CN106557833A - A kind of smooth storing cogeneration system dynamics response performance index forecasting method - Google Patents

Abstract

The invention provides a kind of smooth storing cogeneration system dynamics response performance index forecasting method, by the time serieses for setting up light storing cogeneration system dynamics response performance index Evolution System, Bayes's feed forward-fuzzy control process is carried out to time serieses measurement data, and then light storing cogeneration system dynamics response performance exponential forecasting is calculated based on Markov chain Monte Carlo, obtain light storing cogeneration system dynamics response performance exponential forecasting value.The method can be predicted calculating to light storing cogeneration system dynamics response performance index according to monitoring parameter, in real time light storing cogeneration system and power distribution network are controlled according to result of calculation, the problems such as distribution network system can be prevented effectively from the power that brings being accessed because light is stored up and mismatch, significantly improve power distribution network power system and store up the reliability after association system is accessed and economy in light.

Description

A kind of smooth storing cogeneration system dynamics response performance index forecasting method

Technical field

The invention belongs to technical field of photovoltaic power generation, more particularly to a kind of smooth storing cogeneration system dynamics response performance refers to Number Forecasting Methodology.

Background technology

In power distribution network power system, distributed photovoltaic power generation equipment and energy storage device constitute a complicated system how Light storing cogeneration system dynamics response performance exponential forecasting is carried out according to distributed light-preserved system and power distribution network operation characteristic to comment Estimate, enable each light storing cogeneration system and its power distribution network that accessed is safe and stable, Effec-tive Function, conventional power distribution network The characteristics of grid entry point power-balance index light storing cogeneration system dynamics response performance index calculation method is to ignore distributed Interaction relationship between photovoltaic and photovoltaic energy storage and power distribution network, by each system in regional power grid or light storing cogeneration system Independently carry out the Dynamic Response, it is impossible to effectively utilizes electrical network and distributed photovoltaic power generation service data resource, assess accuracy It is not high with photovoltaic utilization ratio.

In view of this, the present invention provides a kind of smooth storing cogeneration system dynamics response performance index forecasting method, with full Sufficient practical application needs.

The content of the invention

The purpose of the present invention is：To overcome the deficiencies in the prior art, the present invention to provide a kind of smooth storing cogeneration system and move State response performance index forecasting method, so as to obtain light storing cogeneration system dynamics response performance index.

The technical solution adopted in the present invention is：A kind of smooth storing cogeneration system dynamics response performance exponential forecasting side Method, it is characterised in that comprise the steps：

Step 1：Set up the time serieses of light storing cogeneration system dynamics response performance index Evolution System：

In Fixed Time Interval to electricity generation system grid entry point active power, voltage, voltage change ratio, temperature, intensity of illumination Measure, and be defined as follows light storing cogeneration system dynamics response performance index, i.e.,：

Then, in a series of moment tdt₁,tdt_2,...,tdt_n, n is natural number, n=1,2 ..., obtain grid entry point wattful power Rate pdt, voltage vdt, voltage change ratio vddt, temperature Tdt, the measurement data of intensity of illumination sdt：

Step 2：Bayes's feed forward-fuzzy control is set up：

Step 2.1：Set up Bayes's feedforward artificial neuron model：

The measurement data for obtaining is measured with step 1 as the feedforward artificial neuron model for being input into training data is：

y_dt=f (dtx, θ_w) (2)

In formula, θ_wFor the weighting parameter set of hidden layer, output layer in feed forward-fuzzy control；Dtx is by measurement data The training sample data of composition；f(dtx,θ_w) represent many perceptron functional equations, Gaussian distributed, y_dtStore up for light to be asked Combined generating system dynamic response performance index；

Step 2.2：The determination of weighting parameter Bayes posterior probability distribution：

The sample data information that measurement is obtained is combined, the posteriority point of parameter can be obtained according to bayes rule Cloth：

p(θ_w| dtx)～L (θ_w|dtx)p(θ_w) (3)

In formula, p (θ_w| dtx) for θ_wAll of parameter and hyper parameter in the probability under the conditions of the dtx of sample, L (θ_w| Dtx it is) θ_wLikelihood function under the conditions of the dtx of sample, p (θ_w) for θ_wProbability；

Step 2.3：The determination of expectation function：

According to the optimal estimation principle of model prediction mean square error, the mathematic expectaion of posteriority prediction distribution is by being calculated as below Formula is obtained：

Step 3：It is pre- to light storing cogeneration system dynamics response performance index based on Markov chain Monte Carlo Measured value is asked for：

Integration in formula (4) is function y_dt=f (dtx, θ_w) with regard to parameter Posterior distrbutionp p (θ_w| dtx) expectation, this Individual mathematic expectaion can be with Markov chain Monte Carlo come approximately, formula is as follows：

Wherein, n₁For some the initial Markov chains being rejected, n₂To carry out the weight vector sample of self-balancing Posterior distrbutionp Number, θ_sFor parameter θ_wThe sampled value of Posterior distrbutionp, solvesAs light storing cogeneration system dynamics response performance exponential forecasting Value.

The invention has the beneficial effects as follows：The present invention provides a kind of smooth storing cogeneration system dynamics response for photovoltaic electrical network Performance index Forecasting Methodology, carries out real-time monitoring to light-preserved system operational factor and environment parament, and according to monitoring parameter Calculating is predicted to light storing cogeneration system dynamics response performance index, in real time light storage joint is sent out according to result of calculation Electric system and power distribution network are controlled, and can be prevented effectively from distribution network system and ask because light stores up power mismatch that access brings etc. Topic, significantly improves power distribution network power system and stores up the reliability after association system is accessed and economy in light.

Description of the drawings

Object function interative computation figures of the Fig. 1 for the embodiment of the present invention.

Specific embodiment

For a better understanding of the present invention, present disclosure is further elucidated with reference to embodiment, but the present invention Content is not limited solely to the following examples.Those skilled in the art can be made various changes or modifications to the present invention, these The equivalent form of value is equally within the scope of claims listed by the application are limited.

As shown in figure 1, a kind of smooth storing cogeneration system dynamics response performance exponential forecasting provided in an embodiment of the present invention Method, comprises the steps：

Step 1：Set up the time serieses of light storing cogeneration system dynamics response performance index Evolution System：

In Fixed Time Interval to electricity generation system grid entry point active power, voltage, voltage change ratio, temperature, intensity of illumination Measure, and be defined as follows light storing cogeneration system dynamics response performance index, i.e.,：

Then, in a series of moment tdt₁,tdt_2,...,tdt_n, n is natural number, n=1,2 ..., obtain grid entry point wattful power Rate pdt, voltage vdt, voltage change ratio vddt, temperature Tdt, the measurement data of intensity of illumination sdt：

Step 2：Bayes's feed forward-fuzzy control is set up：

Step 2.1：Bayes's feedforward artificial neuron model is set up, transmission function f is Sigmoid functions：

The measurement data for obtaining is measured with step 1 as the feedforward artificial neuron model for being input into training data is：

y_dt=f (dtx, θ_w) (2)

In formula, θ_wFor the weighting parameter set of hidden layer, output layer in feed forward-fuzzy control；Dtx is by measurement data The training sample data of composition；f(dtx,θ_w) represent many perceptron functional equations, Gaussian distributed, y_dtStore up for light to be asked Combined generating system dynamic response performance index.

Step 2.2：The determination of weighting parameter Bayes posterior probability distribution：

The sample data information that measurement is obtained is combined, the posteriority point of parameter can be obtained according to bayes rule Cloth：

p(θ_w| dtx)～L (θ_w|dtx)p(θ_w) (3)

In formula, p (θ_w| dtx) for θ_wAll of parameter and hyper parameter in the probability under the conditions of the dtx of sample, L (θ_w| Dtx it is) θ_wLikelihood function under the conditions of the dtx of sample, p (θ_w) for θ_wProbability.

Step 2.3：The determination of expectation function：

According to the optimal estimation principle of model prediction mean square error, the mathematic expectaion of posteriority prediction distribution is by being calculated as below Formula is obtained：

Step 3：It is pre- to light storing cogeneration system dynamics response performance index based on Markov chain Monte Carlo Measured value is asked for：

Integration in formula (4) is function y_dt=f (dtx, θ_w) with regard to parameter Posterior distrbutionp p (θ_w| dtx) expectation, this Individual mathematic expectaion can be with Markov chain Monte Carlo come approximately, formula is as follows：

Wherein, n₁For some the initial Markov chains being rejected, n₂To carry out the weight vector sample of self-balancing Posterior distrbutionp Number, θ_sFor parameter θ_wThe sampled value of Posterior distrbutionp, in the present embodiment, n₁=10000, n₂=40000, solveAs light storage joins Close electricity generation system dynamic response performance exponential forecasting value.

These are only embodiments of the invention, be not limited to the present invention, therefore, it is all the present invention spirit and Within principle, any modification, equivalent substitution and improvements done etc. are should be included within scope of the presently claimed invention.

Claims (1)

1. a kind of smooth storing cogeneration system dynamics response performance index forecasting method, it is characterised in that comprise the steps：

Step 1：Set up the time serieses of light storing cogeneration system dynamics response performance index Evolution System：

Electricity generation system grid entry point active power, voltage, voltage change ratio, temperature, intensity of illumination are carried out in Fixed Time Interval Measurement, and be defined as follows light storing cogeneration system dynamics response performance index, i.e.,：

Then, in a series of moment tdt₁,tdt₂,...,tdt_n, n is natural number, n=1,2 ..., obtain grid entry point active power Pdt, voltage vdt, voltage change ratio vddt, temperature Tdt, the measurement data of intensity of illumination sdt：

$\left\{\begin{array}{c}{\mathrm{pdt}}_1,{\mathrm{pdt}}_2,...,{\mathrm{pdt}}_n\\ {\mathrm{vdt}}_1,{\mathrm{vdt}}_2,...,{\mathrm{vdt}}_n\\ {\mathrm{vddt}}_1,{\mathrm{vddt}}_2,...,{\mathrm{vddt}}_n\\ {\mathrm{Tdt}}_1,{\mathrm{Tdt}}_2,...,{\mathrm{Tdt}}_n\\ {\mathrm{sdt}}_1,{\mathrm{sdt}}_2,...,{\mathrm{sdt}}_n\end{array}\right.---\left(1\right)$

Step 2：Bayes's feed forward-fuzzy control is set up：

Step 2.1：Set up Bayes's feedforward artificial neuron model：

The measurement data for obtaining is measured with step 1 as the feedforward artificial neuron model for being input into training data is：

y_dt=f (dtx, θ_w) (2)

In formula, θ_wFor the weighting parameter set of hidden layer, output layer in feed forward-fuzzy control；Dtx is made up of measurement data Training sample data；f(dtx,θ_w) represent many perceptron functional equations, Gaussian distributed, y_dtSend out for light storage joint to be asked Electric system dynamic response performance index；

Step 2.2：The determination of weighting parameter Bayes posterior probability distribution：

The sample data information that measurement is obtained is combined, the Posterior distrbutionp of parameter can be obtained according to bayes rule：

p(θ_w| dtx)～L (θ_w|dtx)p(θ_w) (3)

In formula, p (θ_w| dtx) for θ_wAll of parameter and hyper parameter in the probability under the conditions of the dtx of sample, L (θ_w| dtx) for θ_w Likelihood function under the conditions of the dtx of sample, p (θ_w) for θ_wProbability；

Step 2.3：The determination of expectation function：

According to the optimal estimation principle of model prediction mean square error, the mathematic expectaion of posteriority prediction distribution is by being calculated as below formula Obtain：

$\hat{y}=\∫f(dtx,{\mathrm{\θ}}_w)p\left({\mathrm{\θ}}_w\right|dtx){\mathrm{d\θ}}_w---\left(4\right)$

Step 3：Based on Markov chain Monte Carlo to light storing cogeneration system dynamics response performance exponential forecasting value Ask for：

Integration in formula (4) is function y_dt=f (dtx, θ_w) with regard to parameter Posterior distrbutionp p (θ_w| dtx) expectation, this mathematics Expecting can be with Markov chain Monte Carlo come approximately, formula is as follows：

$\hat{y}\≈\frac{1}{n_2}{\mathrm{\Σ}}_{n_1+1}^{n_2+n_1}f(dtx,{\mathrm{\θ}}_s)---\left(5\right)$

Wherein, n₁For some the initial Markov chains being rejected, n₂To come the weight vector sample number of self-balancing Posterior distrbutionp, θ_s For parameter θ_wThe sampled value of Posterior distrbutionp, solvesAs light storing cogeneration system dynamics response performance exponential forecasting value.