CN103123665A - Short-term power load forecasting method based on fuzzy clustering similar day - Google Patents

Abstract

The invention discloses a short-term power load forecasting method based on a fuzzy clustering similar day. The method includes that firstly meteorological factors are divided into temperature, pressure, wind speed, rain and other occasions and then constitute influence factors of the similar day with week styles and date styles, a fuzzy coefficient characteristic mapping table is built through fuzzy rules, clarification is carried out by the method of fuzzy clustering based on the preceding steps, the similar day is chosen according to clustering levels, according to obtained load data of the similar day, load sequences are projected to different scales and low frequency components are obtained by means of wavelet transformation, a support vector machine is optimized by means of a particle swarm optimization algorithm to achieve forecasting for a short-term power load low frequency portion, and forecasting for a high frequency portion is achieved by the method of weighted average. Eventually, application researches are carried out by means of the load data of a power grid in Shanghai city, and good forecasting effects can be achieved in weekdays, at weekends and in holidays.

Description

Based on the fuzzy clustering Short-Term Load Forecasting Method of similar day

Technical field

The present invention relates to a kind of Techniques for Prediction of Electric Loads field, particularly a kind of based on the fuzzy clustering Short-Term Load Forecasting Method of similar day.

Background technology

Load Prediction In Power Systems is the important content of energy management system.By accurate load prediction, can arrange economically Unit Commitment, reduce spinning reserve capacity, and then reduce cost of electricity-generating, increase economic efficiency.Therefore, seek effective load forecasting method, the accuracy that raising predicts the outcome is significant.

In recent ten years, numerous experts and scholars have proposed many Forecasting Methodologies based on nonlinear theory and combination thereof for the characteristics of load prediction.Wherein, artificial neural network (ANN) obtains a wide range of applications in load forecast owing to having the character such as powerful non-linear mapping capability, self-adaptation, self study, fault-tolerance and parallel processing, has obtained a large amount of achievements.Yet artificial neural network especially BP neural network has over-fitting, easily is absorbed in the shortcomings such as local extremum, has affected effect.Support vector machine (SVM) can solve the practical problemss such as small sample, non-linear, high dimension drawn game section minimal point preferably as a kind of new machine learning algorithm, is more and more used in load forecast.But the precision of prediction of SVM depends on choosing of its error penalty parameter c and kernel functional parameter g, adopts particle swarm optimization algorithm (PSO) optimization to choose above-mentioned parameter for this reason.The singularity of considering simultaneously electric load has proposed a kind of short-term load forecasting method of similar day of choosing based on fuzzy clustering.

Summary of the invention

In order to overcome the defective of prior art, the invention discloses a kind ofly based on the similar day Short-Term Load Forecasting Method of fuzzy clustering, it is simple in structure, calculated amount is little, it is higher to operate.

The invention discloses following technical scheme:

A kind of based on the fuzzy clustering Short-Term Load Forecasting Method of similar day, the method comprises the following steps: S1: fuzzy birdsing of the same feather flock together sought similar day; S2: the load data of similar day; S3: wavelet decomposition; S4: the decomposition of low frequency part and HFS; S5: load prediction value.

Preferably, described Short-Term Load Forecasting Method, its S1 is specially: S1.1: for the uncertain factor of weather category, meteorologic factor is subdivided into the situations such as temperature, air pressure, wind speed, wet weather, with week type, date type consist of together the influence factor of similar day; S1.2: set up fuzzy coefficient Feature Mapping table by fuzzy rule, realize the quantification of influence factor.

Preferably, described Short-Term Load Forecasting Method, its S2 adopt fuzzy clustering method to classify, and choose similar day according to the cluster level.

Preferably, described Short-Term Load Forecasting Method, its S3 utilizes wavelet transformation to obtain its low frequency component for for containing the problem of non-Gaussian noise in the similar daily load data of obtaining, and has effectively eliminated the impact of non-Gaussian noise.

Preferably, described Short-Term Load Forecasting Method, its S4 is specially: S4.1: adopt the particle swarm optimization algorithm Support Vector Machines Optimized, obtain optimum error punishment parameter and kernel functional parameter; S4.2: choosing similar day, after simultaneously similar day data having been carried out the denoising operation, the support vector machine after the optimization of employing particle swarm optimization algorithm realizes the prediction of short-term electric load low frequency part; S4.3: adopt weighted average method to realize the prediction of HFS.

Preferably, described Short-Term Load Forecasting Method, its S5 by the method in described S1-S4 can predict accurately at ordinary times, the load datas of working day and festivals or holidays.

Compared with prior art, beneficial effect of the present invention is as follows:

1, set up Short-term Load Forecasting Model;

2, the precision of prediction of the method is higher;

3, improved predetermined speed of common prediction method;

4, simple in structure, calculated amount is little, operability is higher.

Description of drawings

Fig. 1 is the structural representation of specific embodiment of the invention Short-Term Load Forecasting Method;

Fig. 2 is the schematic diagram data of specific embodiment of the invention load prediction.

Embodiment:

The present invention will be further described with specific embodiment by reference to the accompanying drawings in the below:

Embodiment

Be process flow diagram of the present invention as Fig. 1, the steps include:

S1: fuzzy clustering is sought similar day; Described S1 is specially:

S1.1: for the uncertain factor of weather category, meteorologic factor is subdivided into the situations such as temperature, air pressure, wind speed, wet weather, with week type, date type consist of together the influence factor of similar day;

S1.2: set up fuzzy coefficient Feature Mapping table by fuzzy rule, realize the quantification of influence factor; If U=[x ₁, x ₂..., x _n] for predicting n the sample set of day, each sample x _jM characteristic index arranged, i.e. sample x _jCan be expressed as x _j=[x _j1, x _j2..., x _jm] ^T, (j=1,2 ..., n).Set up the fuzzy resembling relation matrix D.D=[d _ij] be used for representing similar matrix.In order to determine fuzzy similarity matrix, adopt correlation coefficient process, as follows:

S2: the load data of similar day; It is specially and adopts fuzzy clustering method to classify, and chooses similar day according to the cluster level; After classifying with fuzzy clustering, introduce not on the same day between the concept of " similarity ".Note X _i=[x _i1, x _i2..., x _im] be the value (supposing total m characteristic quantity) of all characteristic quantities of i day, can calculate the similarity of two days by following formula.This similarity is the included angle cosine between two vectors in m-dimensional space.It describes the degree of closeness of characteristic quantity between two days.

$r_{\mathrm{ij}}=\frac{\underset{k=1}{\overset{m}{\mathrm{\Σ}}}{x}_{\mathrm{ik}}{x}_{\mathrm{jk}}}{\sqrt{\underset{k=1}{\overset{m}{\mathrm{\Σ}}}{x^2}_{\mathrm{ik}}\underset{k=1}{\overset{m}{\mathrm{\Σ}}}{x^2}_{\mathrm{jk}}}}$

S3: wavelet decomposition; It is specially: for containing the problem of non-Gaussian noise in the similar daily load data of obtaining, utilize wavelet transformation to obtain its low frequency component, effectively eliminated the impact of non-Gaussian noise; If function

Be a quadractically integrable function, namely

Claim

Be wavelet function.If its Fourier transform

Satisfy the admissibility condition:

Claim To allow small echo or be called female small echo.

S4: the decomposition of low frequency part and HFS; It is specially:

S4.1: adopt the particle swarm optimization algorithm Support Vector Machines Optimized, obtain optimum error penalty parameter c and kernel functional parameter g;

S4.1.1: read sample number and sample data carried out pre-service, produce at random one group c, g} is as the initial position of particle;

S4.1.2: { c, g} carry out the SVM training, and the square error of calculation training sample is as validation error according to current;

S4.1.3: as adaptive value, and the position of memory individual and group corresponding optimal adaptation value is p with validation error _BestAnd g _Best, { c, g} better according to the search of PSO optimization method;

S4.1.4: repeat S4.1.2 and S4.1.3, until satisfy maximum iterations or satisfy termination condition.

S4.2: choosing similar day, after simultaneously similar day data having been carried out the denoising operation, the support vector machine after the optimization of employing particle swarm optimization algorithm realizes the prediction of short-term electric load low frequency part;

S4.3: adopt weighted average method to realize the prediction of HFS;

S5: the load prediction value, it is specially:

S5.1 is applied to the method the load data of Shanghai City electrical network, checking the method daily load data at ordinary times that can calculate to a nicety;

S5.2 is applied to the method the load data of Shanghai City electrical network, checking the method load data at weekend that can calculate to a nicety;

S5.3 is applied to the method the load data of Shanghai City electrical network, checking the method load data festivals or holidays that can calculate to a nicety;

As Fig. 2, the method for using this paper to carry adopts somewhere, Shanghai Electric Power Co data to carry out load prediction research.Get the data on August 5th, 1 day 1 June in 2011 as the sample data collection.Predict three balance time of on July 22,20 days to 2011 July in 2011, on July 23rd, 2011 and Saturday July 24, Sunday and prediction holiday on National Day on October 4th, 2011.Fig. 2 provides and adopts the error information contrast to predicting day, weekend, festivals or holidays etc. at ordinary times respectively of this paper method.

The preferred embodiment of the present invention just is used for helping to set forth the present invention.Preferred embodiment does not have all details of detailed descriptionthe, does not limit this invention yet and only is described embodiment.Obviously, according to the content of this instructions, can make many modifications and variations.These embodiment are chosen and specifically described to this instructions, is in order to explain better principle of the present invention and practical application, thereby under making, the technical field technician can utilize the present invention well.The present invention only is subjected to the restriction of claims and four corner and equivalent.

Claims (6)

1. one kind based on the similar day Short-Term Load Forecasting Method of fuzzy clustering, it is characterized in that, the method comprises the following steps:

S1: fuzzy clustering is sought similar day;

S2: the load data of similar day;

S3: wavelet decomposition;

S4: the decomposition of low frequency part and HFS;

S5: load prediction value.

2. Short-Term Load Forecasting Method according to claim 1, is characterized in that, described S1 is specially:

S1.1: for the uncertain factor of weather category, meteorologic factor is subdivided into the situations such as temperature, air pressure, wind speed, wet weather, with week type, date type consist of together the influence factor of similar day;

S1.2: set up fuzzy coefficient Feature Mapping table by fuzzy rule, realize the quantification of influence factor.

3. Short-Term Load Forecasting Method according to claim 1, is characterized in that, described S2 adopts fuzzy clustering method to classify, and chooses similar day according to the cluster level.

4. Short-Term Load Forecasting Method according to claim 1, it is characterized in that, described S3 utilizes wavelet transformation to obtain its low frequency component for for containing the problem of non-Gaussian noise in the similar daily load data of obtaining, and has effectively eliminated the impact of non-Gaussian noise.

5. Short-Term Load Forecasting Method according to claim 1, is characterized in that, described S4 is specially:

S4.1: adopt the particle swarm optimization algorithm Support Vector Machines Optimized, obtain optimum error punishment parameter and kernel functional parameter;

S4.2: choosing similar day, after simultaneously similar day data having been carried out the denoising operation, the support vector machine after the optimization of employing particle swarm optimization algorithm realizes the prediction of short-term electric load low frequency part;

S4.3: adopt weighted average method to realize the prediction of HFS.

6. Short-Term Load Forecasting Method according to claim 1, is characterized in that, described S5 by the method in described S1-S4 can predict accurately at ordinary times, the load datas of working day and festivals or holidays.

Images