An unmanned intelligent power station island micro-grid frequency control method

Abstract

The application provides a kind of island micro-grid frequency control method, belong to power grid control technical field.The method provided by the application has the following steps: step one, according to the transient motion equation of synchronous generator, the virtual rotor motion equation of virtual synchronous generator (VSG) is established;Step two, the transfer function of system is obtained by using the virtual rotor motion equation of VSG, the frequency dynamic stability is studied, and the influence relationship of virtual moment of inertia J and damping coefficient D on the system is analyzed;Step three, according to the current operating state of power system, the virtual moment of inertia J and damping coefficient D of VSG system are updated in real time.The beneficial effects of the application are: by adaptive adjustment of damping coefficient, the virtual inertia is regulated by BP neural network algorithm, which can effectively suppress frequency oscillation, reduce overshoot, and improve the stability and dynamic performance of the system.

Description

TECHNICAL FIELD

[0001] The present application relates to the field of power grid control, in particular to a frequency control method for an island micro-grid based on an unmanned intelligent power station. BACKGROUND

[0002] When a large number of distributed power sources are connected to an island micro-grid, the stability of the power grid is difficult to guarantee due to the low inertia, small capacity and poor dynamic characteristics of the traditional synchronous generator. The virtual synchronous generator is a network-type converter, which can greatly improve the inertia of the island micro-grid when connected to the power grid.

[0003] However, the frequency of the VSG parallel system will deviate greatly and oscillate under load disturbance, and the transient frequency stability is poor, which greatly restricts the promotion and application of VSG. In order to suppress frequency oscillation, the control parameters of the virtual synchronous generator are generally set to adaptive adjustment, but the control parameters are not linearly changed, and the regulation effect of the ordinary adaptive function is not ideal. SUMMARY

[0004] To solve the problems in the above background art, the present application provides a frequency control method for an island micro-grid, which can effectively suppress frequency oscillation, reduce overshoot and improve system stability and dynamic performance.

[0005] To achieve the above-mentioned purpose, the present application provides the following technical solution: a frequency control method for an island micro-grid is provided, and the specific steps are as follows:

[0006] Step 1: According to the transient motion equation of the synchronous generator, the virtual rotor motion equation of the virtual synchronous generator is established;

[0007] Step 2: The transfer function of the system is obtained by using the virtual rotor motion equation of the virtual synchronous generator, the frequency dynamic stability is studied, and the influence relationship of the virtual moment of inertia J and the damping coefficient D on the system is analyzed;

[0008] Step 3: According to the current operating state of the power system, the virtual moment of inertia J and the damping coefficient D of the virtual synchronous generator system are updated in real time.

[0009] Further, in S3, the undetermined value D1 of the damping coefficient D is updated adaptively with the operating state, and the relationship is as follows:

[0010]

[0011] where D0 is the initial value of the damping coefficient D, ω, ω n , Δω are the angular speed, rated angular speed and angular speed difference of the virtual synchronous generator, k1, k2 are two adjustment coefficients, k2 is a normal number, and k1 has the following expression:

[0012]

[0013] Where a is the damping coefficient controlling the initial threshold;

[0014] Determine the value of D1. When its value is within the range of the damping coefficient D, update the value of D, that is, let D = D1.

[0015] Furthermore, the range of values ​​for the damping coefficient is:

[0016]

[0017] Among them, P max P min These are the upper and lower limits of the permissible active power output of the virtual synchronous generator, ω. max ω min These are the upper and lower limits of the angular velocity of the virtual synchronous generator, respectively.

[0018] Furthermore, in S3, the virtual moment of inertia J is updated using a backpropagation (BP) neural network algorithm. The BP neural network includes an input layer, a hidden layer, and an output layer. The input layer includes two input quantities, ω and dω / dt. The hidden layer includes m neurons. The output layer outputs the updated value of the virtual moment of inertia J. The update steps are as follows:

[0019] S31 determines the initial value J0 of the virtual rotational inertia, imports the input variables, and denotes the i-th neuron in the input layer as x. i ;

[0020] S32 calculates h of the i-th neuron in the hidden layer. i input value and output value The calculation process is as follows:

[0021]

[0022] Where n is the number of neurons in the input layer. b1 is the hidden layer weight, f is the hidden layer bias, and f is the hidden layer weight. in (·) is the activation function of the hidden layer;

[0023] S33 calculates the input value y of neuron y in the output layer. in and output value y out The calculation process is as follows:

[0024]

[0025] in, b1 is the output layer weight, b2 is the output layer bias, and f is the output layer weight. outis an activation function of the output layer, J' is a to-be-updated value of the virtual moment of inertia J;

[0026] S34 when J' is within the value range of the virtual moment of inertia J, updating the value of the virtual moment of inertia J of the virtual synchronous generator, that is, J = J';

[0027] S35 repeating the above steps until a set maximum number of iterations is reached.

[0028] Further, in the step S33, the neural network is trained using the gradient descent method to iterate the weight and bias of the output layer, and the iteration formula is as follows:

[0029]

[0030] wherein k is the number of neurons of the output layer, w out′ and b2' are the to-be-updated values of the weight and bias of the output layer after iteration, η is a learning rate, E(k) is a performance index function, O(k) is an output value of the neuron of the output layer, and O1(k) is an expected output value of the neuron of the output layer;

[0031] When E(k) ≥ 1, the weight and bias of the output layer are updated, that is:

[0032]

[0033] Further, in the step S32, the activation function f in of the hidden layer is selected as the hyperbolic tangent function, and the expression is as follows:

[0034]

[0035] Further, in the step S33, the activation function f out of the output layer is selected as the sigmoid function, and the expression is as follows:

[0036]

[0037] Further, in the step S34, the value range of the virtual moment of inertia J is:

[0038]

[0039] Further, in the step S1, the virtual rotor motion equation of the virtual synchronous generator is:

[0040]

[0041] wherein P m , P erespectively are the mechanical power and electromagnetic power of the virtual synchronous generator, and θ, θ n respectively are the operating phase angle and the rated phase angle of the virtual synchronous generator.

[0042] Further, in the step S2, the system closed-loop transfer function G v is:

[0043]

[0044] As can be seen from the formula (2), the virtual synchronous generator system is a second-order system, and in the formula (11), there are:

[0045]

[0046] Wherein, ω N is the undamped natural oscillation angular frequency, ξ is the damping ratio, U and U grid respectively are the output voltage and the AC bus voltage, and X is the line inductance, so that the overshoot σ% and the recovery time Ts for evaluating the transient performance of the second-order system can be obtained, and the expressions are as follows:

[0047]

[0048] The method has the advantages that the frequency control method based on the unmanned intelligent power station island micro-grid can adaptively adjust the damping coefficient, control the virtual inertia which is difficult to control by using the BP neural network algorithm, improve the efficiency and effect of control, effectively suppress the frequency oscillation, reduce the overshoot, and improve the stability and dynamic performance of the system. BRIEF DESCRIPTION OF DRAWINGS

[0049] The application will be further described below in combination with the drawings and embodiments:

[0050] Figure 1 is a specific step flowchart of the method of the application;

[0051] Figure 2 is a virtual inertia updating flowchart of the application;

[0052] Figure 3 is a BP neural network algorithm structure diagram of the application. DETAILED DESCRIPTION

[0053] Embodiment 1

[0054] As shown in the drawings, a frequency control method of an island micro-grid is provided, and the specific steps are as follows: Figure 1

[0055] S1, according to the transient motion equation of the synchronous generator, the virtual rotor motion equation of the virtual synchronous generator is established as follows: ​

[0056]

[0057] where P m , P e are the mechanical power and the electromagnetic power of the VSG, respectively, θ, θ n are the operating phase angle and the rated phase angle of the VSG, respectively, J, D are the virtual moment of inertia and the damping coefficient, respectively.

[0058] The expression of P e in the inverter system S2 is:

[0059]

[0060] where U and U grid are the output voltage and the AC bus voltage, respectively, X is the line reactance, and δ is the phase angle of the VSG system.

[0061] Therefore, the closed-loop transfer function G v of the VSG system is:

[0062]

[0063] Therefore, we have:

[0064]

[0065] where ω N is the undamped natural angular frequency of oscillation, and ξ is the damping ratio. It can be seen that the frequency stability of the system is related to the virtual moment of inertia J and the damping coefficient D. The overshoot σ% and the recovery time T s , which are used to evaluate the transient performance of the second-order system, are written as:

[0066]

[0067] According to the VSG control method designed by the University of Leuven, the value range of J and D is given as:

[0068]

[0069] where P max , P min are the upper and lower limits of the active power output allowed by the VSG, ω max , ω min are the upper and lower limits of the VSG angular velocity.

[0070] Step three, according to the current operating state of the power system, the virtual moment of inertia J and the damping coefficient D of the virtual synchronous generator system are updated in real time. For the damping coefficient D, D will change adaptively with the operating state, and the relationship is as follows:

[0071]

[0072] Where, ω, ω n , Δω are the angular speed, rated angular speed and angular speed difference of the virtual synchronous generator, k1, k2 are two adjustment coefficients, k2 is a constant, and k1 has the following expression:

[0073]

[0074] Where, a is the initial threshold value of the damping coefficient control;

[0075] Determine the value of D1, when its value is in the value range of the damping coefficient D, update the value of D, that is, let D = D1;

[0076] As shown in Figure 2 and 3 , for the virtual moment of inertia J, the BP neural network algorithm is used for updating, the BP neural network includes input layer, hidden layer and output layer, the input layer includes two input quantities ω and dω / dt, the hidden layer includes m neurons, and the output layer outputs the updated value of the virtual moment of inertia J, and the updating steps are as follows:

[0077] S31 determines the initial value J0 of the virtual moment of inertia, imports the input variable, and the i-th neuron in the input layer is denoted as x i ;

[0078] S32 calculates the input value and the output value of the i-th neuron h i in the hidden layer, and the calculation process is as follows:

[0079]

[0080] Where, n is the number of input layer neurons, is the weight of the hidden layer, b1 is the bias of the hidden layer, and f in (·) is the activation function of the hidden layer;

[0081] S33 calculates the input value y in and the output value y out of the neuron y in the output layer, and the calculation process is as follows:

[0082]

[0083] Where, For the output layer weight, b2 is the output layer bias, f out (·) is the activation function of the output layer, J' is the to-be-updated value of the virtual moment of inertia J;

[0084] S34 when J' is within the value range of the virtual moment of inertia J, the value of the virtual moment of inertia J is updated, that is, J = J';

[0085] S35 repeat the above steps until the maximum number of iterations is reached;

[0086] Further, the neural network is trained using the gradient descent method, and the weights and biases of the output layer are iterated, and the iteration formula is as follows:

[0087]

[0088] Wherein, k is the number of neurons in the output layer, w out′ , b2' is the to-be-updated value of the output layer weight and bias after iteration, η is the learning rate, E(k) is the performance index function, O(k) is the output value of the output layer neuron, O1(k) is the expected value of the output layer neuron output;

[0089] The size of E(k) is judged, when E(k)≥1, the weight and bias of the output layer are updated, that is:

[0090]

[0091] In the preferred scheme, the activation function f in (·) of the hidden layer is selected as the hyperbolic tangent function, and its expression is:

[0092]

[0093] The activation function f out (·) of the output layer is selected as the sigmoid function, and its expression is:

[0094]

[0095] The above embodiments are only preferred technical solutions of the present application, and should not be regarded as limiting the present application, and the protection scope of the present application should be the technical solutions recited in the claims, including the equivalent replacement solutions of the technical features recited in the claims. That is, within this range, equivalent replacement improvements are also within the protection scope of the present application.

Claims

1. A method for islanded microgrid frequency control, characterized in that: Comprising the following steps: S1, according to the transient motion equation of the synchronous generator, the virtual rotor motion equation of the virtual synchronous generator is established; S2, the transfer function of the system is obtained by using the virtual rotor motion equation of the virtual synchronous generator, the frequency dynamic stability is studied, and the influence of the virtual moment of inertia J and damping coefficient D on the system is analyzed; S3, updating the virtual moment of inertia of the virtual synchronous generator system in real time according to the current operating state of the power system J and damping coefficient D ; Virtual moment of inertia J The BP neural network algorithm is used for updating, the BP neural network comprises an input layer, a hidden layer and an output layer, the input layer comprises two input quantities, the hidden layer comprises a plurality of neurons, and the output layer outputs an updated value of the virtual moment of inertia. The updating steps are as follows. m J ​​ S31 determining an initial value of the virtual moment of inertia J 0, import input variables, the first i neuron in the input layer is recorded as x i ; S32 calculates the input value and the output value of the i-th neuron in the hidden layer i h i The calculation process is as follows:​​​ （1）； wherein, n is the input layer neuron number, is the hidden layer weight, b 1 is the hidden layer bias, f in (·) is the activation function of the hidden layer; S33 calculating input values of neurons in an output layer y and output values of the neurons in the output layer The calculation process is as follows: （2）； wherein, is an output layer weight, b 2 is an output layer bias, f out (·) is an activation function of the output layer, is a virtual moment of inertia J to be updated; S34 when within the value range of the virtual moment of inertia J of the virtual synchronous generator, the value of the virtual moment of inertia J of the virtual synchronous generator is updated, i.e. ; S35 repeat the above steps until the set maximum iteration number is reached.

2. The islanded microgrid frequency control method of claim 1, wherein: The damping coefficient in the S3 D Pending value D 1 will be updated adaptively with the operating state, and the relationship is as follows: （3）； wherein D 0 is a damping coefficient D an initial value of respectively an angular speed, a rated angular speed and an angular speed difference of the virtual synchronous generator, k 1, k 2 are two regulation coefficients, k 2 is a normal number, k 1 has the following expression: （4）； wherein, a is a damping coefficient control initial threshold value; determining D 1 is greater than the value of the damping coefficient D 1, the value of the damping coefficient D 1 is updated, i.e. the value of 1 is set to 3. The islanded microgrid frequency control method of claim 2, wherein: The damping coefficient value range is: （5）； wherein, P max , P min are respectively the upper and lower limit values of the active power allowed to be output by the virtual synchronous generator, are respectively the upper and lower limit values of the angular speed of the virtual synchronous generator.

4. The islanded microgrid frequency control method of claim 1, wherein: In step S33, the neural network is trained using the gradient descent method, and the weights and biases of the output layer are iterated, and the iteration formula is as follows: （6）； wherein, k is the number of neurons of the output layer, is the updated value of the output layer weight and bias after iteration, is the learning rate, E ( k ) is a performance index function, O ( k ) is an output value of the output layer neuron, O 1( k ) is an expected value of the output layer neuron output; right E ( k The size of ) is used to determine the value. At that time, update the weights and biases of the output layer, that is: （7）。 5. The islanded microgrid frequency control method of claim 1, wherein: In the step S32, the activation function of the hidden layer f in (·) The hyperbolic tangent function is selected, and its expression is: （8）。 6. The islanded microgrid frequency control method of claim 1, wherein: In step S33, the output layer activation function f out (·) sigmoid function is selected, and its expression is: （9）。 7. The islanded microgrid frequency control method of claim 1, wherein: In step S34, the virtual inertia J is set to a value in the range （10）。 8. The islanded microgrid frequency control method of claim 1, wherein: In step S1, the virtual rotor motion equation of the virtual synchronous generator is: （11）； wherein, P m , P e are the virtual synchronous generator mechanical power and electromagnetic power, respectively, are the virtual synchronous generator operating phase angle and rated phase angle, respectively.

9. The islanded microgrid frequency control method of claim 1, wherein: The system closed-loop transfer function in step S2 is G v is: （12）； As can be seen from formula (12), the virtual synchronous generator system is a second-order system, and formula (12) has: （13）； wherein, is the undamped natural angular frequency, is the damping ratio, U and U grid are the output voltage and the AC bus voltage, respectively, X is the line reactance, from which the overshoot and the recovery time Ts, which are indices for evaluating the transient performance of a second-order system, can be obtained, and are expressed as follows: （14）。

Images