CN113159438A - Load weighting integrated prediction method based on differential multimode fusion - Google Patents
Load weighting integrated prediction method based on differential multimode fusion Download PDFInfo
- Publication number
- CN113159438A CN113159438A CN202110478395.6A CN202110478395A CN113159438A CN 113159438 A CN113159438 A CN 113159438A CN 202110478395 A CN202110478395 A CN 202110478395A CN 113159438 A CN113159438 A CN 113159438A
- Authority
- CN
- China
- Prior art keywords
- load
- prediction
- day
- state
- selecting
- 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
- 238000000034 method Methods 0.000 title claims abstract description 30
- 230000004927 fusion Effects 0.000 title claims abstract description 12
- 238000013528 artificial neural network Methods 0.000 claims abstract description 16
- 230000010354 integration Effects 0.000 claims abstract description 15
- 230000004069 differentiation Effects 0.000 claims abstract description 4
- 238000012549 training Methods 0.000 claims description 30
- 238000010606 normalization Methods 0.000 claims description 15
- 238000012545 processing Methods 0.000 claims description 13
- 230000007704 transition Effects 0.000 claims description 12
- 239000011159 matrix material Substances 0.000 claims description 9
- 238000012360 testing method Methods 0.000 claims description 7
- 238000006243 chemical reaction Methods 0.000 claims description 6
- 238000012546 transfer Methods 0.000 claims description 6
- 238000013139 quantization Methods 0.000 claims description 4
- 150000001875 compounds Chemical class 0.000 claims description 3
- 210000002569 neuron Anatomy 0.000 claims description 3
- 230000008859 change Effects 0.000 abstract description 7
- 238000011160 research Methods 0.000 abstract description 2
- 238000013135 deep learning Methods 0.000 description 2
- 230000006870 function Effects 0.000 description 2
- 230000008569 process Effects 0.000 description 2
- 238000005070 sampling Methods 0.000 description 2
- 238000004458 analytical method Methods 0.000 description 1
- 238000013473 artificial intelligence Methods 0.000 description 1
- 230000009286 beneficial effect Effects 0.000 description 1
- 238000009960 carding Methods 0.000 description 1
- 238000011161 development Methods 0.000 description 1
- 238000010586 diagram Methods 0.000 description 1
- 230000000694 effects Effects 0.000 description 1
- 238000005265 energy consumption Methods 0.000 description 1
- 238000013213 extrapolation Methods 0.000 description 1
- 238000012986 modification Methods 0.000 description 1
- 230000004048 modification Effects 0.000 description 1
- 238000005457 optimization Methods 0.000 description 1
- 238000010248 power generation Methods 0.000 description 1
- 238000007637 random forest analysis Methods 0.000 description 1
- 230000004044 response Effects 0.000 description 1
- 238000012706 support-vector machine Methods 0.000 description 1
Images
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
- G06Q10/00—Administration; Management
- G06Q10/04—Forecasting or optimisation specially adapted for administrative or management purposes, e.g. linear programming or "cutting stock problem"
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/21—Design or setup of recognition systems or techniques; Extraction of features in feature space; Blind source separation
- G06F18/214—Generating training patterns; Bootstrap methods, e.g. bagging or boosting
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/21—Design or setup of recognition systems or techniques; Extraction of features in feature space; Blind source separation
- G06F18/2163—Partitioning the feature space
-
- 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
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/08—Learning methods
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
- G06Q50/00—Systems or methods specially adapted for specific business sectors, e.g. utilities or tourism
- G06Q50/06—Electricity, gas or water supply
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Physics & Mathematics (AREA)
- Business, Economics & Management (AREA)
- Data Mining & Analysis (AREA)
- General Physics & Mathematics (AREA)
- Artificial Intelligence (AREA)
- Economics (AREA)
- Evolutionary Computation (AREA)
- General Engineering & Computer Science (AREA)
- Health & Medical Sciences (AREA)
- Life Sciences & Earth Sciences (AREA)
- Human Resources & Organizations (AREA)
- Strategic Management (AREA)
- General Health & Medical Sciences (AREA)
- Molecular Biology (AREA)
- Bioinformatics & Computational Biology (AREA)
- Biophysics (AREA)
- Biomedical Technology (AREA)
- Computing Systems (AREA)
- Mathematical Physics (AREA)
- Software Systems (AREA)
- Evolutionary Biology (AREA)
- Computer Vision & Pattern Recognition (AREA)
- General Business, Economics & Management (AREA)
- Tourism & Hospitality (AREA)
- Computational Linguistics (AREA)
- Marketing (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Primary Health Care (AREA)
- Water Supply & Treatment (AREA)
- Public Health (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 relates to a load weighting integration prediction method based on differentiation multimode fusion, which takes a power load in a short period as a research object, and takes the regularity and periodicity of the load and the correlation between the load and factors such as weather and the like into consideration to construct a load prediction regression model based on state similarity days; in consideration of the randomness and the fluctuation of the load, a Markov load prediction model is constructed; considering the diversity of load change, a neural network load prediction model is constructed by taking date attributes, temperature attributes and the like as input variables; finally, the prediction errors of various individual learners are effectively balanced through a weighted integration method. According to the load weighting integration prediction method based on the differential multimode fusion, three types of individual learners are selected in a differentiated mode according to three characteristics of load periodicity, randomness and diversity, the strong learners are formed through weighting integration, prediction errors are reduced, and the load prediction precision is effectively improved through the load weighting integration prediction method based on the differential multimode fusion.
Description
Technical Field
The invention relates to a load weighting integrated prediction method based on differentiation multimode fusion, and belongs to the technical field of power load prediction.
Background
In order to ensure the safe, stable and economic operation of the whole power system, a power grid company needs to make a reasonable power generation and dispatching plan. Accurate load prediction plays an indispensable role in the process, and the smaller the error of the prediction result is, the less the scheduling cost is, so that the load prediction precision needs to be improved by optimizing the prediction method.
The short-time power load prediction mainly refers to prediction of power load of several hours in the future of a day, is an important basis for development of demand response projects, and has important practical significance for peak load optimization, new energy consumption and the like. The short-time power load prediction has the characteristics of periodicity and uncertainty, and the load characteristics of the periodicity appearing on the same type of day may show similarity, such as the same week, the same season, the same holiday and the like; the uncertain phenomenon is that aiming at single load prediction, the short-time load is very easily influenced by various external factors, such as meteorological change, family member structure change and the like, has strong randomness, and aggravates the difficulty of short-time load prediction. The traditional short-time load prediction methods are intensively embodied into two categories, one is a classic load prediction method based on carding statistics and probability theory, and comprises a trend extrapolation method, a wavelet analysis method, a time sequence method and the like; and the other is a modern power load prediction method based on artificial intelligence, such as a support vector machine, a random forest, an artificial neural network and the like. At present, the methods are still in continuous searching, trying and updating, a method with higher prediction precision and more stable performance is determined, and the method becomes a technical problem to be solved urgently in the industry.
Disclosure of Invention
In order to solve the technical problems, the invention provides a load weighting integration prediction method based on differentiation multimode fusion, which takes a power load in a short period as a research object, comprehensively considers the three characteristics of periodicity, randomness and diversity of power load change, and combines the characteristics of an individual learner with the characteristics of the load. Considering the regularity and periodicity of the load and the correlation between the load and factors such as weather, and the like, constructing a load prediction regression model based on state similarity days; in consideration of the randomness and the fluctuation of the load, a Markov load prediction model is constructed; considering the diversity of load change, a neural network load prediction model is constructed by taking date attributes, temperature attributes and the like as input variables; finally, the prediction errors of various individual learners are effectively balanced through a weighting integration method, and the accuracy of load prediction is effectively improved. The specific technical scheme is as follows:
a load weighting integrated prediction method based on differential multimode fusion comprises the following steps:
firstly, an original data packet is segmented to form a training set { x (1), x (2), … x (k), … x (n) } and a test set, wherein in deep learning, an available data set is often divided into the training set and the test set, the training set is used for estimating a model, the test set is used for checking how to finally select the optimal model performance, the two parts are randomly extracted from samples, the data distribution is kept approximately consistent, the data distribution is similar to hierarchical sampling, and the quantity of the training set data is 2/3 to 4/5. Selecting a day D to be predicted;
step two, considering the regularity and periodicity of the load, selecting the training set formed by the division in the step one, and based on the same day F of the last weekwkThe same day of the first month FmtThe first three days of similar load average FavWeather similar day FwtSelecting 4 state similar days (highest temperature, lowest temperature and 24-point temperature) as independent variables to screen similar days of the day to be tested, selecting 4 state similar day load data to perform regression model training, and obtaining a daily load predicted value F of the predicted monthmrThe training formula is as follows:
Fmr＝αFwk+βFmt+γFav+λFwt+η (1)
where α, β, γ, λ are the values of the parameters at which the loss function training model is minimal, and η is an error term (which is a random variable) used to capture any data for F except for 4 state-like daily load datamrThe influence of (c).
Step three, considering the fluctuation and the randomness of the load, counting the load values of the training set optimized in the step two, selecting a proper interval size to divide the load state into N states according to N moment load values in the optimized training set, wherein N is a positive integer greater than 3, and the N state space sequences can be expressed as E ═ E1，e2，…，en]。
Step four, comprehensively considering the randomness of load prediction, and in step three, obtaining a new state space sequence E ═ E1，e2，…，en]On the basis, a load Markov probability transfer matrix is solved, and a Markov probability transfer model is used for predicting load prediction values of days D to be predicted at t time and (t +1) time, so that two-stage load prediction based on a regression-Markov chain can be realized, and the load prediction precision is improved; obtaining n possible states according to the third step, and recording PijIs in slave state EiTransition to State EjThe probability transition matrix is then:
in the formula (I), the compound is shown in the specification,
fijfrequency, f, for one step transition from state i to state jiIs the sum of the number of occurrences of state i.
Step five, setting an initial state Ei of the daily load to be predicted according to the load value at the moment t, and selecting the maximum transition probability P of the (t +1) moment difference value according to the probability transition matrixijCorresponding state EjAnd so on to finish the day D to be predictedLoad prediction at all times of the day to obtain a load prediction curve Fmk；
Step six, considering load diversity, selecting and quantifying load characteristic quantity, comprehensively considering date attribute p, weather attribute q and temperature attribute s, selecting characteristic quantity { p (k), … q (k), … s (k) } and quantifying non-datamation characteristic quantity;
step seven, selecting the training set formed by the division in the step one, and in order to eliminate the influence of magnitude mismatching, carrying out normalization processing on the characteristic quantities { p (k), (… q) (k), … s (k) } selected in the step six to be used as input variables of the neural network load prediction model, wherein a conversion formula of the normalization processing is as follows:
in the formula, x0、xnRespectively representing the values before and after normalization, xmin、xmaxRespectively representing the minimum value and the maximum value in the sample data;
after BP neural network load prediction processing, inverse normalization processing is carried out on the prediction result, so that the information of the original data can be restored, and the inverse normalization conversion formula is as follows:
x0＝xn(xmax-xmin)+xmin (4)
step eight, taking the normalized data obtained in the step seven as input variables, constructing a BP neural network load prediction model, and performing load curve prediction on the day to be predicted to obtain a prediction result Fnn；
The BP neural network consists of an input layer, a hidden layer and an output layer, wherein the input layer is a load characteristic sample containing three types of data of p, q and s; the output layer has a neuron, i.e. a load prediction value Fnn(ii) a The hidden layer is comprehensively determined according to the number of samples of the input layer and the number of output layers;
step nine, adopting a weighted integration combination strategy to the prediction results of the three individual learners to obtain a final prediction result,
F＝aFmk+bFmr+cFnn (5)
wherein: a. b and c are respectively the weights occupied by the three load prediction methods, a + b + c is 1, and the weights are selected and distributed according to the average value, namely a is b is c.
Further, the weather similar day in the second step is an approximate day based on the day maximum temperature, the day minimum temperature and the 24-point temperature.
Further, in the third step, N is 25, and the states are divided as follows:
E1(0～0.5kW)，E2(0.5～1kW)，E3(1～1.5kW)，E4(1.5～2kW)，E5(2～2.5kW)，E6(2.5～3kW)，E7(3～3.5kW)，E8(3.5～4kW)，E9(4～4.5kW)，E10(4.5～5kW)，E11(5～5.5kW)，E12(5.5～6kW)，E13(6 to 6.5kW) is E14(6.5～7kW)，E15(7～7.5kW)，E16(7.5～8kW)，E17(8～8.5kW)，E18(8.5～9kW)，E19(9～9.5kW)，E20(9.5～10kW)，E21(10～10.5kW)，E22(10.5～11kW)，E23(11～11.5kW)，E24(11.5～12kW)，E25(＞12kW)。
Further, in the sixth step, 8 feature quantities including a week type (monday to sunday), a date type (holiday/workday), a highest day temperature, a lowest day temperature, a weather condition, a load of a previous day, an average load of a previous three days, and a load of a same day of a previous week are selected, and the quantization is as follows:
[ monday, tuesday, wednesday, thursday, friday, saturday, sunday ] ═ 1, 2, 3, 4, 5, 6, 7 ];
work day, holiday ] ═ 1, 2;
[ fine, cloudy, rainy, snowy, and extremely bad ] ═ 1, 2, 3, 4, 5, 6 ].
The invention has the beneficial effects that:
the invention comprehensively considers the three characteristics of periodicity, randomness and diversity of power load change and combines the characteristics of the individual learners with the characteristics of the load. Considering the regularity and periodicity of the load and the correlation between the load and factors such as weather, and the like, constructing a load prediction regression model based on state similarity days; in consideration of the randomness and the fluctuation of the load, a Markov load prediction model is constructed; considering the diversity of load change, a neural network load prediction model is constructed by taking date attributes, temperature attributes and the like as input variables; finally, the prediction errors of various individual learners are effectively balanced through a weighting integration method, the accuracy of load prediction is effectively improved, and a better actual load prediction application effect is obtained. The model provides a new idea for forecasting the short-time load of the power system under the intelligent power grid.
Drawings
FIG. 1 is a diagram of the execution logic of the present invention.
Detailed Description
The present invention will now be described in further detail with reference to the accompanying drawings.
As shown in fig. 1, the specific process of the present invention is:
firstly, an original data packet is segmented to form a training set { x (1), x (2), … x (k), … x (n) } and a test set, a day D to be predicted is selected, an available data set is often divided into the training set and the test set in deep learning, the training set is used for estimating a model, the test set is used for checking how to finally select the optimal model performance, the two parts are randomly extracted from samples, the data distribution is kept approximately consistent, the data distribution is similar to hierarchical sampling, and the number of the training set data accounts for 2/3 to 4/5.
Step two, considering the regularity and periodicity of the load, selecting the training set formed by the division in the step one, and based on the same day F of the last weekwkThe same day of the first month FmtThe first three days of similar load average FavWeather similar day Fw1(weather similar day F)wtDate with highest day temperature, lowest day temperature and 24-point temperature closest) 4 state similar days are used as independent variables to screen similar days of the day to be tested, 4 state similar day load data are selected to perform regression model training, and a daily load predicted value F of the month to be predicted is obtainedmrThe training formula is as follows:
Fmr＝αFwk+βFmt+γFav+λFwt+η (1)
where α, β, γ, λ are the values of the parameters at which the loss function training model is minimal, and η is an error term (which is a random variable) used to capture any data for F except for 4 state-like daily load datamrThe influence of (c).
Step three, considering the fluctuation and the randomness of the load, counting the load values of the training set optimized in the step two, selecting a proper interval size to divide the load state into N states according to N moment load values in the optimized training set, wherein N is a positive integer greater than 3, and the N state space sequences can be expressed as E ═ E1，e2，…，en]。
Step four, comprehensively considering the randomness of load prediction, and in step three, obtaining a new state space sequence E ═ E1，e2，…，en]On the basis, a load Markov probability transfer matrix is solved, and a Markov probability transfer model is used for predicting load prediction values of days D to be predicted at t time and (t +1) time, so that two-stage load prediction based on a regression-Markov chain can be realized, and the load prediction precision is improved; obtaining n possible states according to the third step, and recording PijIs in slave state EiTransition to State EjThe probability transition matrix is then:
in the formula (I), the compound is shown in the specification,
fijfrequency, f, for one step transition from state i to state jiIs the sum of the number of occurrences of state i.
Step five, setting an initial state Ei of the daily load to be predicted according to the load value at the moment t, and selecting the maximum transition probability P of the (t +1) moment difference value according to the probability transition matrixijCorresponding state EjAnd by analogy, completing the load prediction of all the time of the day D to be predicted to obtain a load prediction curve Fmk；
Step six, considering load diversity, selecting and quantifying load characteristic quantity, comprehensively considering date attribute p, weather attribute q and temperature attribute s, selecting characteristic quantity { p (k), … q (k), … s (k) } and quantifying non-datamation characteristic quantity;
step seven, selecting the training set formed by the division in the step one, and in order to eliminate the influence of magnitude mismatching, carrying out normalization processing on the characteristic quantities { p (k), (… q) (k), … s (k) } selected in the step six to be used as input variables of the neural network load prediction model, wherein a conversion formula of the normalization processing is as follows:
in the formula, x0、xnRespectively representing the values before and after normalization, xmin、xmaxRespectively representing the minimum value and the maximum value in the sample data;
after BP neural network load prediction processing, inverse normalization processing is carried out on the prediction result, so that the information of the original data can be restored, and the inverse normalization conversion formula is as follows:
x0＝xn(xmax-xmin)+xmin (4)
step eight, taking the normalized data obtained in the step seven as input variables, constructing a BP neural network load prediction model, and performing load curve prediction on the day to be predicted to obtain a prediction result Fnn；
The BP neural network consists of an input layer, a hidden layer and an output layer, wherein the input layer is a load characteristic sample containing three types of data of p, q and s; the output layer has a neuron, i.e. a load prediction value Fnn(ii) a The hidden layer is comprehensively determined according to the number of samples of the input layer and the number of output layers;
step nine, adopting a weighted integration combination strategy to the prediction results of the three individual learners to obtain a final prediction result,
F＝aFmk+bFmr+cFnn (5)
wherein: a. b and c are respectively the weights occupied by the three load prediction methods, a + b + c is 1, and the weights are selected and distributed according to the average value, namely a is b is c.
The load state division performed in step three fully considers the load size, and the interval is as fine as possible to 25, where N is 25, and the state division is as follows:
E1(0～0.5kW)，E2(0.5～1kW)，E3(1～1.5kW)，E4(1.5～2kW)，E5(2～2.5kW)，E6(2.5～3kW)，E7(3～3.5kW)，E8(3.5～4kW)，E9(4～4.5kW)，E10(4.5～5kW)，E11(5～5.5kW)，E12(5.5～6kW)，E13(6 to 6.5kW) is E14(6.5～7kW)，E15(7～7.5kW)，E16(7.5～8kW)，E17(8～8.5kW)，E18(8.5～9kW)，E19(9～9.5kW)，E20(9.5～10kW)，E21(10～10.5kW)，E22(10.5～11kW)，E23(11～11.5kW)，E24(11.5～12kW)，E25(＞12kW)。
In the sixth step, the selection of the load characteristic quantities is based on the consideration of load regularity, the influences of date attributes, weather attributes, temperature attributes and the like on the load are fully considered, quantization processing is performed, 8 characteristic quantities including a week type (Monday-Sunday), a date type (holiday/working day), a day highest temperature, a day lowest temperature, a weather condition, a load on the previous day, an average load on the previous three days and a load on the same day of the previous week are selected, and the quantization is as follows:
[ monday, tuesday, wednesday, thursday, friday, saturday, sunday ] ═ 1, 2, 3, 4, 5, 6, 7 ];
work day, holiday ] ═ 1, 2;
[ fine, cloudy, rainy, snowy, and extremely bad ] ═ 1, 2, 3, 4, 5, 6 ].
In light of the foregoing description of the preferred embodiment of the present invention, many modifications and variations will be apparent to those skilled in the art without departing from the spirit and scope of the invention. The technical scope of the present invention is not limited to the content of the specification, and must be determined according to the scope of the claims.
Claims (4)
1. A load weighting integration prediction method based on differentiation multimode fusion is characterized in that: the method comprises the following steps:
firstly, segmenting an original data packet to form a training set { x (1), x (2), … x (k), … x (n) } and a test set, and selecting a day D to be predicted;
step two, considering the regularity and periodicity of the load, selecting the training set formed by the division in the step one, and based on the same day F of the last weekwkThe same day of the first month FmtThe first three days of similar load average FavWeather similar day FwtSelecting 4 state similar days (highest temperature, lowest temperature and 24-point temperature) as independent variables to screen similar days of the day to be tested, selecting 4 state similar day load data to perform regression model training, and obtaining a daily load predicted value F of the predicted monthmrThe training formula is as follows:
Fmr＝αFwk+βFmt+γFav+λFwt+η (1)
where α, β, γ, λ are the values of the parameters at which the loss function training model is minimal, and η is an error term (which is a random variable) used to capture any data for F except for 4 state-like daily load datamrThe influence of (a);
step three, considering the fluctuation and the randomness of the load, counting the load values of the training set optimized in the step two, selecting a proper interval size to divide the load state into N states according to N moment load values in the optimized training set, wherein N is a positive integer greater than 3, and the N state space sequences can be expressed as E ═ E1，e2，…，en]；
Step four, comprehensively considering the randomness of load prediction, and in step three, obtaining a new state spaceSequence E ═ E1，e2，…，en]On the basis, a load Markov probability transfer matrix is solved, and a Markov probability transfer model is used for predicting load prediction values of days D to be predicted at t time and (t +1) time, so that two-stage load prediction based on a regression-Markov chain can be realized, and the load prediction precision is improved; obtaining n possible states according to the third step, and recording PijIs in slave state EiTransition to State EjThe probability transition matrix is then:
in the formula (I), the compound is shown in the specification,
fijfrequency, f, for one step transition from state i to state jiIs the sum of the number of occurrences of state i;
step five, setting an initial state Ei of the daily load to be predicted according to the load value at the moment t, and selecting the maximum transition probability P of the (t +1) moment difference value according to the probability transition matrixijCorresponding state EjAnd by analogy, completing the load prediction of all the time of the day D to be predicted to obtain a load prediction curve Fmk；
Step six, considering load diversity, selecting and quantifying load characteristic quantity, comprehensively considering date attribute p, weather attribute q and temperature attribute s, selecting characteristic quantity { p (k), … q (k), … s (k) } and quantifying non-datamation characteristic quantity;
step seven, selecting the training set formed by the division in the step one, and in order to eliminate the influence of magnitude mismatching, carrying out normalization processing on the characteristic quantities { p (k), (… q) (k), … s (k) } selected in the step six to be used as input variables of the neural network load prediction model, wherein a conversion formula of the normalization processing is as follows:
in the formula, x0、xnRespectively representing the values before and after normalization, xmin、xmaxRespectively representing the minimum value and the maximum value in the sample data;
after BP neural network load prediction processing, inverse normalization processing is carried out on the prediction result, so that the information of the original data can be restored, and the inverse normalization conversion formula is as follows:
x0＝xn(xmax-xmin)+xmin (4)
step eight, taking the normalized data obtained in the step seven as input variables, constructing a BP neural network load prediction model, and performing load curve prediction on the day to be predicted to obtain a prediction result Fnn；
The BP neural network consists of an input layer, a hidden layer and an output layer, wherein the input layer is a load characteristic sample containing three types of data of p, q and s; the output layer has a neuron, i.e. a load prediction value Fnn(ii) a The hidden layer is comprehensively determined according to the number of samples of the input layer and the number of output layers;
step nine, adopting a weighted integration combination strategy to the prediction results of the three individual learners to obtain a final prediction result,
F＝aFmk+bFmr+cFnn (5)
wherein: a. b and c are respectively the weights occupied by the three load prediction methods, a + b + c is 1, and the weights are selected and distributed according to the average value, namely a is b is c.
2. The load weighting integration prediction method based on differential multi-mode fusion according to claim 1, characterized in that: the weather similar day in the step two is an approximate day based on the highest temperature of the day, the lowest temperature of the day and the 24-point temperature.
3. The load weighting integration prediction method based on differential multi-mode fusion according to claim 1, characterized in that: in the third step, N is 25, and the states are divided as follows:
E1(0～0.5kW)，E2(0.5～1kW)，E3(1～1.5kW)，E4(1.5～2kW)，E5(2～2.5kW)，E6(2.5～3kW)，E7(3～3.5kW)，E8(3.5～4kW)，E9(4～4.5kW)，E10(4.5～5kW)，E11(5～5.5kW)，E12(5.5～6kW)，E13(6 to 6.5kW) is E14(6.5～7kW)，E15(7～7.5kW)，E16(7.5～8kW)，E17(8～8.5kW)，E18(8.5～9kW)，E19(9～9.5kW)，E20(9.5～10kW)，E21(10～10.5kW)，E22(10.5～11kW)，E23(11～11.5kW)，E24(11.5～12kW)，E25(＞12kW)。
4. The load weighting integration prediction method based on differential multi-mode fusion according to claim 1, characterized in that: in the sixth step, 8 feature quantities including a week type (Monday-Sunday), a date type (holiday/working day), a highest day temperature, a lowest day temperature, a weather condition, a load of a previous day, an average load of a previous three days, and a load of the same day of the previous week are selected, and the quantization is as follows:
[ monday, tuesday, wednesday, thursday, friday, saturday, sunday ] ═ 1, 2, 3, 4, 5, 6, 7 ];
work day, holiday ] ═ 1, 2;
[ fine, cloudy, rainy, snowy, and extremely bad ] ═ 1, 2, 3, 4, 5, 6 ].
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202110478395.6A CN113159438A (en) | 2021-04-30 | 2021-04-30 | Load weighting integrated prediction method based on differential multimode fusion |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202110478395.6A CN113159438A (en) | 2021-04-30 | 2021-04-30 | Load weighting integrated prediction method based on differential multimode fusion |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN113159438A true CN113159438A (en) | 2021-07-23 |
Family
ID=76872679
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202110478395.6A Pending CN113159438A (en) | 2021-04-30 | 2021-04-30 | Load weighting integrated prediction method based on differential multimode fusion |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN113159438A (en) |
Cited By (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN113870531A (en) * | 2021-09-17 | 2021-12-31 | 成都理工大学 | Monitoring frequency adjusting method for geological disaster monitoring and early warning system |
Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN106971240A (en) * | 2017-03-16 | 2017-07-21 | 河海大学 | The short-term load forecasting method that a kind of variables choice is returned with Gaussian process |
| CN108229754A (en) * | 2018-01-31 | 2018-06-29 | 杭州电子科技大学 | Short-term load forecasting method based on similar day segmentation and LM-BP networks |
| CN111860979A (en) * | 2020-07-01 | 2020-10-30 | 广西大学 | Short-term load prediction method based on TCN and IPSO-LSSVM combined model |
| CN111932402A (en) * | 2020-08-13 | 2020-11-13 | 河南理工大学 | Short-term power load bidirectional combined prediction method based on similar day and LSTM |
-
2021
- 2021-04-30 CN CN202110478395.6A patent/CN113159438A/en active Pending
Patent Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN106971240A (en) * | 2017-03-16 | 2017-07-21 | 河海大学 | The short-term load forecasting method that a kind of variables choice is returned with Gaussian process |
| CN108229754A (en) * | 2018-01-31 | 2018-06-29 | 杭州电子科技大学 | Short-term load forecasting method based on similar day segmentation and LM-BP networks |
| CN111860979A (en) * | 2020-07-01 | 2020-10-30 | 广西大学 | Short-term load prediction method based on TCN and IPSO-LSSVM combined model |
| CN111932402A (en) * | 2020-08-13 | 2020-11-13 | 河南理工大学 | Short-term power load bidirectional combined prediction method based on similar day and LSTM |
Cited By (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN113870531A (en) * | 2021-09-17 | 2021-12-31 | 成都理工大学 | Monitoring frequency adjusting method for geological disaster monitoring and early warning system |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| Khotanzad et al. | A neuro-fuzzy approach to short-term load forecasting in a price-sensitive environment | |
| CN111861013A (en) | Power load prediction method and device | |
| CN113822481A (en) | Comprehensive energy load prediction method based on multi-task learning strategy and deep learning | |
| CN115907611B (en) | Fitting inventory control method based on fitting market value | |
| CN114676941B (en) | Electric-thermal load combined self-adaptive prediction method and device for park comprehensive energy system | |
| CN113554466A (en) | Short-term power consumption prediction model construction method, prediction method and device | |
| CN111222700A (en) | Day-ahead electricity price probability prediction method based on dynamic network quantile model | |
| Fan et al. | Multi-objective LSTM ensemble model for household short-term load forecasting | |
| Hu et al. | Multi-objective optimization of time-of-use price for tertiary industry based on generalized seasonal multi-model structure | |
| CN113159438A (en) | Load weighting integrated prediction method based on differential multimode fusion | |
| CN115660725A (en) | Method for depicting multi-dimensional energy user portrait | |
| CN114819395A (en) | Industry medium and long term load prediction method based on long and short term memory neural network and support vector regression combination model | |
| CN114091776A (en) | K-means-based multi-branch AGCNN short-term power load prediction method | |
| CN116826745B (en) | Layered and partitioned short-term load prediction method and system in power system background | |
| Chen | Estimating job cycle time in a wafer fabrication factory: A novel and effective approach based on post-classification | |
| CN112288187A (en) | Big data-based electricity sales amount prediction method | |
| CN115481788B (en) | Phase change energy storage system load prediction method and system | |
| CN116307250A (en) | Short-term load prediction method and system based on typical daily feature selection | |
| CN110659775A (en) | LSTM-based improved electric power short-time load prediction algorithm | |
| CN114004530A (en) | Enterprise power credit score modeling method and system based on sequencing support vector machine | |
| Rafi | Short term electric load forecasting using high precision convolutional neural network | |
| Sharma et al. | Synergism of recurrent neural network and fuzzy logic for short term energy load forecasting | |
| Montenegro et al. | A DNN Approach to Improving the Short-Term Investment Criteria for S&P500 Index Stock Market | |
| CN113283638A (en) | Load extreme curve prediction method and system based on fusion model | |
| CN112801415A (en) | Ultra-short-term load prediction method and system based on Markov chain distribution model |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| RJ01 | Rejection of invention patent application after publication | ||
| RJ01 | Rejection of invention patent application after publication |
Application publication date: 20210723 |