CN111708987B - Method for predicting load of multiple parallel transformers of transformer substation - Google Patents

Abstract

A method for predicting the load of a plurality of parallel transformers of a transformer substation comprises the following steps: step 1, acquiring historical data of the transformer substation in a set time period, wherein the historical data comprises transformer substation load data, load data of each transformer and transformer operation mode data in the set time period; step 2, calculating the operation mode C of the transformer substation by using historical data of the transformer substation _j Transformer T during operation _i Load distribution factor F at time t _t (T _i ，C _j ) (ii) a Step 3, the load distribution coefficient F obtained in step 2 _t (T _i ，C _j ) Combined with the load L of the substation _s (t) use of a non-linear regression function G (L) _s (t)，T _i ，C _j ) Quantized load distribution factor F _t (T _i ，C _j ) Load L with substation _s (t) a non-linear mapping relationship; step 4, the nonlinear regression function G (L) obtained in the step 3 _s (t)，T _i ，C _j ) Combined with the load L of the substation _s (t) load L of ith transformer at time t _i (t) performing a prediction. On the premise of improving the accuracy of the load prediction result of the transformer, the workload of load prediction modeling is greatly reduced, and the precision is realizedAnd the double improvement of the efficiency.

Description

A method for predicting load of multiple parallel transformers in substation

Technical Field

The present invention relates to the field of load prediction of electric power equipment, and in particular to a method for predicting the load of multiple parallel transformers in a substation.

Background Art

Transformer load is an important factor affecting the health status and insulation life of equipment. Accurate transformer load prediction is of great significance for optimizing load distribution, equipment status assessment and fault prediction. In practice, in addition to conventional factors such as day type, season and weather, transformer load is also affected by changes in substation operation mode and presents a more complex change pattern than substation-level load, making it impossible for existing power system load prediction technology to accurately predict transformer load. On the other hand, there are a huge number of transformers in the power grid, and establishing a load prediction model for each device separately will generate a heavy workload.

Prior art document 1 (Deng Zhixiang et al. A load forecasting method for distribution transformers and distribution lines [P]. CN110009136A, 2019-07-12.) discloses a load forecasting method for distribution transformers and distribution lines, which initializes various weights such as historical loads, meteorological data, and working day types, and performs neuron calculation and prediction based on the Elman neural network algorithm.

The method disclosed in the prior art document 1 applies the traditional load forecasting method for the power system level to the transformer level load. However, in practice, in addition to conventional factors such as historical load, meteorological data, and workday type, transformer load will also be affected by special factors such as the adjustment of substation operation mode, showing a more complex change pattern than the system-level load. If only conventional factors are considered and the impact of substation operation mode on transformer load is ignored, it will cause a large prediction error. On the other hand, the number of transformers in the power grid is huge, and establishing a load forecasting model for each transformer separately will generate a heavy workload.

Summary of the invention

The purpose of the present invention is to provide a method for predicting the load of multiple parallel transformers in a substation, by defining a "load distribution coefficient" and using a nonlinear regression function to quantify the nonlinear mapping relationship between the parallel transformer load and the substation load under different operating modes of the substation, so as to realize the prediction of the load of multiple parallel transformers.

The present invention adopts the following technical scheme. A load prediction method for multiple transformers in parallel in a substation, wherein _Ti represents the i-th transformer among N transformers in parallel in the substation, N, i is a positive integer, N≥2, 1≤i≤N; _Cj represents the j-th operation mode among K operation modes of the substation, K, j is a positive integer, K≥2, 1≤j≤K; t represents the measurement time, t is a positive integer; _Li (t) represents the load of the i-th transformer at time t; _Ls (t) represents the load of the substation at time t; and the method is characterized in that it comprises the following steps:

Step 1, obtaining historical data of the substation within a set time period, including substation load data, load data of each transformer, and substation operation mode data within the set time period;

Step 2: Use the historical data of the substation to calculate the load sharing coefficient _Ft ( _Tj , _Cj ) of the transformer _Ti at time t when the substation is running in operation mode _Cj using the following formula (1):

Where:

F _t (T _i , C _j ) represents the load sharing factor;

_Ti represents the i-th transformer among the N transformers running in parallel in the substation;

C _j represents the jth operation mode among K operation modes of the substation;

t represents the measurement time;

_Li (t) represents the load of the i-th transformer at time t;

L _s (t) represents the load of the substation at time t;

Step 3: Using the load distribution coefficient _Ft ( _Ti , _Cj ) obtained in step 2 and the load _Ls (t) of the substation, a nonlinear regression function G( _Ls (t), _Ti , _Cj ) is used to quantify the nonlinear mapping relationship between the load distribution coefficient _Ft ( _Ti , _Cj ) and the load _Ls (t) of the substation.

F _t (T _i , C _j ) = G (L _s (t), T _i , C _j ) (2)

Where:

G(L _s (t), _Ti , C _j ) represents a nonlinear regression function including the independent variable as the substation load L _s (t);

Step 4: Using the nonlinear regression function G( _Ls (t), _Ti , _Cj ) obtained in step 3 and the load of the substation _Ls (t), use the following formula (3) to predict the load _Li (t) of the i-th transformer at time t:

_Li (t)=G (L _s (t), _Ti , C _j )·L _s (t) (3).

Preferably, a power function containing a constant term is used as the nonlinear regression function G(L _s (t), _Ti , C _j )

G(L _s (t), T _i , C _j )=aL _s (t) ^b +c (4)

L _s (t) represents the load of the substation at time t;

a, b, c represent parameter values of G(L _s (t), _Ti , C _j ).

Preferably, the parameter values a, b, c of the nonlinear regression function G(L _s (t), _Ti , C _j ) are estimated using the least squares method.

Preferably, the range of the set time period in step 1 is 1 year to 3 years.

Preferably, the substation operation mode data obtained in step 1 refers to the transformer shutdown and operation status that have occurred in the historical operation of the transformer.

Preferably, historical data of at least four historical moments, i.e., t≥4, are obtained, including substation load data, load data of each transformer, and substation operation mode data, for fitting nonlinear functions and making predictions.

The beneficial effect of the present invention lies in that, compared with the prior art, the present invention first defines a "load distribution coefficient" to describe the distribution relationship between parallel transformers and substation-level loads, uses a nonlinear regression function to quantify the nonlinear mapping relationship between the "load distribution coefficient" and the substation-level load under different operating modes of the substation, and finally uses the substation load prediction result as input to realize the prediction of the loads of multiple parallel transformers according to the "load distribution coefficient" of the transformer.

Compared with simplifying to a constant, the nonlinear mapping relationship between the load distribution coefficient _Ft (T _i , C _j ) of each transformer and the substation load _Ls (t) can be quantified by using a nonlinear regression function, which can accurately and reasonably reflect the real dynamic change characteristics of the load distribution coefficient. The method greatly reduces the workload of load prediction modeling while improving the accuracy of transformer load prediction results, and achieves a dual improvement in both accuracy and efficiency.

The present invention exploits the nonlinear mapping relationship between the substation-level load and the load of multiple parallel transformers under different operating modes of the substation, takes the substation load prediction result as input, realizes the load prediction of multiple parallel transformers, and greatly reduces the workload of load prediction while improving the load prediction accuracy.

As long as the amount of historical data collected when the substation is running in a certain state _Cj is not less than 4, including 4 times, that is, not less than 4 points, a nonlinear function can be fitted and predicted. Therefore, the load prediction method for multiple parallel transformers in a substation provided by the present invention has another obvious advantage, that is, it can be used in scenarios where historical load data is scarce.

BRIEF DESCRIPTION OF THE DRAWINGS

Fig. 1 is a scatter diagram of the mapping relationship between the load distribution coefficient _Ft (T _i , C _j ) of transformer T ₁ and the substation load L _s (t) when the substation operation mode is C ₁ ;

FIG2 is a scatter diagram showing the mapping relationship between the load distribution coefficient F _t (T _i , C _j ) of transformer T ₁ and the substation load L _s (t) when the substation operation mode is C ₄ ;

FIG3 is a diagram showing the load prediction results of transformer _T1 ;

FIG4 is a flow chart of a method for predicting loads of multiple transformers in parallel in a substation.

DETAILED DESCRIPTION

The present application is further described below in conjunction with the accompanying drawings. The following embodiments are only used to more clearly illustrate the technical solution of the present invention, and cannot be used to limit the protection scope of the present application.

The present invention provides a load prediction method for multiple parallel transformers in a substation. The substation S includes N transformers running in parallel, _Ti represents the i-th transformer among the N transformers running in parallel in the substation, that is, the transformer number, N,i is a positive integer, N≥2, 1≤i≤N; _Cj represents the j-th operating mode among K operating modes of the substation, K,j is a positive integer, K≥2, 1≤j≤K; t represents a measurement time, t is a positive integer, if the time span for obtaining historical data is n days, 1≤t≤24n; _Li (t) represents the load of the i-th transformer at time t; _Ls (t) represents the load of the substation at time t.

As shown in FIG4 , the load prediction method for multiple parallel transformers in a substation includes the following steps:

Step 1, obtaining historical data of the substation within a set time period, including substation load data, load data of each transformer, and substation operation mode data within the set time period. Preferably, but not restrictively, the range of the set time period is 1 year to 3 years. The substation operation mode data obtained refers to the transformer shutdown and normal operation status that have occurred in the historical operation of the transformer.

Step 2: define the load sharing coefficient _Ft ( _Ti , _Cj ) as the ratio of the load of the i-th transformer _Li (t) to the load of the substation _Ls (t) at a certain time t when the substation is in operation mode Cj. Use the historical data of the substation to calculate the load sharing coefficient _Ft ( _Ti , _{Cj )} of the transformer _Ti at time t when the substation is in operation mode _Cj using the following formula (1):

Where:

F _t (T _i , C _j ) represents the load sharing factor;

_Ti represents the i-th transformer among the N transformers running in parallel in the substation;

C _j represents the jth operation mode among K operation modes of the substation;

t represents the measurement time;

_Li (t) represents the load of the i-th transformer at time t;

L _s (t) represents the load of the substation at time t.

Step 3: Using the load distribution coefficient _Ft ( _Ti , _Cj ) obtained in step 2 and the load _Ls (t) of the substation, a nonlinear regression function G( _Ls (t), _Ti , _Cj ) is used to quantify the nonlinear mapping relationship between the load distribution coefficient _Ft ( _Ti , _Cj ) and the load _Ls (t) of the substation using the following formula (2):

F _t (T _i , C _j ) = G (L _s (t), T _i , C _j ) (2)

Where:

G(L _s (t), _Ti , C _j ) represents the nonlinear regression function of the load L _s (t) of the substation including the independent variable;

_Ti represents the i-th transformer among the N transformers running in parallel in the substation;

_Cj represents the jth operating mode among K operating modes of the substation.

Under ideal conditions, the transformer load sharing coefficient is considered a constant. However, in real scenarios, the various operating parameters of different parallel transformers will be affected by the load level of the substation and will change to varying degrees, and will indirectly affect the load sharing coefficient of each parallel transformer. Therefore, compared with simplifying it to a constant, using a nonlinear regression function to quantify the nonlinear mapping relationship between the load sharing coefficient _Ft (T _i , C _j ) of each transformer and the substation load _Ls (t) can accurately and reasonably reflect the real dynamic change characteristics of the load sharing coefficient.

Preferably, the substation load L _s (t) is taken as the independent variable, the load distribution coefficient F _t (T _i , C _j ) is taken as the dependent variable, and a power function G (L _s (t), T _i , C _j ) containing a constant term is used to quantify the nonlinear mapping relationship between the substation load L _s (t) and the load distribution coefficient F _t (T _i , C _j ), which is expressed as formula (3):

G(L _s (t), T _i , C _j )=aL _s (t) ^b +c (3)

L _s (t) represents the load of the substation at time t;

a, b, c represent parameter values of G(L _s (t), _Ti , C _j ).

The least squares method is used to estimate the parameter values a, b, and c of the nonlinear regression function G( _Ls (t), _Ti , _Cj ). It can be seen that the parameter values a, b, and c change under the influence of the substation load _Ls (t), transformer number _Ti, and substation operation mode _Cj .

Step 4: The nonlinear regression function G( _Ls (t), _Ti , _Cj ) obtained in step 3 is combined with the load of the substation _Ls (t) to predict the load _Li (t) of the i-th transformer at time t using the following formula (4). _Ls (t) can be achieved through substation-level short-term or ultra-short-term load forecasting technology, which is also relatively easy to obtain in the power system because the load at the substation level is required to be predicted by the power grid.

L _i (t) = G (L _s (t), T _i , C _j )·L _s (t) (4)

Where:

_Li (t) represents the load of the i-th transformer at time t;

G(L _s (t), _Ti , C _j ) represents the nonlinear regression function of the load L _s (t) of the substation including the independent variable;

L _s (t) represents the load of the substation at time t.

When the substation operation mode _Ci or transformer number _Ti changes, it is only necessary to replace the corresponding nonlinear function G( _Ls (t), _Ti , _Cj ) to realize the load prediction of substations with different operation states and different parallel transformers.

It is worth noting that as long as the amount of historical data collected when the substation is running in a certain state _Cj is not less than 4, including 4 times, that is, not less than 4 points, a nonlinear function can be fitted and predicted. Therefore, the load prediction method for multiple parallel transformers in a substation provided by the present invention has another obvious advantage, that is, it can be used in scenarios where historical load data is scarce.

Compared with the conventional load forecasting method for the power system level, represented by prior art document 1, which extends the traditional load forecasting method for the transformer level, the present invention exploits the nonlinear mapping relationship between the substation-level load and the load of multiple parallel transformers under different operating modes of the substation, takes the substation load forecasting result as input, and realizes the load forecasting of multiple parallel transformers, thereby greatly reducing the workload of load forecasting while improving the load forecasting accuracy.

The following is an example of implementing the load prediction method for multiple parallel transformers in the substation:

A substation S includes four transformers T ₁ , T ₂ , T ₃ , and T ₄ running in parallel. The energy management system (EMS) stores the historical online monitoring data of active power of the four transformers T ₁ , T ₂ , T ₃ , and T ₄ running in parallel from 2015 to 2016. The substation S has a total of four operating modes in these two years, as shown in Table 1:

In real situations, as shown in Table 2, substations generally operate at C1 most of the time (i.e., all transformers in the substation are operating normally), while C2-C4 are generally uncommon (a transformer in the substation is out of service, while the rest are operating normally). Therefore, C1 often has a large amount of historical data, while C2-C4 historical data is relatively scarce. Since the power function structure containing a constant term is very simple, even if the amount of data used for fitting is very large, it will not have a significant impact on the calculation efficiency.

Table 1 Substation operation mode

Taking _T1 as an example, when the substation operates in _C1 and _C4 modes, the scatter diagrams of the mapping relationship between _T1 's load distribution coefficient _Ft ( _T1 , _C1 ) and _Ft ( _T1 , _C4 ) and the substation load _Ls (t) are shown in Figures 1 and 2 respectively.

According to the scatter plot of the load distribution coefficient and the substation load _Ls (t) in the historical load data as shown in Figures 1 and 2, the nonlinear regression function shown in formula (2) is used to quantify the nonlinear mapping relationship between _Ft ( _T1 , _C1 ) and _Ft ( _T1 , _C4 ) and _Ls (t), and the fitting results are shown in the black curves in Figures 1 and 2, respectively. It can be seen from the figure that the power function with a constant term can accurately quantify and fit the nonlinear mapping relationship between the transformer load distribution coefficient and the substation load. The parameter fitting results of the nonlinear regression function of the parallel transformer under the operating state of the other substations are shown in Table 2.

Table 2 Parameter fitting results of nonlinear regression function

242 historical load values of substation S in a certain period of time were collected. In the two periods [63, 133] and [160, 234], the operation mode of the substation was C ₄ (T ₃ was temporarily out of operation), and the rest of the periods were C _1. Taking T ₁ as an example, equation (4) is used for load forecasting, and the corresponding G(L _s (t), T ₁ , C ₁ ) and G(L _s (t), T ₁ , C ₄ ) are replaced as nonlinear regression functions according to the substation operation mode. The prediction results are shown in Figure 3, and the root mean square error of the prediction value is shown in Table 3. Compared with treating the load distribution coefficient as a constant value, using a nonlinear regression function for load forecasting can significantly improve the accuracy of transformer load forecasting.

Table 3 Prediction results errors of different load forecasting methods

It can be seen from the prediction results of Figure 3 that the transformer load method of the present invention has a high prediction accuracy. When the operation mode of the substation changes, or when it is necessary to predict the load of other parallel transformers, it is only necessary to replace the corresponding nonlinear regression function to accurately predict the load of any transformer under any operation mode. In addition, compared with establishing a separate load prediction model for each transformer, the use of replacing the nonlinear regression function can greatly reduce the workload of load prediction.

The beneficial effect of the present invention lies in that, compared with the prior art, the present invention first defines a "load distribution coefficient" to describe the distribution relationship between parallel transformers and substation-level loads, uses a nonlinear regression function to quantify the nonlinear mapping relationship between the "load distribution coefficient" and the substation-level load under different operating modes of the substation, and finally uses the substation load prediction result as input to realize the prediction of the loads of multiple parallel transformers according to the "load distribution coefficient" of the transformer.

Compared with simplifying to a constant, the nonlinear mapping relationship between the load distribution coefficient _Ft (T _i ,C _j ) of each transformer and the substation load _Ls (t) can be quantified by using a nonlinear regression function, which can accurately and reasonably reflect the real dynamic change characteristics of the load distribution coefficient. The method greatly reduces the workload of load prediction modeling while improving the accuracy of transformer load prediction results, and achieves a dual improvement in both accuracy and efficiency.

As long as the amount of historical data collected when the substation is running in a certain state _Cj is not less than 4, including 4 times, that is, not less than 4 points, a nonlinear function can be fitted and predicted. Therefore, the load prediction method for multiple parallel transformers in a substation provided by the present invention has another obvious advantage, that is, it can be used in scenarios where historical load data is scarce.

The applicant of the present invention has made a detailed explanation and description of the implementation examples of the present invention in conjunction with the drawings in the specification. However, those skilled in the art should understand that the above implementation examples are only preferred implementation schemes of the present invention, and the detailed description is only to help readers better understand the spirit of the present invention, and it is not a limitation on the protection scope of the present invention. On the contrary, any improvements or modifications based on the inventive spirit of the present invention should fall within the protection scope of the present invention.

Claims (5)

1. A load prediction method for multiple parallel transformers in a substation

Indicates that the substation

The first of the transformers in parallel operation

Transformer,

is a positive integer,

;

Indicates substation

The first mode of operation

Types of operation,

is a positive integer,

;

Indicates the measurement time.

is a positive integer;

express

Moment

The load of the transformer;

express

The load of the substation at any moment; characterized in that it comprises the following steps: Step 1, obtaining historical data of the substation within a set time period, including substation load data, load data of each transformer, and substation operation mode data within the set time period; The substation operation mode data obtained refers to the transformer outage and operation status that have occurred in the historical operation of the transformer; Step 2: Use the historical data of the substation using the following formula

Calculate the substation operation mode

Runtime Transformer

exist

Load sharing factor at the moment

Where:

represents the load sharing factor;

Indicates that the substation

The first of the transformers in parallel operation

Transformer;

Indicates substation

The first mode of operation

Types of operation;

Indicates the measurement time;

express

Moment

The load of each transformer;

express

The load of the substation at any moment; Step 3: Use the load sharing factor obtained in step 2

, combined with the load of the substation

, using the nonlinear regression function

Quantifying the load sharing factor

The load of the substation

The nonlinear mapping relationship is

Where:

Indicates that the independent variable is the substation load

Nonlinear regression function of ; Step 4: Use the nonlinear regression function obtained in step 3

, combined with the load of the substation

, using the following formula

right

Moment

Transformer load

Make predictions,

. 2. The load prediction method for multiple parallel transformers in a substation according to claim 1 is characterized in that: Use the following power function with a constant term as the nonlinear regression function

express

The load of the substation at any moment;

express

Parameter value. 3. The load prediction method for multiple parallel transformers in a substation according to claim 2 is characterized in that: Estimate nonlinear regression function using least squares method

Parameter value

. 4. The method for predicting loads of multiple transformers in parallel in a substation according to any one of claims 1 to 3, characterized in that: The range of the time period set in step 1 is 1 year to 3 years. 5. The method for predicting loads of multiple transformers in parallel in a substation according to any one of claims 1 to 3, characterized in that: Get historical data for at least 4 historical moments, namely

, including substation load data, load data of each transformer, and substation operation mode data, which are used to fit nonlinear functions and make predictions.

Images