CN111276977A - Power shortage prediction method after power system disturbance - Google Patents

Abstract

The invention relates to a power shortage prediction method after power system disturbance, which comprises the following steps: step S1, when the frequency of the power system is reduced, the corresponding input characteristic quantity is collected by using a self-adaptive power shortage physical computation model; s2, generating a sample in an actual power grid, standardizing data to obtain a training sample S3, constructing an LSTM prediction model and an LSTM prediction model, inputting the training sample into the prediction model for training and learning, and adjusting parameters of the prediction model to obtain an LSTM predicted power shortage model; step S4, input characteristic quantity is input to a trained LSTM predicted power shortage model for testing after being standardized, and a standardized predicted power shortage value is obtained; and step S5, performing anti-standardization processing according to the obtained standardized power shortage predicted value to obtain the power shortage predicted value. The method can quickly and accurately predict the power shortage after the disturbance of the power system and provide guidance information for the follow-up self-adaptive low-frequency load shedding control.

Description

Power shortage prediction method after power system disturbance

Technical Field

The invention belongs to the technical field of power systems, and particularly relates to a power shortage prediction method after a power system is disturbed.

Background

The power shortage is an important characteristic quantity in the low-frequency load shedding control. The severity of the fault and the chain reaction possibly brought by the fault can be known through predicting the power shortage after the frequency disturbance, and a key guiding effect is played on the load shedding amount of the low-frequency load shedding control. And by combining a deep learning algorithm, analyzing and processing the key characteristic quantity after the fault and performing regression prediction, and a new method is provided for a self-adaptive low-frequency load shedding scheme based on response information.

Disclosure of Invention

In view of this, the present invention provides a power shortage prediction method after a power system disturbance, which only needs frequency and active information of a load side, and can predict a power shortage after an actual power grid disturbance according to an LSTM power shortage prediction model after a frequency disturbance fault occurs in the power system.

In order to achieve the purpose, the invention adopts the following technical scheme:

a power shortage prediction method after power system disturbance comprises the following steps:

step S1, when the frequency of the power system is reduced, the corresponding input characteristic quantity is collected by using a self-adaptive power shortage physical computation model;

step S2, generating samples in the actual power grid, and standardizing the data to obtain training samples

Step S3, constructing an LSTM prediction model and an LSTM prediction model, inputting training samples into the prediction model for training and learning, and adjusting parameters of the prediction model to obtain an LSTM prediction power shortage model;

step S4, input characteristic quantity is input to a trained LSTM predicted power shortage model for testing after being standardized, and a standardized predicted power shortage value is obtained;

and step S5, performing anti-standardization processing according to the obtained standardized power shortage predicted value to obtain the power shortage predicted value.

Further, when the frequency reduction disturbance occurs to the power system, the power shortage includes two parts, namely unbalanced power and active power shortage caused by voltage deviation, and the self-adaptive physical calculation method of the power shortage is as follows:

in the formula:

the change rate of the center frequency of inertia in the system;

are respectively t₁And t₂The rate of change of the moment of inertia center frequency;

is t₁And t₂The moment load has the real power.

The load before the disturbance occurrence moment has active power;

and T is the load active power at the T moment after the disturbance occurs. The input characteristic quantity comprises load frequency and load active information as input characteristic quantity.

Further, the step S2 is specifically: the data were normalized by z-score using the following formula

In the formula: μ is the mean of the sample data and σ is the standard deviation of the sample data.

Further, the LSTM prediction model specifically includes:

f_t＝σ(W_xfx_t+W_hfh_t-1+W_cfc_t-1+b_f)

i_t＝σ(W_xix_t+W_hih_t-1+W_cic_t-1+b_i)

o_t＝σ(W_xox_t+W_hoh_t-1+W_coc_t-1+b_o)

c_t＝f_tc_t-1+i_ttanh(W_xcx_t+W_hch_t-1+b_c)

h_t＝o_ttanh(c_t)

in the formula: x is the number of_tInputting a state for the current moment; f. of_tInputting the current moment for the forgotten door; h is_tIs the output of the LSTM unit at the current moment; o_tOutputting for the output gate at the current moment; c. C_tThe state of the memory unit at the current moment is memorized; w_xi、W_xf、W_xoAnd W_xcIs x_tA correlated weight matrix; w_hi、W_hf、W_hoAnd W_hcIs h_tThe correlation weight matrix of (a); w_ci、W_cfAnd W_coIs c_tThe relative weight value of (a); b_i、b_f、b_oAnd b_cIs the associated offset.

Further, the denormalization is shown as follows:

X＝X^*×σ+μ

in the formula: μ is the mean of the sample data, σ is the standard deviation of the sample data, X^*To normalize the data.

Compared with the prior art, the invention has the following beneficial effects:

the method only needs the frequency and active information of the load side, and can predict the power shortage of the actual power grid after disturbance according to the LSTM power shortage prediction model after the frequency disturbance fault of the power system occurs. When the power system lacks a perfect communication device and a perfect control center, the power shortage of the system can be quickly and accurately output by depending on the information of the load side, important guiding information is provided for the follow-up self-adaptive low-frequency load shedding control, the frequency is effectively prevented from further collapse, and the important role is played in the safe and stable control of the power system.

Drawings

FIG. 1 is a diagram of IEEE39 nodes in one embodiment of the invention;

fig. 2 is a schematic diagram of the mechanism of the LSTM base unit in one embodiment of the invention.

Detailed Description

The invention is further explained below with reference to the drawings and the embodiments.

Referring to fig. 1, in the present embodiment, an IEEE-39 node expansion system is selected, and the IEEE-39 node system is a 10-machine system. Removing in-system balancing machine G₂And a generator G of large inertia₁₀Each of the remaining eight generators is developed into three generators G_i-1、G_i-2、G_i-3(i is 1,3,4,5,6,7,8,9), the output of the three generators is the same.

3 generators in the expanded generators are cut off at the same time in each fault disturbance setting, and the generators are combined to generate in a common energy mode

And simulating the possible generator disturbance fault of the system by using the sample. The simulation time length is 100 cycles each time, the time length is 0.02s when one cycle is carried out, 2s are total, the machine cutting amount disturbance is set when the 5 th cycle is carried out, and 3 generators are cut simultaneously.

Can generate C3 in simulation_3×82024 samples, 1600 samples were randomly selected for training, and 400 samples were randomly drawn for testing for the remaining 424 samples.

The input characteristic quantity of the LSTM prediction model is load frequency deviation and load active information of a time period of 0s-2s, and the predicted time is the power shortage of the time period. The Root Mean Square Error (RMSE) for this scenario, trained with the LSTM algorithm, RNN algorithm, and conventional BP neural network, using 1600 training samples and 400 test sample sets is shown in the table below.

TABLE 1

At the moment of power disturbance, namely before and after the 5 th cycle, the predicted value of the power shortage deviates from the actual value by a certain amount. But after the perturbation occurs for 0.2s, the predicted value can approach the actual value.

The above description is only a preferred embodiment of the present invention, and all equivalent changes and modifications made in accordance with the claims of the present invention should be covered by the present invention.

Claims (5)

1. A power shortage prediction method after power system disturbance is characterized by comprising the following steps:

step S1, when the frequency of the power system is reduced, the corresponding input characteristic quantity is collected by using a self-adaptive power shortage physical computation model;

step S2, generating samples in the actual power grid, and standardizing the data to obtain training samples

Step S3, constructing an LSTM prediction model and an LSTM prediction model, inputting training samples into the prediction model for training and learning, and adjusting parameters of the prediction model to obtain an LSTM prediction power shortage model;

step S4, input characteristic quantity is input to a trained LSTM predicted power shortage model for testing after being standardized, and a standardized predicted power shortage value is obtained;

and step S5, performing anti-standardization processing according to the obtained standardized power shortage predicted value to obtain the power shortage predicted value.

2. The method of claim 1, wherein the input characteristic quantities comprise load frequency and load active information as input characteristic quantities.

3. The method for predicting the power shortage after the disturbance of the power system as claimed in claim 1, wherein the step S2 specifically comprises: the data were normalized by z-score using the following formula

In the formula: μ is the mean of the sample data and σ is the standard deviation of the sample data.

4. The method for predicting the power shortage after the disturbance of the power system as claimed in claim 1, wherein the LSTM prediction model specifically comprises:

f_t＝σ(W_xfx_t+W_hfh_t-1+W_cfc_t-1+b_f)

i_t＝σ(W_xix_t+W_hih_t-1+W_cic_t-1+b_i)

o_t＝σ(W_xox_t+W_hoh_t-1+W_coc_t-1+b_o)

c_t＝f_tc_t-1+i_ttanh(W_xcx_t+W_hch_t-1+b_c)

h_t＝o_ttanh(c_t)

in the formula: x is the number of_tInputting a state for the current moment; f. of_tInputting the current moment for the forgotten door; h is_tIs the output of the LSTM unit at the current moment; o_tOutputting for the output gate at the current moment; c. C_tThe state of the memory unit at the current moment is memorized; w_xi、W_xf、W_xoAnd W_xcIs x_tA correlated weight matrix; w_hi、W_hf、W_hoAnd W_hcIs h_tThe correlation weight matrix of (a); w_ci、W_cfAnd W_coIs c_tThe relative weight value of (a); b_i、b_f、b_oAnd b_cIs the associated offset.

5. The method of claim 1, wherein the denormalization is expressed by the following formula:

X＝X^*×σ+μ

in the formula: μ is the mean of the sample data, σ is the standard deviation of the sample data, X^*To normalize the data.

Images