical data is modified using seasonal index number, by the resident load history number in each month
According to divided by corresponding seasonal index number after, obtain revised resident load historical data；
S4, the data that output and input for determining neural network, and determine optimal hidden layer neuron number, to establish
Neural network based on elman；
S5, place is normalized to revised resident load historical data and corresponding weather history supplemental characteristic
Reason, and then the neural network of foundation is trained according to the data after normalized, by the prediction error control of neural network
System is within a preset range；
S6, the resident load historical data for obtaining prediction the last week day, the weather parameter data and date type for predicting day
As the input of neural network, using the neural network after training to predicting that the resident load of day is predicted, and then will obtain
Prediction data be multiplied by seasonal index number after obtain resident load prediction data；
The date type is divided into day off and working day two types.
Further, preset range described in the step S5 is 5%~10%.
Further, the resident load historical data includes each hour resident load data, the weather history ginseng
Number data include temperature, sunshine-duration and weather pattern.
Further, the step S4, including：
Resident load historical data, weather history supplemental characteristic and the date type that S41, statistics obtain, by any day
Output data of the resident load data as neural network, while by the resident load number of each hour in the last week of this day
According to this and input data as neural network of the weather parameter data of this day and date type；
S42, neural network is initialized, input node unit vector, hidden layer is determined according to input and output sequence
Node unit vector, feedback state vector and output node vector, to set up the neural network based on elman.
Further, the non-linear state space expression of the neural network based on elman is：
Wherein, y (k) indicates that m dimension output node vectors, l (k) indicate that m ties up hidden layer node unit vector, and x (k) indicates u
Dimensional input vector, c (k) indicate that n ties up feedback state vector, w3Indicate hidden layer to the connection weight of output layer, w2Indicate input layer
To the connection weight of hidden layer, w1Indicate to accept layer to the connection weight of hidden layer, the transmission letter of g (*) expression output neurons
Number, f (*) indicate the transmission function of hidden layer neuron.
Further, the step S5, including：
S51, revised resident load historical data and corresponding weather history supplemental characteristic are carried out according to the following formula
Normalized：
Wherein, xkIndicate k-th of parameter value in resident load historical data sequence or weather history supplemental characteristic ordered series of numbers,
K is natural number, xmaxIndicate xkMaximum value in the data sequence of place, xminIndicate xkMinimum value in the data sequence of place；
S52, error calculation, right value update and threshold values are carried out to the neural network of foundation according to the data after normalized
Update, and then within a preset range by the prediction control errors of the neural network based on elman.
Further, the neural network based on elman carries out weight value revision update using BP algorithm, and flat using error
Side and function carry out target function study, and the formula of the target function study is：
E (x) indicates target function in above formula,Indicate target input vector.
Further, the step S6, including：
S61, the resident load historical data for obtaining prediction the last week day, the weather parameter data and date type for predicting day
As the input of neural network, using the neural network after training to predicting that the resident load of day is predicted, and then obtain pre-
Prediction data hourly on the day of surveying day；
S62, after the prediction data of acquisition is multiplied by seasonal index number, resident load prediction data hourly is obtained.
Further, further comprising the steps of after the step S62：
After S63, the actual load data on the day of acquisition prediction day, the resident load prediction data and actual negative of acquisition are calculated
Error amount between lotus data, and by error value back to neural network.
The beneficial effects of the invention are as follows：A kind of resident load prediction technique based on elman neural networks of the present invention, packet
It includes：The resident load historical data of previous year and corresponding weather history supplemental characteristic are obtained, while in the year
Effective number of days carries out date type division；According to the resident load historical data of acquisition, the resident load in each month is calculated
Same period average, and then after the overall average of all same period averages of calculating, each same period average is divided by with overall average
Obtain seasonal index number；Resident load historical data is modified using seasonal index number, by the resident load history in each month
After data divided by corresponding seasonal index number, revised resident load historical data is obtained；Determine the input of neural network and defeated
Go out data, and determine optimal hidden layer neuron number, to establish the neural network based on elman；To revised residence
Burden on the people lotus historical data and corresponding weather history supplemental characteristic are normalized, so according to normalized after
Data are trained the neural network of foundation, within a preset range by the prediction control errors of neural network；Obtain prediction day
The resident load historical data of the last week predicts the input of the weather parameter data and date type of day as neural network, adopts
With the neural network after training to predicting that the resident load of day is predicted, and then the prediction data of acquisition is multiplied by seasonal index number
After obtain resident load prediction data.This method is by the structures of the Elman neural networks based on seasonal index number, in conjunction with relatively
The resident load data for obtaining prediction day can be predicted in area's resident load historical data and corresponding weather history supplemental characteristic,
And arbitrary nonlinear mapping can be approached with arbitrary precision, it is not intended that the influence of external noise, has higher precision,
And with the ability for the seasonal fluctuations for adapting to time-varying characteristics and resident load, it can directly predict and reflect the dynamic of resident load
Characteristic, precision of prediction are higher.And this method increases seasonal index number feature, can overcome the seasonal fluctuations of resident load compared with
Big and historical data utilizes the problems such as incomplete, can effectively improve the precision and prediction stability of prediction data.
Description of the drawings
The invention will be further described with reference to the accompanying drawings and examples.
Fig. 1 is a kind of flow diagram of resident load prediction technique based on elman neural networks of the present invention；
Fig. 2 is the neural network that a kind of resident load prediction technique based on elman neural networks of the present invention is established
Structural schematic diagram.
Specific implementation mode
For the ease of following description, following explanation of nouns is provided first：
Elman networks：J.L.Elman puts forward in nineteen ninety first against speech processing problems, it is a kind of allusion quotation
The local regression network (global feed for ward local recurrent) of type.
BP algorithm：Error Back Propagation Algorithm, error backpropagation algorithm, abbreviation BP algorithm.
Referring to Fig.1, the present invention provides a kind of resident load prediction techniques based on elman neural networks, including：
S1, the resident load historical data for obtaining previous year and corresponding weather history supplemental characteristic, while to this
Effective number of days in year carries out date type division；
S2, the resident load historical data according to acquisition, calculate the same period average of the resident load in each month, in turn
After the overall average for calculating all same period averages, each same period average and overall average are divided by and obtain seasonal index number；
S3, resident load historical data is modified using seasonal index number, by the resident load history number in each month
According to divided by corresponding seasonal index number after, obtain revised resident load historical data；
S4, the data that output and input for determining neural network, and determine optimal hidden layer neuron number, to establish
Neural network based on elman；
S5, place is normalized to revised resident load historical data and corresponding weather history supplemental characteristic
Reason, and then the neural network of foundation is trained according to the data after normalized, by the prediction error control of neural network
System is within a preset range；
S6, the resident load historical data for obtaining prediction the last week day, the weather parameter data and date type for predicting day
As the input of neural network, using the neural network after training to predicting that the resident load of day is predicted, and then will obtain
Prediction data be multiplied by seasonal index number after obtain resident load prediction data；
The date type is divided into day off and working day two types.
It is further used as preferred embodiment, preset range described in the step S5 is 5%~10%.
It is further used as preferred embodiment, the resident load historical data includes each hour resident load number
According to the weather history supplemental characteristic includes temperature, sunshine-duration and weather pattern.
It is further used as preferred embodiment, the step S4, including：
Resident load historical data, weather history supplemental characteristic and the date type that S41, statistics obtain, by any day
Output data of the resident load data as neural network, while by the resident load number of each hour in the last week of this day
According to this and input data as neural network of the weather parameter data of this day and date type；
S42, neural network is initialized, input node unit vector, hidden layer is determined according to input and output sequence
Node unit vector, feedback state vector and output node vector, to set up the neural network based on elman.
It is further used as preferred embodiment, the non-linear state space expression of the neural network based on elman
Formula is：
Wherein, y (k) indicates that m dimension output node vectors, l (k) indicate that m ties up hidden layer node unit vector, and x (k) indicates u
Dimensional input vector, c (k) indicate that n ties up feedback state vector, w3Indicate hidden layer to the connection weight of output layer, w2Indicate input layer
To the connection weight of hidden layer, w1Indicate to accept layer to the connection weight of hidden layer, the transmission letter of g (*) expression output neurons
Number, f (*) indicate the transmission function of hidden layer neuron.
It is further used as preferred embodiment, the step S5, including：
S51, revised resident load historical data and corresponding weather history supplemental characteristic are carried out according to the following formula
Normalized：
Wherein, xkIndicate k-th of parameter value in resident load historical data sequence or weather history supplemental characteristic ordered series of numbers,
K is natural number, xmaxIndicate xkMaximum value in the data sequence of place, xminIndicate xkMinimum value in the data sequence of place；
S52, error calculation, right value update and threshold values are carried out to the neural network of foundation according to the data after normalized
Update, and then within a preset range by the prediction control errors of the neural network based on elman.
It is further used as preferred embodiment, the neural network based on elman carries out weights using BP algorithm and repaiies
Positive update, and target function study is carried out using sum of squared errors function, the formula of the target function study is：
E (x) indicates target function in above formula,Indicate target input vector.
It is further used as preferred embodiment, the step S6, including：
S61, the resident load historical data for obtaining prediction the last week day, the weather parameter data and date type for predicting day
As the input of neural network, using the neural network after training to predicting that the resident load of day is predicted, and then obtain pre-
Prediction data hourly on the day of surveying day；
S62, after the prediction data of acquisition is multiplied by seasonal index number, resident load prediction data hourly is obtained.
It is further used as preferred embodiment, it is further comprising the steps of after the step S62：
After S63, the actual load data on the day of acquisition prediction day, the resident load prediction data and actual negative of acquisition are calculated
Error amount between lotus data, and by error value back to neural network.
With reference to specific embodiment, the present invention will be further described.
Referring to Fig.1, a kind of resident load prediction technique based on elman neural networks, including：
S1, the resident load historical data for obtaining previous year and corresponding weather history supplemental characteristic, while to this
Effective number of days in year carries out date type division, and date type is divided into two type of day off and working day by the present embodiment
Type；
S2, the resident load historical data according to acquisition, calculate the same period average of the resident load in each month, in turn
After the overall average for calculating all same period averages, each same period average and overall average are divided by and obtain seasonal index number；
S3, resident load historical data is modified using seasonal index number, by the resident load history number in each month
According to divided by corresponding seasonal index number after, obtain revised resident load historical data；
S4, the data that output and input for determining neural network, and determine optimal hidden layer neuron number, to establish
Neural network based on elman；Neural network based on elman includes input layer, hidden layer, accepts layer and output layer, is accepted
Layer is used to remember the output valve of hidden layer previous moment and the output valve is returned to the input of hidden layer, increases feedback, right
Historical data is more sensitive, relatively stable；
S5, place is normalized to revised resident load historical data and corresponding weather history supplemental characteristic
Reason, and then the neural network of foundation is trained according to the data after normalized, by the prediction error control of neural network
System is within a preset range；In the present embodiment, preset range is 5%~10%；Control errors are carried out, can be approached with arbitrary accuracy
Arbitrary nonlinear mapping；
S6, the resident load historical data for obtaining prediction the last week day, the weather parameter data and date type for predicting day
As the input of neural network, using the neural network after training to predicting that the resident load of day is predicted, and then will obtain
Prediction data be multiplied by seasonal index number after obtain resident load prediction data；
Resident load historical data includes each hour resident load data, weather history supplemental characteristic include temperature,
Sunshine-duration and weather pattern.
Specifically, step S4 includes step S41~S42：
Resident load historical data, weather history supplemental characteristic and the date type that S41, statistics obtain, by any day
Output data of the resident load data as neural network, while by the resident load number of each hour in the last week of this day
According to this and input data as neural network of the weather parameter data of this day and date type；Resident load number in input data
According to total 168 load points, the resident load data of output data amount to 24 load points；
S42, neural network is initialized, determines that u ties up input node unit vector according to input and output sequence (X, Y)
X, n ties up hidden layer node unit vector l, n dimension feedback state vector c and m and ties up output node vector y, is based on to set up
The neural network of elman, the neural network training model that the present embodiment is established are as shown in Figure 2.Wherein, X1, X2Xu are
The node of input layer, the weather parameter data of the prediction day of corresponding input, resident's load and date class after correcting within upper one week
Type；Y1 is the node of output layer, the prediction day system resident load of corresponding output；L1, l2lN are hidden layers
Node, wherein node in hidden layer n (i.e. optimal hidden layer neuron number) by gradually increasing cut-and-try, i.e., according to by
The method that cumulative additional examination is visited determines；C1, C2CN are the nodes for accepting layer, for remembering implicit layer unit previous moment
Output valve and return to the input of hidden layer.
In the present embodiment, the non-linear state space expression of the neural network based on elman is：
Wherein, y (k) indicates that m dimension output node vectors, l (k) indicate that m ties up hidden layer node unit vector, and x (k) indicates u
Dimensional input vector, c (k) indicate that n ties up feedback state vector, w3Indicate hidden layer to the connection weight of output layer, w2Indicate input layer
To the connection weight of hidden layer, w1Indicate to accept layer to the connection weight of hidden layer, the transmission letter of g (*) expression output neurons
Number, f (*) indicate that the transmission function of hidden layer neuron, f (*) generally use S function.
Specifically, step S5 includes step S51~S52：
S51, using minimax method, according to the following formula to revised resident load historical data and corresponding history day
Gas supplemental characteristic is normalized：
Wherein, xkIndicate k-th of parameter value in resident load historical data sequence or weather history supplemental characteristic ordered series of numbers,
K is natural number, xmaxIndicate xkMaximum value in the data sequence of place, xminIndicate xkMinimum value in the data sequence of place；
S52, error calculation, right value update and threshold values are carried out to the neural network of foundation according to the data after normalized
Update, and then within a preset range by the prediction control errors of the neural network based on elman.
Due to the particularity of date type, 1 will be labeled as day off here, working day is labeled as 0, also corresponds to maximum most
Small normalization principle is normalized date type and does not interfere with its occurrence.
In step S52, the neural network based on elman carries out weight value revision update using BP algorithm, and flat using error
Side and function carry out target function study, and the formula of target function study is：
E (x) indicates target function in above formula,Indicate target input vector.
Specifically, step S6 includes step S61~S63：
S61, the resident load historical data for obtaining prediction the last week day, the weather parameter data and date type for predicting day
As the input of neural network, using the neural network after training to predicting that the resident load of day is predicted, and then obtain pre-
Prediction data hourly, that is, obtain 24 load points of prediction on the day of surveying day；
S62, after the prediction data of acquisition is multiplied by seasonal index number, resident load prediction data hourly is obtained；
After S63, the actual load data on the day of acquisition prediction day, the resident load prediction data and actual negative of acquisition are calculated
Error amount between lotus data, and by error value back to neural network.After obtaining the actual load data on the day of predicting day,
It is compared with 24 load points in prediction data, calculates separately the error amount of 24 load points and actual load data and anti-
It is fed to neural network, the neural network training model of foundation can be adjusted, make it closer to actual conditions.
The resident load prediction technique of the present invention can be with by setting up the neural network prediction model of resident load
Arbitrary precision approaches arbitrary nonlinear mapping, it is not intended that the influence of external noise, has higher precision, and with adaptation
The ability of the seasonal fluctuations of time-varying characteristics and resident load can directly predict and reflect the dynamic characteristic of resident load, prediction
Precision is higher.And this method increases seasonal index number feature, the seasonal fluctuations of resident load can be overcome larger and history
Data utilize the problems such as incomplete, can effectively improve the precision and prediction stability of prediction data.
In addition, changing the residential electricity consumption historical data to be formed according to residential electricity consumption demand, seasonal climate, predict that it is used in the recent period
Electricity is beneficial to residential households and using electricity wisely and slows down the urgency of electric power resource scarcity consciously.At the same time, lead to
It crosses and predicts to be combined with the utilization of new energy by resident load, can more accomplish such as solar energy, the various forms of new energy such as wind energy
Source makes full use of, and avoids the waste of the energy, and the economic benefit of bigger is created for user.
It is to be illustrated to the preferable implementation of the present invention, but the invention is not limited to embodiment above, it is ripe
Various equivalent variations or replacement can also be made under the premise of without prejudice to spirit of that invention by knowing those skilled in the art, these
Equivalent modification or replacement is all contained in the application claim limited range.
Claims (6)
1. a kind of resident load prediction technique based on elman neural networks, which is characterized in that including：
S1, the resident load historical data for obtaining previous year and corresponding weather history supplemental characteristic, while to the year
In effective number of days carry out date type division；
S2, the resident load historical data according to acquisition calculate the same period average of the resident load in each month, and then calculate
After the overall average of all same period averages, each same period average and overall average are divided by and obtain seasonal index number；
S3, resident load historical data is modified using seasonal index number, the resident load historical data in each month is removed
After corresponding seasonal index number, revised resident load historical data is obtained；
S4, the data that output and input for determining neural network, and determine optimal hidden layer neuron number, it is based on to establish
The neural network of elman；
S5, revised resident load historical data and corresponding weather history supplemental characteristic are normalized, into
And the neural network of foundation is trained according to the data after normalized, by the prediction control errors of neural network pre-
If in range；
S6, the resident load historical data for obtaining prediction the last week day, the weather parameter data for predicting day and date type conduct
The input of neural network, using the neural network after training to predicting that the resident load of day is predicted, and then by the pre- of acquisition
Measured data obtains resident load prediction data after being multiplied by seasonal index number；
The date type is divided into day off and working day two types；
The resident load historical data includes each hour resident load data, and the weather history supplemental characteristic includes gas
Temperature, sunshine-duration and weather pattern；
The step S4, including：
Resident load historical data, weather history supplemental characteristic and the date type that S41, statistics obtain, by any day resident
Output data of the load data as neural network, at the same by the resident load data of each hour in the last week of this day with
And input data of the weather parameter data and date type of this day as neural network；
S42, neural network is initialized, input node unit vector, hidden layer node is determined according to input and output sequence
Unit vector, feedback state vector and output node vector, to set up the neural network based on elman；
The neural network based on elman using BP algorithm carry out weight value revision update, and using sum of squared errors function into
Row index function learning, the formula that the target function learns are：
E (x) indicates target function in above formula,Indicate target input vector.
2. a kind of resident load prediction technique based on elman neural networks according to claim 1, which is characterized in that
Preset range described in the step S5 is 5%~10%.
3. a kind of resident load prediction technique based on elman neural networks according to claim 1, which is characterized in that
The non-linear state space expression of the neural network based on elman is：
Wherein, y (k) indicates that m dimension output node vectors, l (k) indicate that m ties up hidden layer node unit vector, and x (k) indicates that u dimensions are defeated
Incoming vector, c (k) indicate that n ties up feedback state vector, w3Indicate hidden layer to the connection weight of output layer, w2Indicate input layer to hidden
Connection weight containing layer, w1Indicate to accept layer to the connection weight of hidden layer, the transmission function of g (*) expression output neurons, f
(*) indicates the transmission function of hidden layer neuron.
4. a kind of resident load prediction technique based on elman neural networks according to claim 1, which is characterized in that
The step S5, including：
S51, normalizing is carried out to revised resident load historical data and corresponding weather history supplemental characteristic according to the following formula
Change is handled：
Wherein, xkIndicate that k-th of parameter value in resident load historical data sequence or weather history supplemental characteristic ordered series of numbers, k are certainly
So number, xmaxIndicate xkMaximum value in the data sequence of place, xminIndicate xkMinimum value in the data sequence of place；
S52, error calculation, right value update and threshold value are carried out more to the neural network of foundation according to the data after normalized
Newly, and then by the prediction control errors of the neural network based on elman within a preset range.
5. a kind of resident load prediction technique based on elman neural networks according to claim 4, which is characterized in that
The step S6, including：
S61, the resident load historical data for obtaining prediction the last week day, the weather parameter data for predicting day and date type conduct
The input of neural network using the neural network after training to predicting that the resident load of day is predicted, and then obtains prediction day
Same day prediction data hourly；
S62, after the prediction data of acquisition is multiplied by seasonal index number, resident load prediction data hourly is obtained.
6. a kind of resident load prediction technique based on elman neural networks according to claim 5, which is characterized in that
It is further comprising the steps of after the step S62：
After S63, the actual load data on the day of acquisition prediction day, the resident load prediction data and actual load number of acquisition are calculated
Error amount between, and by error value back to neural network.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201510031651.1A CN104636822B (en) | 2015-01-21 | 2015-01-21 | A kind of resident load prediction technique based on elman neural networks |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201510031651.1A CN104636822B (en) | 2015-01-21 | 2015-01-21 | A kind of resident load prediction technique based on elman neural networks |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN104636822A CN104636822A (en) | 2015-05-20 |
| CN104636822B true CN104636822B (en) | 2018-08-31 |
Family
ID=53215546
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201510031651.1A Active CN104636822B (en) | 2015-01-21 | 2015-01-21 | A kind of resident load prediction technique based on elman neural networks |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN104636822B (en) |
Families Citing this family (22)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN106845770A (en) * | 2016-12-16 | 2017-06-13 | 深圳市麦澜创新科技有限公司 | A kind of energy storage management system based on neutral net |
| CN106971238A (en) * | 2017-03-10 | 2017-07-21 | 东华大学 | The Short-Term Load Forecasting Method of Elman neutral nets is obscured based on T S |
| CN107038502A (en) * | 2017-04-18 | 2017-08-11 | 国网安徽省电力公司芜湖供电公司 | Consider the improvement wavelet packet electricity demand forecasting method of Seasonal Characteristics |
| CN107590562A (en) * | 2017-09-05 | 2018-01-16 | 西安交通大学 | A kind of Short-Term Load Forecasting of Electric Power System based on changeable weight combination predicted method |
| CN108038568A (en) * | 2017-12-05 | 2018-05-15 | 国家电网公司 | A kind of changeable weight combination Short-Term Load Forecasting of Electric Power System based on particle cluster algorithm |
| CN108053070A (en) * | 2017-12-15 | 2018-05-18 | 国网江苏省电力有限公司淮安供电分公司 | The node load Forecasting Methodology of meter and high permeability distributed generation resource |
| CN109376907B (en) * | 2018-09-25 | 2022-06-28 | 国网浙江省电力有限公司经济技术研究院 | High-voltage distribution network transformer substation load prediction method suitable for integrated planning of transmission and distribution network |
| CN110009136A (en) * | 2019-03-12 | 2019-07-12 | 国网江西省电力有限公司电力科学研究院 | A kind of load forecasting method of distribution transformer and distribution line |
| CN109996247B (en) * | 2019-03-27 | 2022-03-11 | 中国电子科技集团公司信息科学研究院 | Networked resource allocation method, device, equipment and storage medium |
| CN110046765A (en) * | 2019-04-15 | 2019-07-23 | 国网甘肃省电力公司电力科学研究院 | A kind of power system stabilizer, PSS implementation method based on Elman neural network |
| CN110569981A (en) * | 2019-07-22 | 2019-12-13 | 国网冀北电力有限公司电力科学研究院 | Method and device for determining fuel consumption of scattered coal |
| CN110689208A (en) * | 2019-10-14 | 2020-01-14 | 国网山东省电力公司泰安供电公司 | Medium-and-long-term load prediction method based on uncertain load high permeability |
| CN111080477A (en) * | 2019-12-06 | 2020-04-28 | 北京国电通网络技术有限公司 | Household power load prediction method and system |
| CN111275571B (en) * | 2020-01-14 | 2020-12-11 | 河海大学 | Resident load probability prediction deep learning method considering microclimate and user mode |
| CN111563827B (en) * | 2020-04-10 | 2023-08-04 | 杭州电子科技大学 | Load decomposition method based on physical characteristics of electrical appliance and electricity utilization behaviors of residents |
| CN111898799B (en) * | 2020-06-24 | 2022-09-27 | 南京理工大学 | BFA-Elman-based power load prediction method |
| CN112200384B (en) * | 2020-10-28 | 2024-05-17 | 宁波立新科技股份有限公司 | EWT neural network-based short-time prediction method for power load |
| CN113139673A (en) * | 2020-11-04 | 2021-07-20 | 西安天和防务技术股份有限公司 | Method, device, terminal and storage medium for predicting air quality |
| CN113191559B (en) * | 2021-05-08 | 2024-04-19 | 苏州瑞城电力科技有限公司 | Improved neural network short-term resident load prediction method based on autoregressive selection |
| CN113300416B (en) * | 2021-07-07 | 2022-10-21 | 广东电网有限责任公司 | Power grid standby capacity configuration method, system, equipment and computer medium |
| CN114997477A (en) * | 2022-05-20 | 2022-09-02 | 国网江苏省电力有限公司苏州供电分公司 | Low-voltage residential district load prediction method and system |
| CN115829152B (en) * | 2022-12-21 | 2023-07-07 | 杭州易龙电安科技有限公司 | Power load prediction method, device and medium based on machine learning algorithm |
Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN101383023A (en) * | 2008-10-22 | 2009-03-11 | 西安交通大学 | Neural network short-term electric load prediction based on sample dynamic organization and temperature compensation |
| CN103093285A (en) * | 2013-01-22 | 2013-05-08 | 清华大学 | Short-term load forecast method based on artificial neural network |
| CN103793887A (en) * | 2014-02-17 | 2014-05-14 | 华北电力大学 | Short-term electrical load on-line predicting method based on self-adaptation enhancing algorithm |
| CN103854068A (en) * | 2013-12-06 | 2014-06-11 | 国家电网公司 | Method for forecasting residential quarter short-term loads |
| CN104268638A (en) * | 2014-09-11 | 2015-01-07 | 广州市香港科大霍英东研究院 | Photovoltaic power generation system power predicting method of elman-based neural network |
-
2015
- 2015-01-21 CN CN201510031651.1A patent/CN104636822B/en active Active
Patent Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN101383023A (en) * | 2008-10-22 | 2009-03-11 | 西安交通大学 | Neural network short-term electric load prediction based on sample dynamic organization and temperature compensation |
| CN103093285A (en) * | 2013-01-22 | 2013-05-08 | 清华大学 | Short-term load forecast method based on artificial neural network |
| CN103854068A (en) * | 2013-12-06 | 2014-06-11 | 国家电网公司 | Method for forecasting residential quarter short-term loads |
| CN103793887A (en) * | 2014-02-17 | 2014-05-14 | 华北电力大学 | Short-term electrical load on-line predicting method based on self-adaptation enhancing algorithm |
| CN104268638A (en) * | 2014-09-11 | 2015-01-07 | 广州市香港科大霍英东研究院 | Photovoltaic power generation system power predicting method of elman-based neural network |
Also Published As
| Publication number | Publication date |
|---|---|
| CN104636822A (en) | 2015-05-20 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN104636822B (en) | A kind of resident load prediction technique based on elman neural networks | |
| Ye et al. | Predicting electricity consumption in a building using an optimized back-propagation and Levenberg–Marquardt back-propagation neural network: Case study of a shopping mall in China | |
| CN111353656B (en) | Steel enterprise oxygen load prediction method based on production plan | |
| CN102426674B (en) | Power system load prediction method based on Markov chain | |
| CN107301472A (en) | Distributed photovoltaic planing method based on scene analysis method and voltage-regulation strategy | |
| CN102509173B (en) | A kind of based on markovian power system load Accurate Prediction method | |
| CN109376950A (en) | A kind of polynary Load Forecasting based on BP neural network | |
| CN105447509A (en) | Short-term power prediction method for photovoltaic power generation system | |
| CN110635518A (en) | Source network load and storage optimization method based on photovoltaic high permeability | |
| CN109325880A (en) | A kind of Mid-long term load forecasting method based on Verhulst-SVM | |
| Shah et al. | A novel prediction error-based power forecasting scheme for real pv system using pvusa model: A grey box-based neural network approach | |
| CN114429248A (en) | Transformer apparent power prediction method | |
| CN116722561A (en) | Reactive power compensation system | |
| CN109858668B (en) | Coordination prediction method for power load region in thunder and lightning climate | |
| CN114595542A (en) | Line loss prediction method | |
| CN116663870B (en) | Heat supply system scheduling method and system based on cloud computing | |
| Feng et al. | Peak operation optimization of cascade hydropower reservoirs and solar power plants considering output forecasting uncertainty | |
| CN111079966B (en) | Generalized load space prediction method and system | |
| Huang et al. | Wind prediction based on improved bp artificial neural network in wind farm | |
| Ye et al. | Multi-scenario stochastic optimal scheduling for power systems with source-load matching based on pseudo-inverse Laguerre polynomials | |
| CN112348235A (en) | Wind-solar bus load self-adaptive prediction method and device and computer equipment | |
| Xu et al. | Adaptive Feature Selection and GCN with Optimal Graph Structure-based Ultra-short-term Wind Farm Cluster Power Forecasting Method | |
| CN116683472B (en) | Reactive power compensation method, device, equipment and storage medium | |
| Baeva et al. | Most Favorable Results for Forecasting Methods for Natural Gas and Photovoltaic Energy Consumption | |
| CN116073449B (en) | Controllable photovoltaic participation peak shaving method based on low carbon benefit and uncertainty |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| C06 | Publication | ||
| PB01 | Publication | ||
| C10 | Entry into substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |