CN105069519A - Intelligent power grid park terminal user energy demand condition dynamic prediction system and method - Google Patents

Abstract

The invention discloses an intelligent power grid park terminal user energy demand condition dynamic prediction system and a method. The method includes: conducting a main constituent analysis of meteorological factors for influencing the cooling and heating load demand of intelligent power grid park terminal users; converting related variables to a few of linear independent random variables; quantifying the weather factor and the day type, conducting an analysis with historical load data by employing a fuzzy clustering method, and forming a sample; representing load characteristics of various types of loads and various types of distributed energy supply systems in the intelligent power grid park in load curves; and finally solving a model according to the process of a BP neural network algorithm, and obtaining a cooling and heating load prediction result. The system comprises a main constituent analysis module, an analysis sample formation module, a load characteristic curve module, and a load prediction module. According to the method and the system, the network size is reduced, the prediction precision is improved, and advantages of a BP neural network for large-scale parallel processing and adaptive learning ability are fully developed.

Description

A kind of intelligent grid garden terminal user's energy demand situation Dynamic Forecasting System and method

Technical field

The invention belongs to technical field of power systems, particularly relate to a kind of intelligent grid garden terminal user's energy demand situation Dynamic Forecasting System and method.

Background technology

Load prediction as the important component part of intelligent grid garden energy management, for the links such as planning construction, operation optimum management of intelligent grid garden provides important decision support.But intelligent grid garden user terminal One's name is legion, type is complicated, and the users such as industrial user, data center and public organizations there are differences with energy curve, and have bigger difference to the demand status of cool and thermal power load, load prediction exists certain difficulty.

Traditional load forecasting method mainly contains time series method and regression analysis, time series method does not consider the impact of weather on load, and regression analysis is difficult to solve dynamic, the nonlinear relation between the variable such as load and weather, more for garden user type, the situation that on-load demand there are differences, the precision of prediction of classic method can not be satisfactory, in the face of mass data process with excavate also Shortcomings.How mass data mining technology being applied in energy demand performance prediction technically, playing the advantage of data mining and intelligent algorithm, improve precision and the efficiency of prediction, is the outstanding problem of pendulum in face of intelligent grid garden supvr.

Summary of the invention

The object of the present invention is to provide a kind of intelligent grid garden terminal user's energy demand situation Dynamic Forecasting System and method, being intended to solve the precision of prediction that traditional load forecasting method exists can not be satisfactory, in the face of mass data processing and the problem excavating also Shortcomings.

The present invention is achieved in that a kind of intelligent grid garden terminal user's energy demand situation dynamic prediction method, and described intelligent grid garden terminal user's energy demand situation dynamic prediction method comprises:

First on affecting the maximum temperature of intelligent grid garden terminal user's cool and thermal power workload demand, minimum temperature, medial temperature, rainfall amount, wind speed, intensity of solar radiation meteorologic factor carry out principal component analysis (PCA);

Secondly correlated variables is converted to the stochastic variable of minority linear independence; By weather conditions and day type quantize, carry out analysis with historical load data acquisition fuzzy clustering method and form sample;

Then the part throttle characteristics of load polytype in intelligent grid garden, polytype distributing-supplying-energy system is all embodied in load curve;

Finally according to BP neural network algorithm flow process, model is solved, obtain cool and thermal power load prediction results.

Further, described intelligent grid garden terminal user's energy demand situation dynamic prediction method specifically comprises the following steps:

The first step, uses principal component analysis (PCA) to analyze the many factors affecting load; Based on method of fuzzy cluster analysis, user type is classified; BP neural network learnt and trains, after network convergence, performance prediction being carried out to the cool and thermal power load of intelligent grid garden;

Second step, the classification using the shortest classification of Euclidean distance as prediction day, sets up BP neural network and predicts, obtain intelligent grid garden terminal user's cool and thermal power load data.

Further, described utilization principal component analysis (PCA) comprises affecting the step that load many factors analyzes:

Step one, standardization is carried out to sample data:

Raw data matrix:

In formula, n is number of samples; P is each sample dimension; X _ijbe the jth dimension value of i-th sample, use X ₁, X ₂..., X _peach column vector of representing matrix X respectively, has:

$X_j^{*}=\frac{X_j-E\left(X_j\right)}{\sqrt{Vax\left(X_j\right)}};$

E (X in formula _j) and Vax (X _j) represent X respectively _javerage and variance;

Step 2, calculating correlation matrix R:

$r_{ij}=\frac{Cov(X_i,X_j)}{\sqrt{Var\left(X_i\right)Var\left(X_j\right)}}=\frac{1}{n}\underset{k=1}{\overset{n}{\Σ}}{X}_{ki}^{*}{X}_{kj}^{*};$

Cov (X in formula _i, X _j) represent the i-th row and jth in data matrix arrange between covariance;

Step 3, ask orthogonal matrix and eigenwert thereof:

P ^TRP＝diag(λ ₁,λ ₂,…,λ _p)；

λ in formula ₁>=λ ₂>=...>=λ _pbe p the eigenwert of R, diag represents diagonal matrix.

Further, described fuzzy cluster analysis comprises historical load data clusters step:

Step one, sample data carried out to normalization process:

x' _jk＝(x _jk-x _kmin)/(x _kmax-x _kmin)；

In formula, x _kmax, x _kminbe respectively x _1k, x _2k..., x _nkmaximal value and minimum value; X' _jkfor the data after normalization;

Step 2, set up fuzzy resembling relation matrix R={r _ij}:

$r_{ij}=\exp (-\underset{k=1}{\overset{m}{\Σ}}|x_{ik}^{,}-x_{jk}^{,}\left|\right);$

Step 3, carry out dynamic clustering:

$C_i=\frac{{\mathrm{\λ}}_{i-1}-{\mathrm{\λ}}_i}{n_i-n_{i-1}};$

In formula, i is λ polymerization order number from high to low, n _iand n _i-1be respectively the element number of i-th time and the i-th-1 time cluster; λ _iand λ _i-1be respectively confidence level when i-th time and the i-th-1 time cluster, if C _i=max (C _j), then think the confidence level λ of i-th cluster _ifor optimal threshold;

Step 4, computational prediction day and above-mentioned all kinds of Euclidean distance:

$d_{ij}=\frac{1}{\sqrt{m}}\sqrt{\underset{k=1}{\overset{m}{\Σ}}{(x_{ik}^{,}-x_{jk}^{,})}^2};$

X' in formula _ikfor predicting the characteristic index vector of day, x' _jkfor the characteristic index vector of each classification.

Another object of the present invention is to provide a kind of intelligent grid garden terminal user's energy demand situation Dynamic Forecasting System, described intelligent grid garden terminal user's energy demand situation Dynamic Forecasting System comprises:

Principal component analysis (PCA) module, on affecting the maximum temperature of intelligent grid garden terminal user's cool and thermal power workload demand, minimum temperature, medial temperature, rainfall amount, wind speed, intensity of solar radiation meteorologic factor carry out principal component analysis (PCA);

Analyzing samples forms module, correlated variables is converted to the stochastic variable of minority linear independence; By weather conditions and day type quantize, carry out analysis with historical load data acquisition fuzzy clustering method and form sample;

Load characteristic curve module, is all embodied in the part throttle characteristics of load polytype in intelligent grid garden, polytype distributing-supplying-energy system in load curve;

Load prediction module, solves model according to BP neural network algorithm flow process, obtains cool and thermal power load prediction results.

Terminal user's energy demand situation Dynamic Forecasting System in intelligent grid garden provided by the invention and method, the feature of cool and thermal power workload demand in combined with intelligent electrical network garden, based on data mining technology, consider the historical load factor and meteorologic factor that affect workload demand, time series method and regression analysis are combined, achieves gratifying precision of prediction by combination forecasting method.The present invention has also taken into full account difference and the feature of intelligent grid garden cool and thermal power load, data mining technology is utilized to reduce network size and improve precision of prediction, also give full play to the advantage of BP neural network in massively parallel processing and adaptive learning ability simultaneously, improve the precision of intelligent grid garden cool and thermal power load prediction.

The feature of cool and thermal power workload demand in combined with intelligent electrical network garden of the present invention, based on data mining technology, consider the historical load factor and meteorologic factor that affect workload demand, time series method and regression analysis are combined, achieves gratifying precision of prediction by combination forecasting method; The present invention uses principal component analysis (PCA), converts the numerous correlated variabless affecting cool and thermal power workload demand to a few major component, reduces the scale of network; Use Fuzzy Cluster Analysis method, carry out cluster to historical load data, get the classification of the shortest class of Euclidean distance as prediction day, such sample comprised is the training sample of BP network; The weights and bias of Reasonable adjustment BP network, after network convergence, predicts the cool and thermal power load of intelligent grid garden terminal user.

Accompanying drawing explanation

Fig. 1 is intelligent grid garden terminal user's energy demand situation dynamic prediction method process flow diagram that the embodiment of the present invention provides;

Fig. 2 is the intelligent grid garden electric load curve synoptic diagram that the embodiment of the present invention provides;

Fig. 3 is the intelligent grid garden thermal load curve synoptic diagram that the embodiment of the present invention provides;

Fig. 4 is the cold and hot lotus curve synoptic diagram in intelligent grid garden that the embodiment of the present invention provides;

Fig. 5 is the process flow diagram that utilization BP nerual network technique that the embodiment of the present invention provides carries out load prediction;

Fig. 6 be the embodiment of the present invention provide carry out electric load according to BP nerual network technique and predict the prediction load and actual load comparison diagram that obtain;

Fig. 7 be the embodiment of the present invention provide carry out prediction load that heat load prediction obtains and actual load comparison diagram according to BP nerual network technique;

Fig. 8 be the embodiment of the present invention provide carry out prediction load that cooling load prediction obtains and actual load comparison diagram according to BP nerual network technique.

Embodiment

In order to make object of the present invention, technical scheme and advantage clearly understand, below in conjunction with embodiment, the present invention is further elaborated.Should be appreciated that specific embodiment described herein only in order to explain the present invention, be not intended to limit the present invention.

Intelligent grid garden terminal user's energy demand situation dynamic prediction method based on data mining technology and BP neural network of the present invention, for the technical matters such as solve that Consideration in existing intelligent grid garden cool and thermal power load prediction process is not enough, the data mining degree of depth is not enough and precision of prediction is lower; Data mining technology is combined with intelligent algorithm, is improved the precision of cool and thermal power load prediction by combination forecasting method.

Below in conjunction with accompanying drawing, application principle of the present invention is explained in detail.

As shown in Figure 1, intelligent grid garden terminal user's energy demand situation dynamic prediction method of the embodiment of the present invention comprises the following steps:

S101: principal component analysis (PCA) is carried out on the meteorologic factor such as maximum temperature, minimum temperature, medial temperature, rainfall amount, wind speed, intensity of solar radiation affecting intelligent grid garden terminal user's cool and thermal power workload demand;

S102: stochastic variable correlated variables being converted to minority linear independence;

S103: by weather conditions and day type quantize, adopt fuzzy clustering method to carry out analysis together with historical load data and form sample.

S104: the part throttle characteristics of load polytype in intelligent grid garden, polytype distributing-supplying-energy system is all embodied in load curve;

S105: according to BP neural network algorithm flow process, model is solved, obtain cool and thermal power load prediction results.

The part throttle characteristics of load polytype in intelligent grid garden, polytype distributing-supplying-energy system is all embodied in load curve, cool and thermal power load respectively as shown in Figure 2, Figure 3 and Figure 4, then according to the BP neural network algorithm flow process shown in Fig. 5, model is solved, obtain the cool and thermal power load prediction results as shown in Fig. 6, Fig. 7 and Fig. 8.

Below in conjunction with specific embodiment, application principle of the present invention is further described:

First, principal component analysis (PCA) is used to analyze the many factors affecting load; Based on method of fuzzy cluster analysis, user type is classified; BP neural network learnt and trains, after network convergence, performance prediction being carried out to the cool and thermal power load of intelligent grid garden;

In implementation process, utilize load prediction results to dispatch powering device and energy storage device, realize the Optimum utilization of the intelligent grid garden energy.

Wherein, use principal component analysis (PCA) to analyze the many factors affecting load, its step comprises:

Step one, standardization is carried out to sample data:

Raw data matrix:

In formula, n is number of samples; P is each sample dimension; X _ijit is the jth dimension value of i-th sample.Use X ₁, X ₂..., X _peach column vector of representing matrix X respectively, has:

$X_j^{*}=\frac{X_j-E\left(X_j\right)}{\sqrt{Vax\left(X_j\right)}};$

E (X in formula _j) and Vax (X _j) represent X respectively _javerage and variance.

Step 2, calculating correlation matrix R:

$r_{ij}=\frac{Cov(X_i,X_j)}{\sqrt{Var\left(X_i\right)Var\left(X_j\right)}}=\frac{1}{n}\underset{k=1}{\overset{n}{\Σ}}{X}_{ki}^{*}{X}_{kj}^{*};$

Cov (X in formula _i, X _j) represent the i-th row and jth in data matrix arrange between covariance.

Step 3, ask orthogonal matrix and eigenwert thereof:

P ^TRP＝diag(λ ₁,λ ₂,…,λ _p)；

λ in formula ₁>=λ ₂>=...>=λ _pbe p the eigenwert of R, diag represents diagonal matrix.

Each principal set has suffered the different common trait of each component of stochastic variable X, and the dimension of such variable reduces, below based on fuzzy cluster analysis to historical load data clusters, its step comprises:

Step one, sample data carried out to normalization process:

x' _jk＝(x _jk-x _kmin)/(x _kmax-x _kmin)；

In formula, x _kmax, x _kminbe respectively x _1k, x _2k..., x _nkmaximal value and minimum value; X' _jkfor the data after normalization;

Step 2, set up fuzzy resembling relation matrix R={r _ij}:

$r_{ij}=\exp (-\underset{k=1}{\overset{m}{\Σ}}|x_{ik}^{,}-x_{jk}^{,}\left|\right);$

Step 3, carry out dynamic clustering:

$C_i=\frac{{\mathrm{\λ}}_{i-1}-{\mathrm{\λ}}_i}{n_i-n_{i-1}};$

In formula, i is λ polymerization order number from high to low, n _iand n _i-1be respectively the element number of i-th time and the i-th-1 time cluster; λ _iand λ _i-1be respectively confidence level when i-th time and the i-th-1 time cluster, if C _i=max (C _j), then think the confidence level λ of i-th cluster _ifor optimal threshold;

Step 4, computational prediction day and above-mentioned all kinds of Euclidean distance:

$d_{ij}=\frac{1}{\sqrt{m}}\sqrt{\underset{k=1}{\overset{m}{\Σ}}{(x_{ik}^{,}-x_{jk}^{,})}^2};$

X' in formula _ikfor predicting the characteristic index vector of day, x' _jkfor the characteristic index vector of each classification.

Finally, the classification using the shortest classification of Euclidean distance as prediction day, sets up BP neural network and predicts, obtain intelligent grid garden terminal user's cool and thermal power load data, in this, as effective support of garden energy management platform.

The foregoing is only preferred embodiment of the present invention, not in order to limit the present invention, all any amendments done within the spirit and principles in the present invention, equivalent replacement and improvement etc., all should be included within protection scope of the present invention.

Claims (5)

1. an intelligent grid garden terminal user energy demand situation dynamic prediction method, is characterized in that, described intelligent grid garden terminal user's energy demand situation dynamic prediction method comprises:

First on affecting the maximum temperature of intelligent grid garden terminal user's cool and thermal power workload demand, minimum temperature, medial temperature, rainfall amount, wind speed, intensity of solar radiation meteorologic factor carry out principal component analysis (PCA);

Secondly correlated variables is converted to the stochastic variable of minority linear independence; By weather conditions and day type quantize, carry out analysis with historical load data acquisition fuzzy clustering method and form sample;

Then the part throttle characteristics of load polytype in intelligent grid garden, polytype distributing-supplying-energy system is all embodied in load curve;

Finally according to BP neural network algorithm flow process, model is solved, obtain cool and thermal power load prediction results.

2. intelligent grid garden terminal user's energy demand situation dynamic prediction method as claimed in claim 1, it is characterized in that, described intelligent grid garden terminal user's energy demand situation dynamic prediction method specifically comprises the following steps:

The first step, uses principal component analysis (PCA) to analyze the many factors affecting load; Based on method of fuzzy cluster analysis, user type is classified; BP neural network learnt and trains, after network convergence, performance prediction being carried out to the cool and thermal power load of intelligent grid garden;

Second step, the classification using the shortest classification of Euclidean distance as prediction day, sets up BP neural network and predicts, obtain intelligent grid garden terminal user's cool and thermal power load data.

3. intelligent grid garden terminal user's energy demand situation dynamic prediction method as claimed in claim 2, is characterized in that, described utilization principal component analysis (PCA) comprises affecting the step that load many factors analyzes:

Step one, standardization is carried out to sample data:

Raw data matrix:

In formula, n is number of samples; P is each sample dimension; X _ijbe the jth dimension value of i-th sample, use X ₁, X ₂..., X _peach column vector of representing matrix X respectively, has:

$X_j^{*}=\frac{X_j-E\left(X_j\right)}{\sqrt{Vax\left(X_j\right)}};$

E (X in formula _j) and Vax (X _j) represent X respectively _javerage and variance;

Step 2, calculating correlation matrix R:

$r_{ij}=\frac{Cov(X_i,X_j)}{\sqrt{Var\left(X_i\right)Var\left(X_j\right)}}=\frac{1}{n}\underset{k=1}{\overset{n}{\mathrm{\Σ}}}{X}_{ki}^{*}{X}_{kj}^{*};$

Cov (X in formula _i, X _j) represent the i-th row and jth in data matrix arrange between covariance;

Step 3, ask orthogonal matrix and eigenwert thereof:

P ^TRP＝diag(λ ₁,λ ₂,…,λ _p)；

λ in formula ₁>=λ ₂>=...>=λ _pbe p the eigenwert of R, diag represents diagonal matrix.

4. intelligent grid garden terminal user's energy demand situation dynamic prediction method as claimed in claim 2, it is characterized in that, described fuzzy cluster analysis comprises historical load data clusters step:

Step one, sample data carried out to normalization process:

x' _jk＝(x _jk-x _kmin)/(x _kmax-x _kmin)；

In formula, x _kmax, x _kminbe respectively x _1k, x _2k..., x _nkmaximal value and minimum value; X' _jkfor the data after normalization;

Step 2, set up fuzzy resembling relation matrix R={r _ij}:

$r_{ij}=\exp \left(-\underset{k=1}{\overset{m}{\Σ}}\left|x_{ik}^{,}-x_{jk}^{,}\right|\right);$

Step 3, carry out dynamic clustering:

$C_i=\frac{{\mathrm{\λ}}_{i-1}-{\mathrm{\λ}}_i}{n_i-n_{i-1}};$

In formula, i is λ polymerization order number from high to low, n _iand n _i-1be respectively the element number of i-th time and the i-th-1 time cluster; λ _iand λ _i-1be respectively confidence level when i-th time and the i-th-1 time cluster, if C _i=max (C _j), then think the confidence level λ of i-th cluster _ifor optimal threshold;

Step 4, computational prediction day and above-mentioned all kinds of Euclidean distance:

$d_{ij}=\frac{1}{\sqrt{m}}\sqrt{\underset{k-1}{\overset{m}{\Σ}}{(x_{ik}^{,}-x_{jk}^{,})}^2};$

X' in formula _ikfor predicting the characteristic index vector of day, x' _jkfor the characteristic index vector of each classification.

5. an intelligent grid garden terminal user energy demand situation Dynamic Forecasting System, is characterized in that, described intelligent grid garden terminal user's energy demand situation Dynamic Forecasting System comprises:

Principal component analysis (PCA) module, on affecting the maximum temperature of intelligent grid garden terminal user's cool and thermal power workload demand, minimum temperature, medial temperature, rainfall amount, wind speed, intensity of solar radiation meteorologic factor carry out principal component analysis (PCA);

Analyzing samples forms module, correlated variables is converted to the stochastic variable of minority linear independence; By weather conditions and day type quantize, carry out analysis with historical load data acquisition fuzzy clustering method and form sample;

Load characteristic curve module, is all embodied in the part throttle characteristics of load polytype in intelligent grid garden, polytype distributing-supplying-energy system in load curve;

Load prediction module, solves model according to BP neural network algorithm flow process, obtains cool and thermal power load prediction results.