CN105404973A - Power transmission and transformation equipment state prediction method and system - Google Patents

Abstract

The present ivnetion relates to a power transmission and transformation equipment state prediction method and system. The method comprises: acquiring observation data of to-be-predicted equipment and constructing a characteristic sequence; acquiring a classification maximizing a probability of each characteristic vector in the characteristic sequence from a preset probability classification model; and according to the characteristic vectors and an optimal predictive coefficient corresponding to a probability classification model class of the classification maximizing the probability of each characteristic vector, performing a calculation to obtain predicted data corresponding to the to-be predicted equipment, and outputting the predicted data. The characteristic sequence is constructed according to the observation data of the to-be-predicted equipment, and the probability classification model class constructed by historical observation data of power transmission and transformation equipment is combined to obtain the corresponding optimal predictive coefficient so as to complete data prediction on the to-be-predicted equipment. According to the power transmission and transformation equipment state prediction method and system provided by the present invention, a characteristic changing trend is fully excavated from the historical data of the equipment, and the optimal predictive coefficient is adopted to predict an equipment state, so that a better prediction effect can be obtained, the method and the system can take a good effect of carrying out state monitoring, early-warning diagnosis and the like on the power transmission and transformation equipment, and reliability is improved.

Description

Power transmission and transformation equipment state Forecasting Methodology and system

Technical field

The present invention relates to electric power network technical field, particularly relate to a kind of power transmission and transformation equipment state Forecasting Methodology and system.

Background technology

Power transmission and transforming equipment is the important component part of electrical network, and the availability of power transmission and transforming equipment and stability directly have influence on the safe operation of electrical network.Along with the development of power industry, electrical network scale development on the one hand, power transmission and transforming equipment quantity is increased sharply, and user requires to improve constantly to power supply reliability; The level of informatization of equipment is more and more higher on the other hand, and equipment condition monitoring technology is increasingly mature, and equipment operating data and test data are increased sharply.Therefore, in power transmission and transforming equipment operational process, find possible fault in advance and prevented and get rid of extremely important.

Traditional power transmission and transforming equipment monitoring mode is mainly main based on scheduled overhaul.Because scheduled overhaul has blindness and mandatory, easily overhaul of the equipments that is poor to state, that there is potential faults is not enough, overequipment in good condition is overhauled, not only can not avoid the generation of operation troubles in time, and the reduction of equipment operational reliability and larger economic loss can be caused.There is the low shortcoming of reliability in traditional power transmission and transforming equipment monitoring method.

Summary of the invention

Based on this, be necessary for the problems referred to above, the power transmission and transformation equipment state Forecasting Methodology providing a kind of reliability high and system.

A kind of power transmission and transformation equipment state Forecasting Methodology, comprises the following steps:

Gather the observation data of equipment to be predicted and construction feature sequence;

According to from the probabilistic classification models class preset, obtain the classification making the maximum probability of each eigenvector in described characteristic sequence, described probabilistic classification models class is the classification obtained according to the history observation data construct of power transmission and transforming equipment;

According to each described eigenvector and the optimum prediction coefficient that makes the probabilistic classification models class of the classification of the maximum probability of eigenvector corresponding, calculate the corresponding predicted data of described equipment to be predicted and export.

A kind of power transmission and transformation equipment state prognoses system, comprising:

Data acquisition module, for gathering the observation data of equipment to be predicted and construction feature sequence;

Data processing module, for from the probabilistic classification models class preset, obtains the classification making the maximum probability of each eigenvector in described characteristic sequence, and described probabilistic classification models class is the classification obtained according to the history observation data construct of power transmission and transforming equipment;

State prediction module, for according to each described eigenvector and the optimum prediction coefficient that makes the probabilistic classification models class of the classification of the maximum probability of eigenvector corresponding, calculates the corresponding predicted data of described equipment to be predicted and exports.

Above-mentioned power transmission and transformation equipment state Forecasting Methodology and system, gather the observation data of equipment to be predicted and construction feature sequence.From the probabilistic classification models class preset, obtain the classification making the maximum probability of each eigenvector in characteristic sequence.According to eigenvector and the optimum prediction coefficient making the probabilistic classification models class of the classification of the maximum probability of eigenvector corresponding, calculate the corresponding predicted data of equipment to be predicted and export.Observation data according to equipment to be predicted sets up characteristic sequence, and obtains corresponding optimum prediction coefficient in conjunction with the probabilistic classification models class of the history observation data construct of power transmission and transforming equipment, thus completes the data prediction treating predict device.From device history data, fully excavate changing features trend, and adopt optimum prediction coefficient to carry out the state of predict device, better prediction effect can be obtained, status monitoring, early warning diagnosis etc. can be carried out to power transmission and transforming equipment and play good effect.Compared with traditional power transmission and transforming equipment monitoring mode, improve reliability.

Accompanying drawing explanation

Fig. 1 is the process flow diagram of power transmission and transformation equipment state Forecasting Methodology in an embodiment;

Fig. 2 gathers the observation data of equipment to be predicted and the process flow diagram of construction feature sequence in one embodiment;

Fig. 3 is that acquisition makes the process flow diagram of the classification of the maximum probability of each eigenvector in characteristic sequence from the probabilistic classification models class preset in an embodiment;

Fig. 4 is the process flow diagram of power transmission and transformation equipment state Forecasting Methodology in another embodiment;

Fig. 5 is the structural drawing of power transmission and transformation equipment state prognoses system in an embodiment;

Fig. 6 is the structural drawing of data acquisition module in an embodiment;

Fig. 7 is the structural drawing of data processing module in an embodiment;

Fig. 8 is the structural drawing of power transmission and transformation equipment state prognoses system in another embodiment.

Embodiment

A kind of power transmission and transformation equipment state Forecasting Methodology, as shown in Figure 1, comprises the following steps:

Step S140: gather the observation data of equipment to be predicted and construction feature sequence.Namely equipment to be predicted refer to the power transmission and transforming equipment needing to carry out status predication, can obtain data that equipment to be predicted recorded as observation data in immediate data storehouse.Wherein in an embodiment, as shown in Figure 2, step S140 comprises step S142 and step S144.

Step S142: the observation data gathering equipment to be predicted, obtains observation data sequence.Be specially:

A＝(a ₁,a ₂,…,a _p)

Wherein, A is observation data sequence, a _pfor equipment to be predicted p observation data.

Step S144: data sequence construction feature sequence according to the observation.Be specially:

A′＝(A′ ₁,A′ ₂,…,A′ _p-n)

A′ _i＝(a′ _i,a′ _i+1,…,a′ _i+n-1)

Wherein, A ' is characteristic sequence, A ' _irepresentation feature sequence A ' in i-th eigenvector, p-n is the number of eigenvector, a ' _ifor i-th observation data in observation data sequence A.

To predict transformer state, transformer to be detected have recorded 20 observation datas, needs prediction the 21st observation data, and 20 observation datas form sequence A=(a ₁, a ₂..., a ₂₀).Construction feature sequence A '=(A ' ₁, A ' ₂..., A ' ₁₇), wherein A ' _i=(a ' _i, a ' _i+1..., a ' _i+3), namely in the present embodiment, each eigenvector is made up of 4 observation datas.

Step S150: from the probabilistic classification models class preset, obtain the classification making the maximum probability of each eigenvector in characteristic sequence.

Probabilistic classification models class is the classification obtained according to the history observation data construct of power transmission and transforming equipment, obtains the classification making eigenvector maximum probability respectively according to probabilistic classification models class.Wherein in an embodiment, as shown in Figure 3, step S150 comprises step S152 to step S156.

Step S152: the probabilistic classification expression formula building each eigenvector according to the probabilistic classification models class preset.Particularly, probabilistic classification expression formula is:

p(A′ _i|λ _z)＝∫p(D′ _i|λ _z)da _i+n

Wherein, p (A ' _i| λ _z) be characteristic sequence A ' _iprobabilistic classification expression formula, p (D ' _i| λ _z) be probabilistic classification models in z probabilistic classification models class, λ _zbe the parameter of z probabilistic classification models class, a _i+nrepresent the predicted data of equipment to be predicted.For each probabilistic classification models class, by the substitution above formula of each probabilistic classification models in classification, obtain the probabilistic classification expression formula of each eigenvector.

Step S154: the posterior probability calculating each eigenvector according to probabilistic classification expression formula.Can obtain from Bayes theory, maximum a posteriori probability represents as follows:

$p({\mathrm{\λ}}_z/a_i^{\′},a_{i+1}^{\′},\mathrm{...},a_{i+n-1}^{\′})=\frac{p(a_i^{\′},a_{i+1}^{\′},\mathrm{...},a_{i+n-1}^{\′}/{\mathrm{\λ}}_z)}{p(a_i^{\′},a_{i+1}^{\′},\mathrm{...},a_{i+n-1}^{\′})}$

P (λ _z/ a ' _i, a ' _i+1..., a ' _i+n-1) represent posterior probability, p (a ' _i, a ' _i+1..., a ' _i+n-1) represent a ' _i, a ' _i+1..., a ' _i+n-1priori joint probability.Wherein

$p(a_i^{\′},a_{i+1}^{\′},\mathrm{...},a_{i+n-1}^{\′}/{\mathrm{\λ}}_z)=\underset{t=0}{\overset{n-1}{\Π}}p(a_{i+t}^{\′}/{\mathrm{\λ}}_z)$

Its logarithm is asked to above formula, can obtain:

$\log p(a_i^{\′},a_{i+1}^{\′},\mathrm{...},a_{i+n-1}^{\′}/{\mathrm{\λ}}_z)=\underset{t=0}{\overset{n-1}{\Σ}}\log p(a_{i+t}^{\′}/{\mathrm{\λ}}_z)$

Step S156: probabilistic classification models class when acquisition makes posterior probability maximum is as the classification making character pair vector maximum probability.

Due to probabilistic classification expression formula p (λ _z) prior probability unknown, therefore can suppose to concentrate probabilistic classification expression formula p (λ closing _z) possibility is equal.When obtain one determine eigenvector time, p (a ' _i, a ' _i+1..., a ' _i+n-1) can obtain being a value determined, all equal to all classifications.Therefore, ask maximum a posteriori probability be namely by try to achieve make p (a ' _i, a ' _i+1..., a ' _i+n-1| λ _z) maximum acquisition.

Distinguishing feature vector A ' _i=(a ' _i, a ' _i+1..., a ' _i+n-1) belonging to classification, namely by ask p (a ' _i, a ' _i+1..., a ' _i+n-1| λ _z) make required by maximum that corresponding model is:

z ^*＝argmaxp(a′ _i,a′ _i+1,…,a′ _i+n-1|λ _z)

Wherein, p (a ' _i, a ' _i+1..., a ' _i+n-1| λ _z) by p (a ' _i, a ' _i+1..., a ' _i+n-1| λ _z)=∫ p (D ' _i| λ _z) da _i+ncan obtain, z ^*be exactly eigenvector A ' _iaffiliated classification, just can determine the classification making this eigenvector maximum probability thus.

In the present embodiment according to preset probabilistic classification models class, calculate make p (a ' ₁₈, a ' ₁₉, a ' ₂₀| λ _z) classification of maximum probability is A ' _i=(a ' ₁₈, a ' ₁₉, a ' ₂₀) belonging to classification, and then obtain making A ' _i=(a ' ₁₈, a ' ₁₉, a ' ₂₀) classification of maximum probability.

Step S160: according to each eigenvector and the optimum prediction coefficient that makes the probabilistic classification models class of the classification of the maximum probability of eigenvector corresponding, calculate the corresponding predicted data of equipment to be predicted and export.Probabilistic classification models class all has corresponding optimum prediction coefficient, obtains the optimum prediction coefficient making the classification of the maximum probability of eigenvector, just can calculate predicted data in conjunction with eigenvector.Prediction of output data can be direct displays, also can be sent to master controller.Wherein in an embodiment, step S160 computational prediction data are specially;

$a_{i+n}=\underset{t=0}{\overset{n-1}{\Σ}}{\mathrm{\ρ}}_t*a_{i+t}$

Wherein, a _i+nfor predicted data, a _i+ti-th+t observation data in representation feature vector, ρ _tfor the optimum prediction coefficient making the probabilistic classification models class of the classification of eigenvector maximum probability corresponding.Use a _i, a _i+1..., a _i+n-1prediction a _i+n, thus try to achieve the next state parameter of equipment to be predicted.

According to the characteristic sequence A ' built in the present embodiment _i=(a ' ₁₈, a ' ₁₉, a ' ₂₀) belonging to classification obtain the optimum prediction coefficient of corresponding classification, according to optimum prediction coefficient a ' ₁₈, a ' ₁₉, a ' ₂₀prediction a ' ₂₁, obtain the next state parameter a ' of transformer ₂₁, can a ' be expressed as ₂₁=ρ ₀a ' ₁₈+ ρ ₁a ' ₁₉+ ρ ₂a ' ₂₀.

Wherein in an embodiment, as shown in Figure 4, before step S150, power transmission and transformation equipment state Forecasting Methodology also comprises step S110 to step S130.Step S110 to step S130 specifically can before step S140, also can be after step s 140.

Step S110: obtain the history observation data of power transmission and transforming equipment and build training characteristics sequence.The history observation data of power transmission and transforming equipment can directly obtain from counting database one by one, and in the present embodiment, step S110 comprises step 112 and step 114.

Step 112: the history observation data construct history observation data sequence obtaining power transmission and transforming equipment, is specially:

D＝(d ₁,d ₂,…,d _m)

Wherein, D is history observation data sequence, d _mfor power transmission and transforming equipment m history observation data.Comprise the Life cycle detection record of 3000 transformers with the historical data base of power transmission and transforming equipment, wherein every bar record comprises 100 history observation data instances.The history observation data sequence D=(d of each transformer is obtained after obtaining history observation data ₁, d ₂..., d ₁₀₀).

Step 114: obtain training characteristics sequence according to history observation data sequence.Be specially:

D′＝(D′ ₁,D′ ₂,…,D′ _m-n)

D′ _i＝(d _i,d _i+1,…,d _i+n)

Wherein, D ' is training characteristics sequence, D ' _irepresent training characteristics sequence D ' in i-th eigenvector, m-n is the number of eigenvector, d _ifor i-th history observation data in history observation data sequence D.In the present embodiment based on the history observation data observation sequence construction feature sequence D respectively of each transformer '=(D ' ₁, D ' ₂..., D ' ₉₇), wherein D ' _i=(d _i, d _i+1..., d _i+3), each eigenvector comprises 4 history observation datas.

Step S120: respectively the training characteristics vector in training characteristics sequence set up probabilistic classification models and classified, obtaining probabilistic classification models class.

The probabilistic classification models set up according to training characteristics vector in the present embodiment is mixed Gaussian probabilistic classification models, is specially:

$p\left(D_i^{\′}\right|{\mathrm{\λ}}_z)=\underset{j=1}{\overset{M}{\Σ}}{w}_j{b}_j\left(D_i^{\′}\right)$

Wherein, p (D ' _i| λ _z) represent according to training characteristics vector D ' _ithe mixed Gaussian probabilistic classification models set up, w _jfor the weights of each gaussian component, and weights need meet for each gaussian component probability density function, j=1,2 ..., M; λ _zfor the parameter of probabilistic classification models.If for the mean value vector of each gaussian component, ∑ _jfor the covariance matrix of each Gaussian function.Therefore the parameter of probabilistic classification models can be expressed as:

${\mathrm{\λ}}_z=(M,w,\stackrel{\→}{u},\Σ)$

The process of establishing of probabilistic classification models is exactly estimated parameter λ _zthe process of corresponding optimal parameter group.Optimal parameter λ _zadopt maximal possibility estimation (MaximumLikelihood, ML), namely at given eigenvector D ' _iwhen, find the parameter lambda that likelihood ratio is maximum _z.

If training set D ' _i=(d _i, d _i+1..., d _i+n), Gaussian mixture number is M, and likelihood score function can be expressed as follows:

$p(D_i^{\′}/{\mathrm{\λ}}_z)=\underset{t=1}{\overset{n}{\Π}}p(d_{i+t}/{\mathrm{\λ}}_z)$

Wherein, ${\mathrm{\λ}}_z=(M,w_{m=\mathrm{1...}M},{\stackrel{\‾}{u}}_{m=\mathrm{1...}M},{\Σ}_{m=\mathrm{1...}M}).$

In the training process, the model λ that initial is first adopted ₀, estimate with expectation maximization (ExpectationMaximization, EM) algorithm, obtain a new model parameter

$p(D_i^{\′}/\stackrel{\‾}{\mathrm{\λ}})\≥p(D_i^{\′}/\mathrm{\λ})$

And then with model parameter for initial model parameter, application EM algorithm carries out iteration according to above formula, and iteration, until meet the condition of convergence, finally obtains making likelihood function model parameter when reaching maximum be the simulated target result of probability model parameter estimation.Obtain the parameter of each probabilistic classification models, complete each training characteristics vector D ' _iestablish probabilistic classification models p (D ' _i| λ _z) after, namely to each training characteristics sequence D '=(D ' ₁, D ' ₂..., D ' _m-n) establish different probabilistic classification models.

The characteristic sequence D ' in the present embodiment, whole transformer observation sequence in database built=(D ' ₁, D ' ₂..., D ' ₉₇), set up probabilistic classification models p (D ' _i| λ _z), the wherein parameter of each probabilistic classification models employing maximal possibility estimation draws.By parameter lambda _zidentical training characteristics vector D ' _ibe divided into a class, obtain probabilistic classification models class, complete the classification to training characteristics vector.

Step S130: according to the optimum prediction coefficient of the corresponding probabilistic classification models class of training characteristics Vector operation.

For each training characteristics vector D ' in each classification _i, calculate by d _i, d _i+1..., d _i+n-1to d _i+noptimum prediction coefficient, can be write as

$d_{i+n}=\underset{t=0}{\overset{n-1}{\Σ}}{\mathrm{\ρ}}_t*d_{i+t}$

Wherein, ρ _trepresent predictive coefficient.

Predicated error determine by mean square deviation criterion:

$\mathrm{\ϵ}=E\[e^2\]=E\left\{{\[d_{i+n}-\underset{t=0}{\overset{n-1}{\Σ}}{\mathrm{\ρ}}_t*d_{i+t}\]}^2\right\}$

For making mean square deviation ε minimum, to optimum prediction coefficient ρ _task local derviation, and make it be zero.

$E\{\[d_{i+n}-\underset{t=0}{\overset{n-1}{\Σ}}{\mathrm{\ρ}}_t*d_{i+t}\]d_{i+j}\}=0,$ j＝0,1,…,n-1

The predictive coefficient ρ obtained thus _tbe optimum prediction coefficient.

For the characteristic sequence belonging to each probabilistic classification models in the present embodiment, set up by d by mean square deviation criterion by predicated error _i, d _i+1..., d _i+n-1calculate d _i+noptimum prediction coefficient.Calculate optimum prediction coefficient in conjunction with multiple characteristic sequence by mean square deviation criterion predicated error, improve and calculate accuracy.

Above-mentioned power transmission and transformation equipment state Forecasting Methodology, observation data according to equipment to be predicted sets up characteristic sequence, and obtain corresponding optimum prediction coefficient in conjunction with the probabilistic classification models class of the history observation data construct of power transmission and transforming equipment, thus complete the data prediction treating predict device.From device history data, fully excavate changing features trend, and adopt optimum prediction coefficient to carry out the state of predict device, better prediction effect can be obtained, status monitoring, early warning diagnosis etc. can be carried out to power transmission and transforming equipment and play good effect.Compared with traditional power transmission and transforming equipment monitoring mode, improve reliability.

Present invention also offers a kind of power transmission and transformation equipment state prognoses system, as shown in Figure 5, comprise data acquisition module 140, data processing module 150 and state prediction module 160.

Data acquisition module 140 is for gathering the observation data of equipment to be predicted and construction feature sequence.Namely equipment to be predicted refer to the power transmission and transforming equipment needing to carry out status predication, can obtain data that equipment to be predicted recorded as observation data in immediate data storehouse.Wherein in an embodiment, as shown in Figure 6, data acquisition module 140 comprises the first collecting unit 142 and the second collecting unit 144.

First collecting unit 142, for gathering the observation data of equipment to be predicted, obtains observation data sequence.Be specially:

A＝(a ₁,a ₂,…,a _p)

Wherein, A is observation data sequence, a _pfor equipment to be predicted p observation data.

Second collecting unit 144 is for data sequence construction feature sequence according to the observation.Be specially:

A′＝(A′ ₁,A′ ₂,…,A′ _p-n)

A′ _i＝(a′ _i,a′ _i+1,…,a′ _i+n-1)

Wherein, A ' is characteristic sequence, A ' _irepresentation feature sequence A ' in i-th eigenvector, p-n is the number of eigenvector, a ' _ifor i-th observation data in observation data sequence A.

To predict transformer state, transformer to be detected have recorded 20 observation datas, needs prediction the 21st observation data, and 20 observation datas form sequence A=(a ₁, a ₂..., a ₂₀).Construction feature sequence A '=(A ' ₁, A ' ₂..., A ' ₁₇), wherein A ' _i=(a ' _i, a ' _i+1..., a ' _i+3), namely in the present embodiment, each eigenvector is made up of 4 observation datas.

Data processing module 150, for from the probabilistic classification models class preset, obtains the classification making the maximum probability of each eigenvector in characteristic sequence.

Probabilistic classification models class is the classification obtained according to the history observation data construct of power transmission and transforming equipment, obtains the classification making eigenvector maximum probability respectively according to probabilistic classification models class.Wherein in an embodiment, as shown in Figure 7, data processing module 150 comprises the first processing unit 152, second processing unit 154 and the 3rd processing unit 156.

First processing unit 152 is for building the probabilistic classification expression formula of each eigenvector according to the probabilistic classification models class preset.Particularly, probabilistic classification expression formula is:

p(A′ _i|λ _z)＝∫p(D′ _i|λ _z)da _i+n

Wherein, p (A ' _i| λ _z) be characteristic sequence A ' _iprobabilistic classification expression formula, p (D ' _i| λ _z) be probabilistic classification models in z probabilistic classification models class, λ _zbe the parameter of z probabilistic classification models class, a _i+nrepresent the predicted data of equipment to be predicted.For each probabilistic classification models class, by the substitution above formula of each probabilistic classification models in classification, obtain the probabilistic classification expression formula of each eigenvector.

Second processing unit 154 is for calculating the posterior probability of each eigenvector according to probabilistic classification expression formula.Can obtain from Bayes theory, maximum a posteriori probability represents as follows:

$p({\mathrm{\λ}}_z/a_i^{\′},a_{i+1}^{\′},\mathrm{...},a_{i+n-1}^{\′})=\frac{p(a_i^{\′},a_{i+1}^{\′},\mathrm{...},a_{i+n-1}^{\′}/{\mathrm{\λ}}_z)}{p(a_i^{\′},a_{i+1}^{\′},\mathrm{...},a_{i+n-1}^{\′})}$

P (λ _z/ a ' _i, a ' _i+1..., a ' _i+n-1) represent posterior probability, p (a ' _i, a ' _i+1..., a ' _i+n-1) represent a ' _i, a ' _i+1..., a ' _i+n-1priori joint probability.Wherein

$p(a_i^{\′},a_{i+1}^{\′},\mathrm{...},a_{i+n-1}^{\′}/{\mathrm{\λ}}_z)=\underset{t=0}{\overset{n-1}{\Π}}p(a_{i+t}^{\′}/{\mathrm{\λ}}_z)$

Its logarithm is asked to above formula, can obtain:

$\log p(a_i^{\′},a_{i+1}^{\′},\mathrm{...},a_{i+n-1}^{\′}/{\mathrm{\λ}}_z)=\underset{t=0}{\overset{n-1}{\Σ}}\log p(a_{i+t}^{\′}/{\mathrm{\λ}}_z)$

3rd processing unit 156 is for obtaining probabilistic classification models class when making posterior probability maximum as the classification making character pair vector maximum probability.

Due to probabilistic classification expression formula p (λ _z) prior probability unknown, therefore can suppose to concentrate probabilistic classification expression formula p (λ closing _z) possibility is equal.When obtain one determine eigenvector time, p (a ' _i, a ' _i+1..., a ' _i+n-1) can obtain being a value determined, all equal to all classifications.Therefore, ask maximum a posteriori probability be namely by try to achieve make p (a ' _i, a ' _i+1..., a ' _i+n-1| λ _z) maximum acquisition.

Distinguishing feature vector A ' _i=(a ' _i, a ' _i+1..., a ' _i+n-1) belonging to classification, namely by ask p (a ' _i, a ' _i+1..., a ' _i+n-1| λ _z) make required by maximum that corresponding model is:

z ^*＝argmaxp(a′ _i,a′ _i+1,…,a′ _i+n-1|λ _z)

Wherein, p (a ' _i, a ' _i+1..., a ' _i+n-1| λ _z) by p (a ' _i, a ' _i+1..., a ' _i+n-1| λ _z)=∫ p (D ' _i| λ _z) da _i+ncan obtain, z ^*be exactly eigenvector A ' _iaffiliated classification, just can determine the classification making this eigenvector maximum probability thus.

In the present embodiment according to preset probabilistic classification models class, calculate make p (a ' ₁₈, a ' ₁₉, a ' ₂₀| λ _z) classification of maximum probability is A ' _i=(a ' ₁₈, a ' ₁₉, a ' ₂₀) belonging to classification, and then obtain making A ' _i=(a ' ₁₈, a ' ₁₉, a ' ₂₀) classification of maximum probability.

State prediction module 160, for according to each eigenvector and the optimum prediction coefficient that makes the probabilistic classification models class of the classification of the maximum probability of eigenvector corresponding, calculates the corresponding predicted data of equipment to be predicted and exports.

Probabilistic classification models class all has corresponding optimum prediction coefficient, obtains the optimum prediction coefficient making the classification of the maximum probability of eigenvector, just can calculate predicted data in conjunction with eigenvector.Prediction of output data can be direct displays, also can be sent to master controller.Wherein in an embodiment, state prediction module 160, according to each eigenvector and the optimum prediction coefficient that makes the probabilistic classification models class of the classification of the maximum probability of eigenvector corresponding, calculates the corresponding predicted data of equipment to be predicted, is specially:

$a_{i+n}=\underset{t=0}{\overset{n-1}{\Σ}}{\mathrm{\ρ}}_t*a_{i+t}$

Wherein, a _i+nfor predicted data, a _i+ti-th+t observation data in representation feature vector, ρ _tfor the optimum prediction coefficient making the probabilistic classification models class of the classification of eigenvector maximum probability corresponding.

Wherein in an embodiment, as shown in Figure 8, power transmission and transformation equipment state prognoses system also can comprise data acquisition module 110, classification based training module 120 and coefficients calculation block 130.

Data acquisition module 110 for data processing module 150 from preset probabilistic classification models class, obtain make the classification of the maximum probability of each eigenvector in characteristic sequence before, obtain power transmission and transforming equipment history observation data and build training characteristics sequence.The history observation data of power transmission and transforming equipment can directly obtain from counting database one by one, data acquisition module 110 build the detailed process of training characteristics sequence and step S110 similar, do not repeat them here.

Classification based training module 120, for setting up probabilistic classification models to the training characteristics vector in training characteristics sequence and classify respectively, obtains probabilistic classification models class.

The probabilistic classification models set up according to training characteristics vector in the present embodiment is the parameter that mixed Gaussian probabilistic classification models obtains each probabilistic classification models, completes each training characteristics vector D ' _iestablish probabilistic classification models p (D ' _i| λ _z), the parameter of each probabilistic classification models employing maximal possibility estimation draws.By parameter lambda _zidentical training characteristics vector D ' _ibe divided into a class, obtain probabilistic classification models class, complete the classification to training characteristics vector.Classification based training module 120 divide obtain the detailed process of probabilistic classification models class and step S120 similar, do not repeat them here.

Coefficients calculation block 130 is for the optimum prediction coefficient according to the corresponding probabilistic classification models class of training characteristics Vector operation.For the characteristic sequence belonging to each probabilistic classification models, set up by d by mean square deviation criterion by predicated error _i, d _i+1..., d _i+n-1calculate d _i+noptimum prediction coefficient.Coefficients calculation block 130 calculate the detailed process of the optimum prediction coefficient of corresponding probabilistic classification models class and step S130 similar, do not repeat them here.Calculate optimum prediction coefficient in conjunction with multiple characteristic sequence by mean square deviation criterion predicated error, improve and calculate accuracy.

Above-mentioned power transmission and transformation equipment state prognoses system, observation data according to equipment to be predicted sets up characteristic sequence, and obtain corresponding optimum prediction coefficient in conjunction with the probabilistic classification models class of the history observation data construct of power transmission and transforming equipment, thus complete the data prediction treating predict device.Fully changing features trend is excavated from device history data, and adopt optimum prediction coefficient to carry out the state of predict device, better prediction effect can be obtained, status monitoring, early warning diagnosis etc. can be carried out to power transmission and transforming equipment and play good effect, improve reliability.

Each technical characteristic of the above embodiment can combine arbitrarily, for making description succinct, the all possible combination of each technical characteristic in above-described embodiment is not all described, but, as long as the combination of these technical characteristics does not exist contradiction, be all considered to be the scope that this instructions is recorded.

The above embodiment only have expressed several embodiment of the present invention, and it describes comparatively concrete and detailed, but can not therefore be construed as limiting the scope of the patent.It should be pointed out that for the person of ordinary skill of the art, without departing from the inventive concept of the premise, can also make some distortion and improvement, these all belong to protection scope of the present invention.Therefore, the protection domain of patent of the present invention should be as the criterion with claims.

Claims (10)

1. a power transmission and transformation equipment state Forecasting Methodology, is characterized in that, comprises the following steps:

Gather the observation data of equipment to be predicted and construction feature sequence;

From the probabilistic classification models class preset, obtain the classification making the maximum probability of each eigenvector in described characteristic sequence, described probabilistic classification models class is the classification obtained according to the history observation data construct of power transmission and transforming equipment;

According to each described eigenvector and the optimum prediction coefficient that makes the probabilistic classification models class of the classification of the maximum probability of eigenvector corresponding, calculate the corresponding predicted data of described equipment to be predicted and export.

2. power transmission and transformation equipment state Forecasting Methodology according to claim 1, is characterized in that, the observation data of described collection equipment to be predicted, and the step of observation data construction feature sequence according to described equipment to be predicted, comprise the following steps:

Gather the observation data of described equipment to be predicted, obtain observation data sequence, be specially:

A＝(a ₁,a ₂,…,a _p)

Wherein, A is observation data sequence, a _pfor equipment to be predicted p observation data;

According to described observed data sequence construct characteristic sequence, be specially:

A′＝(A′ ₁,A′ ₂,…,A′ _p-n)

A′ _i＝(a′ _i,a′ _i+1,…,a′ _i+n-1)

Wherein, A ' is characteristic sequence, A ' _irepresentation feature sequence A ' in i-th eigenvector, p-n is the number of eigenvector, a ' _ifor i-th observation data in observation data sequence A.

3. power transmission and transformation equipment state Forecasting Methodology according to claim 1, is characterized in that, the described probabilistic classification models class from presetting, obtaining the step making the classification of the maximum probability of each eigenvector in described characteristic sequence, comprising the following steps:

The probabilistic classification expression formula of each described eigenvector is built according to the probabilistic classification models class preset;

The posterior probability of each described eigenvector is calculated according to described probabilistic classification expression formula;

Probabilistic classification models class when acquisition makes described posterior probability maximum is as the classification making character pair vector maximum probability.

4. power transmission and transformation equipment state Forecasting Methodology according to claim 1, it is characterized in that, described according to each eigenvector and the optimum prediction coefficient that makes the probabilistic classification models class of the classification of the maximum probability of eigenvector corresponding, calculate the corresponding predicted data of described equipment to be predicted, be specially:

$a_{i+n}=\underset{t=0}{\overset{n-1}{\Σ}}{\mathrm{\ρ}}_t*a_{i+t}$

Wherein, a _i+nfor predicted data, a _i+ti-th+t observation data in representation feature vector, ρ _tfor the optimum prediction coefficient making the probabilistic classification models class of the classification of eigenvector maximum probability corresponding.

5. power transmission and transformation equipment state Forecasting Methodology according to claim 1, is characterized in that, the described probabilistic classification models class from presetting, obtains before making the step of the classification of the maximum probability of each eigenvector in described characteristic sequence, further comprising the steps of:

Obtain the history observation data of power transmission and transforming equipment and build training characteristics sequence;

Respectively the training characteristics vector in described training characteristics sequence set up probabilistic classification models and classified, obtaining probabilistic classification models class;

According to the optimum prediction coefficient of the corresponding probabilistic classification models class of described training characteristics Vector operation.

6. a power transmission and transformation equipment state prognoses system, is characterized in that, comprising:

Data acquisition module, for gathering the observation data of equipment to be predicted and construction feature sequence;

Data processing module, for from the probabilistic classification models class preset, obtains the classification making the maximum probability of each eigenvector in described characteristic sequence, and described probabilistic classification models class is the classification obtained according to the history observation data construct of power transmission and transforming equipment;

State prediction module, for according to each described eigenvector and the optimum prediction coefficient that makes the probabilistic classification models class of the classification of the maximum probability of eigenvector corresponding, calculates the corresponding predicted data of described equipment to be predicted and exports.

7. power transmission and transformation equipment state prognoses system according to claim 6, is characterized in that, described data acquisition module comprises:

First collecting unit, for gathering the observation data of described equipment to be predicted, obtaining observation data sequence, being specially:

A＝(a ₁,a ₂,…,a _p)

Wherein, A is observation data sequence, a _pfor equipment to be predicted p observation data;

Second collecting unit, for according to described observed data sequence construct characteristic sequence, is specially:

A′＝(A′ ₁,A′ ₂,…,A′ _p-n)

A′ _i＝(a′ _i,a′ _i+1,…,a′ _i+n-1)

Wherein, A ' is characteristic sequence, A ' _irepresentation feature sequence A ' in i-th eigenvector, p-n is the number of eigenvector, a ' _ifor i-th observation data in observation data sequence A.

8. power transmission and transformation equipment state prognoses system according to claim 6, is characterized in that, described data processing module comprises:

First processing unit, for building the probabilistic classification expression formula of each described eigenvector according to the probabilistic classification models class preset;

Second processing unit, for calculating the posterior probability of each described eigenvector according to described probabilistic classification expression formula;

3rd processing unit, for obtaining probabilistic classification models class when making described posterior probability maximum as the classification making character pair vector maximum probability.

9. power transmission and transformation equipment state prognoses system according to claim 6, it is characterized in that, described state prediction module is according to each described eigenvector and the optimum prediction coefficient that makes the probabilistic classification models class of the classification of the maximum probability of eigenvector corresponding, calculate the corresponding predicted data of described equipment to be predicted, be specially:

$a_{i+n}=\underset{t=0}{\overset{n-1}{\Σ}}{\mathrm{\ρ}}_t*a_{i+t}$

Wherein, a _i+nfor predicted data, a _i+ti-th+t observation data in representation feature vector, ρ _tfor the optimum prediction coefficient making the probabilistic classification models class of the classification of eigenvector maximum probability corresponding.

10. power transmission and transformation equipment state prognoses system according to claim 6, is characterized in that, also comprise:

Data acquisition module, for described data processing module from preset probabilistic classification models class, obtain make the classification of the maximum probability of each eigenvector in described characteristic sequence before, obtain power transmission and transforming equipment history observation data and build training characteristics sequence;

Classification based training module, for setting up probabilistic classification models to the training characteristics vector in described training characteristics sequence and classify respectively, obtains probabilistic classification models class;

Coefficients calculation block, for the optimum prediction coefficient according to the corresponding probabilistic classification models class of described training characteristics Vector operation.