TWI420337B - Method of establishing loop power information model - Google Patents

Description

Method of establishing a loop power information model

The present invention relates to a power monitoring method, and more particularly to a method for establishing a loop power information model that can be used as a reference for estimating the use and power consumption state of various electrical appliances in a home.

Nowadays, global awareness of environmental protection and energy conservation is high, and the green energy industry that saves energy and reduces carbon is a very popular project. In terms of household electricity, according to the 2008 US electricity consumption statistics, household electricity consumption accounts for 37% of total electricity consumption. Therefore, only by understanding the behavior of using electricity, can we effectively achieve the goal of energy saving at home. In order to accurately analyze the behavior of household electricity, it is most common to install a power meter on each electrical appliance to measure the electrical condition of the electrical appliance.

However, the installation of multiple power meters is quite costly, and the more power meters will consume more power. Therefore, the present invention proposes a method for monitoring household power information by using a loop type power meter for the above-mentioned problems, by establishing a loop power information model including the order and correlation of the use of the electric appliance, and can be identified by the loop power information model. The power consumption status of the electrical appliances on the circuit allows the user to keep abreast of the current statistical information of the state of the electricity consumption information, and can provide accurate power saving suggestions by analyzing the electricity usage data.

The main object of the present invention is to provide a method for establishing a loop power information model, which is to quantify the order and correlation of electrical appliances usage by establishing a state probability model of the loop electrical appliance, thereby effectively improving the recognition degree of the electrical state combination on the loop.

Another object of the present invention is to provide a method for establishing a loop power information model, which is monitored by a loop type power meter and can be centrally managed at a power distribution panel, which will greatly reduce system cost and power consumption.

A further object of the present invention is to provide a method for establishing a loop power information model, which can be easily combined with a smart meter, and can be combined with various power statistics algorithms for real-time monitoring of power to analyze and count user details. The use of electricity will greatly increase the goal of energy conservation.

In order to achieve the above object, the present invention provides a method for establishing a loop power information model, which comprises the steps of utilizing a loop meter and a socket meter to collect loop power information and electrical status information of a plurality of electrical appliances on the loop, and The real power and the spent power of the plurality of time segments of the loop power information are sequentially extracted, and the feature vector of each time segment is calculated by the real power and the spent power, and the electrical state information is integrated according to the time segment. And the feature vector is a plurality of probability model training data, and the time series of the training data of each probability model is calculated, and then the time value is used to calculate the probability value of the electrical usage state of the training data of the relative probability model to establish the probability of using the state of the circuit electrical appliance. model.

The purpose, technical contents, features and effects achieved by the present invention will be more readily understood by the detailed description of the embodiments and the accompanying drawings.

The invention provides a method for establishing a loop power information model, which considers the order and correlation of the use of a plurality of electrical appliances on the same circuit, and establishes the probability model training data by using the loop power information and the electrical state information collected from the loop. Then, through the time series of the computer rate model training data, the probability value of the electrical usage state of the training data of the corresponding probability model is calculated to establish a probability model of the use state of the primary circuit electrical appliance. The technical features of the present invention will be described in detail below with reference to preferred embodiments.

The first figure is a flow chart of establishing a loop power information model according to the present invention. As shown in the figure, first, in step S10, at least one loop meter is used on at least one loop to fix the sampling frequency, and the voltage and power on the loop are obtained. Factor and apparent power are used as loop power information. And a plurality of electrical appliances on the circuit utilize a plurality of socket meters to measure the operating power of each electrical appliance as a status information of the electrical appliance in a fixed time range, and obtain various states of the electrical appliances by calculating electrical state information. The maximum apparent power and the minimum apparent power consumed, and the plurality of state slitting points of the electrical appliance status information are marked according to the operation of the electrical appliance to distinguish different operating states of the electrical appliances.

Then, in step S12, the sliding window mode is used to sequentially capture the real power and the spent power of the plurality of time segments of the loop power information, and then calculate each time zone according to the obtained real power and the spent power. One of the feature vectors of the segment, each feature vector includes a plurality of eigenvalues such as an average value, a root mean square value, a variance number, a maximum value, a minimum value, a maximum difference value, and a waveform factor, and the waveform factor is obtained by the maximum value The Crest Factor, the Form Factor, the Peak to Average Ratio, and the Wavelet Transform are calculated from the mean and root mean square values.

Finally, in step S14, the electrical state information and the feature vector are integrated into the plurality of probability model training materials according to the time segment, and the electrical state information and the feature vector of each training model training data are generated to generate the corresponding probability model training data. An electrical characteristic equation, an electrical state sequence characteristic equation, and an electrical appliance use a dependent characteristic equation to represent a time series of the probability model training data, and then calculate a probability value of the electrical usage state of the probability model training data by using a time series to establish The primary circuit appliance uses a state probability model.

The above is an overall step description of establishing a loop power information model according to the present invention. The following will take a lamp, an electric fan and an electric cooker in the first loop as an example, and establish a loop electric appliance use state for integrating the electrical state information and the feature vector in step S14. The probability model is described in further detail, please refer to the second and third figures.

In the above step S14, the electrical state information and the feature vector are integrated into the plurality of probability model training materials according to the time segment. Please refer to the second figure, the schematic diagram of the probability model training data of the present invention, as shown in the figure, each probability model The training data 10 is composed of the electrical status information of the desk lamp, the electric fan and the electric cooker, and the corresponding feature vector (not shown), and the electric appliances of the desk lamp, the electric fan and the electric cooker are occupied in the probability model training data 10. Most of the operational status will represent the main operational status of this probability model training material 10. Therefore, in the probability model training data 10 shown in the second figure, the main operating state of the desk lamp is on, the main operating state of the electric fan is the amount of stroke, and the main operating state of the electric pot is off. And the correlation between the use of each electrical appliance, taking the desk lamp as an example, when the desk lamp is turned on, the probability that the electric fan is set to a stroke amount is higher than the probability that the electric kettle is set to be insulated, and the order of the electric appliance combination conversion is in the desk lamp and electricity. When the fans are used together, the next time point is that the combination of the same table lamp and the electric fan is higher.

In order to clearly establish the correlation between the use of lamps, electric fans, electric cookers and other electrical appliances, as well as the sequential characteristics of the electrical combination conversion, the time series of each probability model training data 10 will be calculated. Please also refer to the second figure. As shown in the three figures, the appliance state information and feature vector of the training model 10 according to the probability model are calculated, and the electrical characteristic equation 20 of the corresponding probability model training data 10, the electrical state sequence characteristic equation 22, and the appliance using the dependent characteristic equation 24 are expressed as The time series of the probability model training data 10 is representative of the relevance of the electrical appliance usage and the sequence of the electrical components. Taking the desk lamp as an example, the electrical characteristic equation 20 is expressed as (on): 1 time; (on): 3 times. Its electrical state sequence characteristic equation 22 is expressed as (on, on): 3 times. The electrical appliance uses the dependent characteristic equation 24 to be divided into an electric appliance overall combination and a two-phase electrical appliance combination. The overall electrical appliance combination is expressed as (open, closed, closed): 1 time; (open, stroke volume, closed): 3 times. The two-phase electrical combination table lamp and electric fan are expressed as (on, off): 1 time; (on, stroke volume): 3 times. The two-phase electric combination table lamp and electric cooker are expressed as (open, closed): 4 times.

Calculate the time series of each probability model training data 10, and then calculate the electrical usage state probability value representing each probability model training data 10 by time series, and the electrical usage state probability value and time series system will satisfy the following Equation (1):

among them, Representing the electrical characteristic equation 20, s _i is the electrical state information and feature vector combination of each electrical appliance at the same time, i is a natural number, In order to correspond to the feature weight value of s _i , the feature weight value is the correlation weight of the operational state of each appliance and the corresponding feature vector combination. Representing the electrical state sequence characteristic equation 22, u _j is the electrical state information conversion combination of the same electrical appliance, j is a natural number, In order to correspond to the order weight value of u _{j ,} the order weight value is the sequential weight of the operational change of each appliance. Representing the appliance using the dependent characteristic equation 24, v _k is the combination of electrical state information between the appliances at the same time, k is a natural number, In order to correspond to the dependent weight value of v _{k ,} the dependent weight value is the dependence of the overall operating state of the electrical appliance and the operating state of any two electrical appliances. Z is a normalized denominator used to normalize the value of the exponential term in equation (1) to a value between 0 and 1 to satisfy the probability characteristic. p _λ is the value of the expected state of use of the appliance. When calculating the probability value of the electrical usage state of each probability model training material 10 via the above formula (1), the optimization algorithm may be adjusted by a Newton method, a gradient ascending method, a belief transfer algorithm or a repeated classification algorithm. The weight value, the order weight value, and the dependent weight value are used to maximize the probability of using the appliance. And the characteristic weight value, the sequence weight value, the dependent weight value, and the electrical activity state probability value of each probability model training data 10 obtained are stored to establish a loop electrical appliance usage state probability of being used as an identification circuit electrical appliance. model.

The above is a description of the probability state model of the loop electrical appliance state established by integrating the electrical state information and the feature vector in step S14, and the use state of the loop electrical appliance usage state probability model to identify the use state of the loop electrical appliance will be described below.

The fourth figure is a flow chart of the circuit state usage probability model of the present invention applied to the identification circuit power state. As shown in the figure, first, in step S20, the loop power meter is used to detect the loop power consumption information on the circuit. Then, in step S22, the power characteristics of the loop power information are retrieved. Finally, in step S24, a Viterbi Algorithm is used in conjunction with the loop electrical appliance to use a state probability model to perform a single electrical inference for each electrical appliance to match the electrical characteristic to the loop electrical appliance usage state probability model, and the maximum probability is obtained. The solution is the state of use of the appliance. Then use the cross classification algorithm (lterative Classification Algorithm) to match the state probability model of the loop electrical appliance, and carry out the cross-electrical inference for the dependent electrical appliances to confirm the use state of the final electrical appliance to identify the use status of the plurality of electrical appliances on the loop.

It can be seen from the description of the embodiments that the present invention establishes a loop electrical appliance usage state probability model as a loop power information model by quantifying the order and correlation of electrical appliances. And through the algorithm with this model can be used to subdivide the use of household appliances recorded on the meter to identify the operating status of each appliance.

The embodiments described above are merely illustrative of the technical spirit and the features of the present invention, and the objects of the present invention can be understood by those skilled in the art, and the scope of the present invention cannot be limited thereto. That is, the equivalent variations or modifications made by the spirit of the present invention should still be included in the scope of the present invention.

10. . . Probability model training data

20. . . Electrical characteristic equation

twenty two. . . Electrical state sequence characteristic equation

twenty four. . . Electrical appliance using dependent characteristic equation

The first figure is a flow chart of establishing a loop power information model of the present invention.

The second figure is a schematic diagram of the probability model training data of the present invention.

The third figure is a schematic diagram of the time value sequence for calculating the usage state of the appliance in the present invention.

The fourth figure is a flow chart of the circuit state probability model of the circuit appliance used in the present invention applied to the identification circuit power state.

Claims (10)

A method for establishing a loop power information model, comprising the steps of: collecting loop power information of at least one loop and electrical status information of a plurality of electrical appliances on the loop; and capturing a plurality of time segments of the loop power information Real power and power consumption, and corresponding to the power consumption and the power consumption calculation corresponding to one of the feature vectors of each time segment; and integrating the electrical state information and the feature vector into a plurality of probability model training materials, and based on Calculating the time series of the training data of the probability model according to the correlation between the use of the electrical appliances and the sequential characteristics of the electrical combination conversion, and calculating the probability value of the electrical usage state of the training data corresponding to the probability model according to the time series, And a characteristic weight value, a sequential weight value, a dependent weight value and a probability value of the electrical usage state of each of the probability model training materials obtained are stored, so as to be established as an available circuit electrical appliance. The primary circuit appliance uses a state probability model. The method for establishing a loop power information model according to claim 1, wherein the step of collecting the loop power information of the loop and the electrical status information of the appliance on the loop is through at least one loop The meter, at a fixed sampling frequency, measures the voltage, power factor and apparent power of the circuit to become the power consumption information of the circuit. The method for establishing a loop power information model according to claim 1, wherein in the step of collecting the loop power information of the loop and the electrical status information of the appliance on the loop, the plurality of sockets are The electric meter, in a fixed time range, measures the operating power of each electrical appliance to become the electrical state information, and calculates the maximum apparent power and the minimum apparent power of the electrical state information, and according to the operating electrical marking of the electrical appliance A plurality of state cut points for the electrical status information. The method for establishing a loop power information model according to claim 1, wherein in the step of extracting the real power and the power of the time segment of the loop power information, the feature vector has a plurality of Eigenvalue, which is the mean value, mean square Root value, variance number, maximum value, minimum value, maximum difference value, and waveform factor. The method for establishing a loop power information model according to claim 1, wherein in the step of integrating the electrical state information and the feature vector into the probability model training data, calculating the electrical state of the probability model training data The information and the feature vector generate an electrical characteristic equation, an electrical state sequence characteristic equation, and an electrical appliance use the dependent characteristic equation to represent the time series. The method for establishing a loop power information model according to claim 5, wherein in the step of integrating the electrical state information and the feature vector is the probability model training data, calculating the relative probability model training data according to the time series The value of the state of use of the appliance is such that the following conditions are met: among them, For the characteristic equation of the electrical appliance, s _i is the electrical state information of each electrical appliance at the same time combined with the characteristic vector, i is a natural number, Corresponding to the feature weight value of s _i , the feature weight value is the correlation weight of the operating state of each electrical appliance and the corresponding combination of the feature vectors. For the electrical state sequence characteristic equation, u _j is the electrical state information conversion combination of the same electrical appliance, j is a natural number, which is a sequential weight value corresponding to u _j , and the sequential weight value is for each electrical appliance The sequential weight of change, Using a dependent characteristic equation for the electrical appliance, v _{k is} a combination of the electrical state information of the electrical appliance at the same time, k is a natural number, In order to correspond to the dependent weight value of v _{k ,} the dependent weight value is the dependence of the overall operating state of the electrical appliance and the operating state of any two electrical appliances, Z is a normalized denominator, and p _λ is a probability value of the electrical use state. The electrical usage state probability value of each of the probability model training materials is calculated through the above formula (1) to establish the loop electrical appliance usage state probability model. A method for establishing a loop power information model as described in claim 6 wherein the step of integrating the appliance state information and the feature vector is the probability model training data In the step, the feature weight value, the order weight value, and the dependent weight value are adjusted by an optimization algorithm to maximize the usage probability value of the appliance. For example, the method for establishing a loop power information model described in claim 7 may be a Newton method, a gradient rising method, a belief transfer algorithm or a repeated classification algorithm. The method for establishing a loop power information model according to claim 1, wherein in the step of extracting the real power and the power of the time segment of the loop power information, a sliding window manner is utilized. The real power and the consumed power of the time segment of the loop power information are sequentially captured and the corresponding feature vector is calculated. The method for establishing a loop power information model according to claim 4, wherein the waveform factor is a Crest Factor calculated by the maximum value, the average value, and the root mean square value. Form Factor, Peak to Average Ratio, and Wavelet Transform.