An emergency power control-oriented micro-grid regulation threshold intelligent setting method

Abstract

The application discloses a micro-grid regulation threshold intelligent setting method for emergency power control, and comprises the following steps: step 1, constructing a micro-grid emergency power control scene set; step 2, constructing a generative adversarial network model, the generative adversarial network model comprises an emergency power control scene generator and a micro-grid regulation threshold effect discriminator, the micro-grid emergency power control scene set is input into the generative adversarial network model for multi-round adversarial learning, the parameter of a reward function of the emergency power control scene generator is solved by an improved plant cell colony algorithm optimization, thereby obtaining a current micro-grid regulation threshold, and whether the current micro-grid regulation threshold meets the micro-grid emergency power regulation requirement is judged by the micro-grid regulation threshold effect discriminator; if not, the iteration adversarial learning is continued; if yes, the iteration is stopped and the current micro-grid regulation threshold is output. The application can effectively improve the reliability and stability of the micro-grid in a complex power system.

Description

TECHNICAL FIELD

[0001] The application relates to the field of micro-grid regulation methods, and in particular to a micro-grid regulation threshold intelligent setting method for emergency power control. BACKGROUND

[0002] With the development of new power systems and large-scale access of renewable energy, micro-grids are increasingly important in modern power systems. Micro-grids not only improve power quality and power supply reliability, but also provide rapid and accurate responses in emergency situations to ensure the stable operation of the power system. However, the setting of the regulation threshold in the emergency power control of micro-grids is a key issue. A suitable regulation threshold can not only improve the operation efficiency of micro-grids, but also effectively avoid misoperation and ensure the safety and stability of the system. Therefore, it is of great practical significance and academic value to study an intelligent regulation threshold setting method.

[0003] Although existing research has made some progress in the setting of micro-grid regulation thresholds, there are still many limitations. Traditional methods mostly rely on fixed thresholds or empirical values, which are difficult to adapt to different operating conditions, resulting in limited response speed and regulation accuracy of micro-grids in emergency situations, and even may cause misoperation, affecting the stability of the system, which is not satisfactory in dealing with complex and variable power system operating environments. In addition, existing optimization algorithms often have low search efficiency and are prone to local optimization when dealing with micro-grid regulation threshold problems, which are difficult to meet the requirements of high precision and fast response. These limitations significantly affect the response speed and regulation accuracy of micro-grids in emergency situations, restricting their application effect. SUMMARY

[0004] The application provides a micro-grid regulation threshold intelligent setting method for emergency power control to solve the problems of low efficiency and easy to fall into local optimization when existing optimization algorithms deal with micro-grid regulation thresholds in the prior art.

[0005] In order to achieve the above purpose, the technical scheme adopted by the application is as follows:

[0006] A micro-grid regulation threshold intelligent setting method for emergency power control, comprising the following steps:

[0007] Step 1, obtaining power grid emergency power regulation demand information, micro-grid output data, and micro-grid state information, and constructing a micro-grid emergency power control scenario set with the power grid emergency power regulation demand information, the micro-grid output data, and the micro-grid state information;

[0008] Step 2, a generative adversarial network model for micro-grid regulation threshold setting is constructed, the generative adversarial network model includes an emergency power control scene generator and a micro-grid regulation threshold effect discriminator, and the micro-grid emergency power control scene set obtained in step 1 is input into the emergency power control scene generator in the generative adversarial network model to perform multi-round adversarial learning of the generative adversarial network model.

[0009] In each round of adversarial learning, the emergency power control scene generator optimizes the parameters of its own reward function based on the improved plant cell colony algorithm, generates a current micro-grid emergency power control scene and a current micro-grid regulation threshold corresponding to the current micro-grid emergency power control scene under the guidance of the optimized reward function, and the micro-grid regulation threshold effect discriminator discriminates whether the current micro-grid regulation threshold meets the micro-grid emergency power regulation requirement.

[0010] If the discrimination result of the micro-grid regulation threshold effect discriminator is that the current micro-grid regulation threshold cannot meet the micro-grid emergency power regulation requirement, the discrimination result is fed back to the emergency power control scene generator by the micro-grid regulation threshold effect discriminator, and the generative adversarial network model repeats the above process for the next round of adversarial learning.

[0011] If the discrimination result is that the current micro-grid regulation threshold can meet the micro-grid emergency power regulation requirement, the generative adversarial network model stops adversarial learning, and the emergency power control scene generator outputs the current micro-grid regulation threshold, thereby completing the micro-grid regulation threshold setting.

[0012] Further, in step 1, the micro-grid output data includes micro-grid photovoltaic output data and micro-grid wind power output data.

[0013] Further, in step 1, the micro-grid state information includes micro-grid energy storage device state, micro-grid load state and micro-grid adjustable resource state.

[0014] Further, in step 2, noise is superimposed on the micro-grid emergency power control scene set before being input into the emergency power control scene generator, and the micro-grid operation uncertainty is simulated by superimposing noise.

[0015] Further, in step 2, the emergency power control scene generator is composed of multiple layers of fully connected layers and a general activation function. When the microgrid emergency power control scene set is input into the emergency power control scene generator, each layer of fully connected layers of the emergency power control scene generator extracts the features of the microgrid emergency power control scene set, including the scene state and the threshold value, and uses the general activation function to determine whether the extracted features are transmitted to the next layer of fully connected layers. If the result is yes, the features extracted by the current fully connected layer can be transmitted to the next fully connected layer. The transmission is performed successively until the extracted features of the microgrid emergency power control scene set are transmitted to the end of the emergency power control scene generator, and the current emergency power control scene set state and threshold value are obtained.

[0016] Further, in step 2, the microgrid regulation threshold effect discriminator is composed of multiple layers of fully connected layers and a general activation function. When the current emergency power control scene set state and threshold value generated by the emergency power control scene generator are input into the microgrid regulation threshold effect discriminator, each layer of fully connected layers of the microgrid regulation threshold effect discriminator extracts the microgrid operation safety state under the current emergency power control scene set state and threshold value, and uses the general activation function to determine whether the extracted features are transmitted to the next layer of fully connected layers. If the result is yes, the features extracted by the current fully connected layer can be transmitted to the next fully connected layer. The transmission is performed successively until the extracted microgrid operation safety state features are transmitted to the end of the microgrid regulation threshold effect discriminator, and the current microgrid operation safety state features are obtained, and then it is determined whether the current microgrid operation safety state features meet the microgrid operation requirements.

[0017] Further, in step 2, the current microgrid regulation threshold obtained by the emergency power control scene generator includes a regulation power shortage threshold and a regulation delay time.

[0018] Further, in step 2, the reward function of the emergency power control scene generator is as follows:

[0019] R(λ)=ω(λ n -λ n-1 )+b,2≤n≤N

[0020] wherein R(λ) is the reward function of the emergency power control scene generator; n is the sample generated by the emergency power control scene generator for the nth time; ω is the sensitivity coefficient of the emergency power control scene generator. b is the bias coefficient of the emergency power control scene generator; N is the total number of samples generated by the emergency power control scene generator;

[0021] In each round of adversarial learning, the emergency power control scene generator optimizes and solves the sensitivity coefficient ω and the bias coefficient b in the reward function based on the improved plant cell colony algorithm.

[0022] The application proposes an intelligent setting method combining generative adversarial network and improved plant cell colony algorithm. By using the method combining generative adversarial network and improved plant cell colony algorithm, the algorithm parameters can be dynamically adjusted at different stages, the search efficiency is improved, and the local optimal solution is avoided, so as to realize the optimization setting of the control threshold.

[0023] The application solves the limitations of the traditional method in dealing with the complex and changeable power system operation environment. Through the adjustment of dynamic learning factors, the performance of the generative adversarial network at different stages is improved, and the improved plant cell colony algorithm is used for secondary optimization of the generative adversarial network to balance the global exploration ability and local mining ability of the algorithm. In this way, not only the setting accuracy of the control threshold can be significantly improved, the response speed and control accuracy of the microgrid can be improved, the fast response and efficient operation of the microgrid in emergency can be ensured, but also the demand under different operation conditions can be adapted to ensure the stable operation of the system and avoid misoperation. Through the application of the method, the microgrid control threshold can be quickly set, and the reliability and stability of the microgrid in the complex power system can be effectively improved, which provides solid technical support for the intelligent development of the power system. BRIEF DESCRIPTION OF DRAWINGS

[0024] Figure 1 is the principle diagram of the method of the embodiment of the application.

[0025] Figure 2 is the simulation analysis result diagram of the performance of three different methods in the experiment of the embodiment of the application.

[0026] Figure 3 is the simulation analysis result diagram of the optimization performance comparison of the improved plant cell colony algorithm in the experiment of the embodiment of the application. DETAILED DESCRIPTION

[0027] The application will be further described below in combination with the drawings and embodiments.

[0028] As shown in the drawings, Figure 1 The embodiment discloses a microgrid control threshold intelligent setting method for emergency power control, which comprises the following steps:

[0029] Step 1, constructing a microgrid emergency power control scene set

[0030] In this embodiment, the grid emergency power regulation demand information, micro-grid output data and micro-grid state information are obtained from the micro-grid, wherein the micro-grid output data includes micro-grid photovoltaic output data and micro-grid wind power output data, and the micro-grid state information includes micro-grid energy storage device state, micro-grid load state and micro-grid adjustable resource state. Thus, the micro-grid emergency power control scenario set is constructed based on the grid emergency power regulation demand information, micro-grid photovoltaic output data, micro-grid wind power output data, micro-grid energy storage device state, micro-grid load state and micro-grid adjustable resource state.

[0031] Specifically, the grid emergency power regulation demand information is a time sequence power demand vector, as shown in formula (1):

[0032]

[0033] In formula (1), is a time sequence power demand vector; are power demand amounts at time 1, 2, …n, respectively.

[0034] The micro-grid photovoltaic output data can be represented as shown in formula (2):

[0035]

[0036] In formula (2), Y PV is a photovoltaic output data set; are photovoltaic output data at time 1, 2, …n, respectively.

[0037] The micro-grid wind power output data can be represented as shown in formula (3):

[0038]

[0039] In formula (3), Y W is a wind power output data set; are wind power output data at time 1, 2, …n, respectively.

[0040] The micro-grid energy storage device state can be represented as shown in formula (4):

[0041]

[0042] In formula (4), Y ES is a micro-grid energy storage device state vector; are micro-grid energy storage device state of charge collected at time 1, 2, …n, respectively.

[0043] The micro-grid load state can be represented as shown in formula (5):

[0044]

[0045] Y (5) in the formula (5), Y LOAD is a micro-grid load state time sequence vector; respectively, the first, second…n micro-grid load power collected at the collection time.

[0046] The micro-grid adjustable resource state can be expressed as shown in formula (6):

[0047]

[0048] Y (6) in the formula (6), Y FR is a micro-grid adjustable resource state time sequence vector; respectively, the first, second…n micro-grid adjustable resource adjustable power collected at the collection time.

[0049] Therefore, the final micro-grid emergency power control scenario set is

[0050]

[0051] Y (7) in the formula (7), Y WS is a micro-grid emergency power control scenario set; Y DK is a micro-grid regulation threshold.

[0052] Step 2, build a generative adversarial network model for micro-grid regulation threshold setting, the generative adversarial network model includes an emergency power control scenario generator and a micro-grid regulation threshold effect discriminator.

[0053] Superimpose Gaussian noise on the micro-grid photovoltaic output data, micro-grid wind power output data, micro-grid load state, and micro-grid adjustable resource state in the micro-grid emergency power control scenario set obtained in step 1, thereby simulating the uncertainty of the micro-grid operation. Then, input the micro-grid emergency power control scenario set superimposed with Gaussian noise into the emergency power control scenario generator in the generative adversarial network model to perform multi-round adversarial learning of the generative adversarial network model. The adversarial learning process is as follows:

[0054] (1) In each round of adversarial learning, the emergency power control scenario generator optimizes the parameters of its reward function based on the improved plant cell colony algorithm, and generates the current micro-grid emergency power control scenario and the current micro-grid regulation threshold corresponding to the current micro-grid emergency power control scenario under the guidance of the optimized reward function.

[0055] In this embodiment, the emergency power control scene generator is composed of multiple layers of fully connected layers and a general activation function. When the microgrid emergency power control scene set is input into the emergency power control scene generator, each layer of fully connected layers of the emergency power control scene generator extracts the features (including scene state and threshold) of the microgrid emergency power control scene set, and uses the general activation function to determine whether the extracted features are transmitted to the next layer of fully connected layers; if the result is yes, the features extracted by the current fully connected layer can be transmitted to the next fully connected layer; the transmission is performed successively until the extracted features of the microgrid emergency power control scene set are transmitted to the end of the emergency power control scene generator, and the current emergency power control scene set state and threshold are obtained. Moreover, the current microgrid regulation threshold generated by the emergency power control scene generator includes a regulation power shortage threshold and a regulation delay time.

[0056] In this embodiment, the reward function of the emergency power control scene generator is as shown in formula (7):

[0057] R(λ)=ω(λ n -λ n-1 )+b,2≤n≤N

[0058] wherein R(λ) is the reward function of the emergency power control scene generator; n is the sample generated by the emergency power control scene generator for the nth time; ω is the sensitivity coefficient of the emergency power control scene generator. b is the bias coefficient of the emergency power control scene generator; N is the total number of samples generated by the emergency power control scene generator;

[0059] In each round of adversarial learning, the emergency power control scene generator optimizes and solves the sensitivity coefficient ω and the bias coefficient b in the reward function based on the improved plant colony algorithm.

[0060] In this embodiment, the improved plant colony algorithm is obtained by improving the growth position update based on the existing plant colony algorithm. The dynamic growth parameter is added in the growth position update function to realize the dynamic adjustment of the global exploration ability and the local mining ability of the plant colony algorithm when updating the growth position. The optimization and solving process of the improved plant colony algorithm for the sensitivity coefficient ω and the bias coefficient b in the reward function of the emergency power control scene generator is as follows:

[0061] S1, initialize the parameters of the improved plant colony algorithm, including the number of plant cells of the improved plant colony algorithm, the optimization dimension of the improved plant colony algorithm, and the initial position (initial random solution) of the plant cells of the improved plant colony algorithm. The optimization dimension of the improved plant colony algorithm is 2, including the sensitivity coefficient ω in the reward function of the emergency power control scene generator and the bias coefficient b of the emergency power control scene generator.

[0062] S2, a growth position updating function of a plant cell in the improved plant colony algorithm is designed, and the growth position updating function is used to find the parameter optimal solution of the sensitivity coefficient ω and the bias coefficient b of the emergency power control scene generator reward function. The growth position updating function is shown in formula (8):

[0063]

[0064] In formula (8), d are the growth positions of the plant cell in the improved plant colony algorithm at t and t+1 time, respectively; d g is a random growth distance of the plant cell in the improved plant colony algorithm; G c is a dynamic growth constant in the improved plant colony algorithm; is the growth position of the optimal position plant cell at t in the improved plant colony algorithm.

[0065] S3, the updated growth position of the plant cell is used as the output solution of the sensitivity coefficient ω and the bias coefficient b of the emergency power control scene generator reward function at the current iteration round.

[0066] (2) When the emergency power control scene generator generates the current micro-grid regulation threshold value, the micro-grid regulation threshold value effect discriminator discriminates whether the current micro-grid regulation threshold value meets the micro-grid emergency power regulation demand.

[0067] In the embodiment, the micro-grid regulation threshold value effect discriminator is composed of multiple full connection layers and a general activation function. When the current emergency power control scene set state and threshold value generated by the emergency power control scene generator are input into the micro-grid regulation threshold value effect discriminator, each full connection layer of the micro-grid regulation threshold value effect discriminator extracts the micro-grid operation safety state under the current emergency power control scene set state and threshold value, and uses the general activation function to judge whether the extracted features are transmitted to the next full connection layer. If the judgment result is yes, the features extracted by the current full connection layer can be transmitted to the next full connection layer; the transmission is performed successively until the extracted micro-grid operation safety state features are transmitted to the end of the micro-grid regulation threshold value effect discriminator, the current micro-grid operation safety state features are obtained, and then it is judged whether the current micro-grid operation safety state features meet the micro-grid operation demand.

[0068] In this embodiment, if the determination result of the micro-grid regulation threshold effect discriminator is that the current micro-grid regulation threshold cannot meet the emergency power regulation demand of the micro-grid, the determination result is fed back to the emergency power control scene generator by the micro-grid regulation threshold effect discriminator, and the emergency power control scene generator in the generative adversarial network model continues to optimize and solve the parameters of the reward function by improving the plant cell group algorithm, and generates the micro-grid emergency power control scene and the micro-grid regulation threshold corresponding to the micro-grid emergency power control scene under the guidance of the optimized reward function, that is, the above process is repeated to continue the next round of adversarial learning.

[0069] If the determination result is that the current micro-grid regulation threshold can meet the emergency power regulation demand of the micro-grid, the emergency power control scene generator in the generative adversarial network model stops improving the plant cell group algorithm, at this time the generative adversarial network model stops adversarial learning, and the current micro-grid regulation threshold is output by the emergency power control scene generator, thereby completing the setting of the micro-grid regulation threshold.

[0070] In this embodiment, the optimization target of the generative adversarial network model for micro-grid regulation threshold setting is shown in formula (9):

[0071]

[0072] In formula (9), G represents the emergency power control scene generator; D represents the micro-grid regulation threshold effect discriminator; x is the data input to the emergency power control scene generator; z is Gaussian noise; p z () is a normal distribution function; E() is a mathematical expectation function; V is an optimization target function.

[0073] In order to make the purpose, technical scheme and advantages of the present application more clear, the following will further explain the embodiment in combination with experiments.

[0074] The running data of the micro-grid in a certain province in the east of China in 2022 is selected for case analysis. The micro-grid data includes power grid operation data and environmental meteorological data. The generative adversarial network is constructed based on a deep school framework, and the computer hardware for training includes an Intel I9 central processing unit, 32 GB of memory, and a GTX 4070TI GPU. The initial number of plant cells of the improved plant cell group algorithm is 100, the optimization dimension of the improved plant cell group algorithm is 2, and the initial position of the plant cells of the improved plant cell group algorithm is [0, 0]. The maximum number of iterations of the improved plant cell group algorithm is set to 500.

[0075] In order to compare the performance of the method in this embodiment, three different methods are set for simulation analysis in the experimental case analysis, and the results are as follows Figure 2As shown in the figure. Method 1: the method proposed in this embodiment (the proposed method curve in the figure); method 2: setting method based on particle swarm algorithm (PSO algorithm) (the PSO curve in the figure); method 3: setting method based on genetic algorithm (GA algorithm) (the GA curve in the figure). The abscissa in the figure represents the number of emergency power control, and the ordinate represents the completion effect of emergency power control. It can be seen that the micro-grid regulation threshold setting effect based on the method proposed in this embodiment is better, which can better enable the micro-grid to respond to the emergency power control demand of the power grid.

[0076] In order to compare the optimization performance of the improved plant colony algorithm proposed in this embodiment, two different methods are set in the experimental case analysis for simulation analysis, and the results are as shown in the figure. Figure 3 As shown in the figure. Method 1: the method proposed in this embodiment (the improved method curve in the figure); method 2: setting method based on traditional plant colony algorithm (the traditional method curve in the figure). It can be seen that the improved plant colony algorithm proposed in this embodiment can converge faster and better, thereby improving the overall optimization ability of the micro-grid regulation threshold intelligent setting algorithm.

[0077] The preferred embodiments of the present application are described in detail above in combination with the drawings, and the embodiments described in the present application are only used to describe the preferred embodiments of the present application, and do not limit the concept and scope of the present application. In the above specific embodiments, each specific technical feature described above can be combined in any appropriate manner without contradiction, and such combination should also be considered as disclosed by the present disclosure as long as it does not deviate from the technical concept of the present application. In order to avoid unnecessary repetition, the present application will not further describe various possible combinations.

[0078] The present application is not limited to the specific details in the above embodiments, and various modifications and improvements of the technical solutions of the present application made by those skilled in the art within the technical concept of the present application and without departing from the design idea of the present application should fall within the protection scope of the present application. The technical content claimed by the present application has been fully recorded in the claims.

Claims

1. A method for intelligent setting of microgrid regulation threshold for emergency power control, characterized in that, The method comprises the following steps: Step 1, obtaining power grid emergency power regulation demand information, micro-grid output data and micro-grid state information, and constructing a micro-grid emergency power control scene set based on the power grid emergency power regulation demand information, the micro-grid output data and the micro-grid state information; Step 2, constructing a generative adversarial network model for micro-grid regulation threshold setting, the generative adversarial network model comprising an emergency power control scene generator and a micro-grid regulation threshold effect discriminator, and inputting the micro-grid emergency power control scene set obtained in step 1 into the emergency power control scene generator in the generative adversarial network model to perform multi-round adversarial learning of the generative adversarial network model; In each round of adversarial learning, the emergency power control scene generator optimizes and solves parameters of a reward function based on an improved plant cell colony algorithm, generates a current micro-grid emergency power control scene and a current micro-grid regulation threshold corresponding to the current micro-grid emergency power control scene under the guidance of the optimized reward function, and the micro-grid regulation threshold effect discriminator discriminates whether the current micro-grid regulation threshold meets the micro-grid emergency power regulation demand; If the discrimination result of the micro-grid regulation threshold effect discriminator is that the current micro-grid regulation threshold cannot meet the micro-grid emergency power regulation demand, the discrimination result is fed back to the emergency power control scene generator by the micro-grid regulation threshold effect discriminator, and the generative adversarial network model repeats the above process to perform the next round of adversarial learning; If the discrimination result is that the current micro-grid regulation threshold can meet the micro-grid emergency power regulation demand, the generative adversarial network model stops adversarial learning, and the emergency power control scene generator outputs the current micro-grid regulation threshold, thereby completing the micro-grid regulation threshold setting.

2. The method of claim 1, wherein, In step 1, the micro-grid output data comprises micro-grid photovoltaic output data and micro-grid wind power output data.

3. The method of claim 1, wherein the method further comprises: In step 1, the micro-grid state information comprises micro-grid energy storage device state, micro-grid load state and micro-grid adjustable resource state.

4. The method of claim 1, wherein, In step 2, noise is superimposed on the micro-grid emergency power control scene set before being input into the emergency power control scene generator, and the micro-grid operation uncertainty is simulated through the superimposed noise.

5. The method of claim 1, wherein, In step 2, the emergency power control scene generator is composed of multiple fully connected layers and a general activation function, when the micro-grid emergency power control scene set is input into the emergency power control scene generator, each fully connected layer of the emergency power control scene generator extracts features of the micro-grid emergency power control scene set, and the general activation function is used to determine whether the extracted features are transmitted to the next fully connected layer; If the determination result is yes, the features extracted by the current fully connected layer can be transmitted to the next fully connected layer; The features are transmitted successively until the extracted micro-grid emergency power control scene set features are transmitted to the end of the emergency power control scene generator, and the current emergency power control scene set state and threshold are obtained.

6. The method of claim 1, wherein, In step 2, the micro-grid regulation threshold effect discriminator is composed of multiple full connection layers and a general activation function, when the current emergency power control scene set state and threshold generated by the emergency power control scene generator are input into the micro-grid regulation threshold effect discriminator, each full connection layer of the micro-grid regulation threshold effect discriminator extracts the micro-grid operation safety state under the current emergency power control scene set state and threshold, and judges whether the extracted feature is transmitted to the next full connection layer by using the general activation function; If the judgment result is yes, the feature extracted by the current full connection layer can be transmitted to the next full connection layer; The transmission is performed successively until the extracted micro-grid operation safety state feature is transmitted to the end of the micro-grid regulation threshold effect discriminator, the current micro-grid operation safety state feature is obtained, and then it is judged whether the current micro-grid operation safety state feature meets the micro-grid operation requirement.

7. The method of claim 1, wherein, In step 2, the current micro-grid regulation threshold obtained by the emergency power control scene generator includes a regulation power shortage threshold and a regulation delay time. 8.The method of claim 1, wherein, In step 2, the reward function of the emergency power control scene generator is as follows: R(λ) = ω(λ n -λ n-1 )+b,2≤n≤N Wherein: R(λ) is the reward function of the emergency power control scene generator; n is the sample generated by the emergency power control scene generator for the nth time; ω is the sensitivity coefficient of the emergency power control scene generator; b is the bias coefficient of the emergency power control scene generator; N is the total number of samples generated by the emergency power control scene generator; In each round of adversarial learning, the emergency power control scene generator optimizes and solves the sensitivity coefficient ω and the bias coefficient b in the reward function based on the improved plant cell colony algorithm.

Images