Smart power grid dispatching optimization method and system

The smart grid dispatch optimization method, which combines LSTM prediction model and rime optimization algorithm, solves the problem that traditional grid dispatch methods cannot effectively consider complexity and uncertainty, and realizes efficient, safe and environmentally friendly power supply of the grid.

CN118970895AInactive Publication Date: 2024-11-15HUAIYIN INSTITUTE OF TECHNOLOGY
View PDF 6 Cites 0 Cited by

Patent Information

Application Number
CN202411003677.0
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Filing Date
2024-07-25
Publication Date
2024-11-15
Estimated Expiration
Not applicable · inactive patent

AI Technical Summary

Technical Problem

Traditional power grid dispatching methods cannot fully consider the complexity and uncertainty of the power system, resulting in poor dispatching performance.

Method used

A smart grid dispatch optimization method combining an LSTM prediction model and a rime optimization algorithm is proposed. By monitoring voltage, frequency, and power angle in real time, the method uses a training dataset for accurate prediction and optimization, thereby achieving efficient operation and quality optimization of the power grid.

Benefits of technology

It improves the power supply efficiency and security of the power grid, reduces equipment costs, reduces environmental pollution, conforms to the trend of green energy development, and ensures the stable operation of the power grid.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN118970895A_ABST
    Figure CN118970895A_ABST
Patent Text Reader

Abstract

The invention discloses an intelligent power grid dispatching optimization method and system. Belongs to the field of power grid AI testing and comprises a regulator module, a sensor module, a data processing module and an AI system module, the regulator module regulates required parameters, provides data required by projects and enables equipment power operation to be optimal according to instructions of a central control module, and a training data set and data collected by the sensor module jointly act to achieve optimal power operation. In the data processing module, analysis and processing are carried out through rime algorithm optimization and an LSTM prediction model, data credibility measurement is tested, and an optimal AI regulation and control strategy is obtained by detecting the data credibility of voltage stability, frequency stability and power angle stability of the AI-assisted dispatching power grid; compared with traditional power grid dispatching, scientific information support and refined equipment management can be provided, the AI utilization safety is improved, the power supply efficiency is improved, the equipment cost is reduced, and wide application prospects are achieved.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The present invention belongs to the field of high-quality power supply of power grids, and relates to a power grid dispatch optimization method - AI test data credibility measurement to evaluate power grid power supply quality; specifically, it relates to a smart grid dispatch optimization method and system. Background Art

[0002] The technical background of using AI test data to optimize power grid dispatching is based on the application of artificial intelligence and data analysis. Power grid dispatching refers to the reasonable arrangement of the allocation and dispatching of power resources according to the load demand and power generation capacity of the power system to ensure the stable operation of the power grid and the reliability of power supply. Traditional power grid dispatching methods are usually based on empirical rules and mathematical models, but due to the complexity and uncertainty of the power system, these methods often fail to fully consider the combined impact of various factors, resulting in poor dispatching results. The method of using AI test data to optimize power grid dispatching can more accurately predict the load demand, power generation capacity and various potential fault conditions of the power system by collecting and analyzing a large amount of power system operation data and combining artificial intelligence algorithms, thereby optimizing power grid dispatching. Specifically, AI test data can include historical load data, weather data, equipment operation status data, etc. By deep learning and model training of these data, prediction models and optimization models of the power system can be established. These models can help power grid dispatchers more accurately predict future load demand and power generation capacity, and adjust and optimize according to real-time data to achieve efficient operation and quality optimization of the power grid. In summary, the technical background of using AI test data to optimize power grid dispatching is based on artificial intelligence and data analysis methods. By collecting and analyzing a large amount of power system operation data, combined with deep learning and model training, accurate prediction and optimization of power grid dispatching can be achieved, and the quality and reliability of the power grid can be improved. Summary of the invention

[0003] In view of the above problems, the purpose of the present invention is to provide a smart grid scheduling optimization method and system, provide scientific information support and refined equipment management, improve power supply efficiency, increase equipment safety, and reduce equipment costs.

[0004] The technical solution of the present invention is: a smart grid dispatch optimization method described in the present invention comprises the following steps:

[0005] (1) The regulator module adjusts the required parameters according to the instructions of the central control module and provides the data required for the project;

[0006] (2) Obtain preliminary internal environmental data of the power grid dispatch device and pre-process the environmental data;

[0007] (3) Build an LSTM prediction model based on the acquired environmental data; use the rime algorithm, use the training data set, and take the minimum error as the optimization goal to obtain the optimal AI control strategy;

[0008] (4) Based on the obtained optimized AI control strategy, the equipment adjustment is controlled again to achieve control of the AI ​​power grid dispatch.

[0009] Furthermore, in step (2), the environmental data includes voltage, frequency and isoelectric angle.

[0010] Furthermore, in step (3), the LSTM prediction model establishment process is as follows:

[0011] Use LSTM to model each modal component separately;

[0012] The LSTM model is mainly composed of forget gate, input gate, memory gate and output gate;

[0013] The forget gate is used to selectively forget the AI ​​power grid data input by the previous node, the input gate is used to input the AI ​​power grid data, the memory gate is used to selectively remember the input AI power grid data, and the output gate is used to output the AI ​​power grid data; the calculation formula is as follows:

[0014] f t =σ(W f ·[x t ,h t-1 ]+b f )

[0015] i t =σ(W i ·[x t ,h t-1 ]+b i )

[0016] C t =tanh(W c ·[x t ,h t-1 ]+b c )

[0017] C t =f t ·C t-1 +i t ·C t

[0018] O t =σ(W o ·[x t ,h t-1 ]+b o )

[0019] ht =O t tanh(C t )

[0020] In the formula, f t 、i t , C t , O t is the information update process of the forget gate, input gate, memory gate, and output gate, W f , W i , W c , W o Its corresponding weight matrix, b f 、b i 、b c 、b o is the corresponding bias constant; C t is a temporary state; σ is the sigmoid function; tanh is the hyperbolic tangent function; x t It is the voltage, frequency and power angle data of the current state; h t-1 is the voltage, frequency, and power angle data of the previous node; h t The output voltage, frequency, and power angle data are obtained; through the LSTM model, an accurate prediction of the AI-controlled power grid dispatching volume within a certain period of time is obtained.

[0021] Furthermore, the objective function in step (3) is:

[0022]

[0023] Where n is the number of samples; X obs,i is the true value; X model,i is the predicted value.

[0024] Furthermore, in step (3), the specific steps of using the rime optimization algorithm are:

[0025] (3.1), initialization parameters and population;

[0026] The rime population R consists of n sprays S i Each fog agent consists of d rime particles x ij Therefore, the rime population can also be directly used ij Represented; the formula is as follows:

[0027]

[0028] In the formula, i is the ordinal number of the aerosol, and j is the ordinal number of the aerosol particle;

[0029] (3.2), bring the initial population into the objective function for evaluation;

[0030] Each rime population is made to generate a corresponding set of optimal weight matrices and bias constants. The objective function formula is as follows:

[0031]

[0032] Where n is the number of samples; X obs,i is the true value; X model,i is the predicted value;

[0033] (3.3), using the step-by-step rime search strategy to update the population;

[0034] The formula for the position of rime particles is as follows:

[0035]

[0036] In the formula, is the new position of the particle after update, that is, the optimal weight matrix and bias constant; P best,j is the jth particle of the best fogger in the rime population; r1 is a random number uniformly distributed between [0,1]; cosθ controls the movement direction of the particle;

[0037] As the number of iterations changes, the formula becomes as follows:

[0038]

[0039] In the formula, t is the current number of iterations, and T is the maximum number of iterations of the algorithm;

[0040] β is the environmental factor, which can simulate the influence of the external environment as the number of iterations changes to ensure the convergence of the algorithm. The formula is as follows:

[0041]

[0042] Where β is the step function; [·] indicates rounding; the default value of ω is 5, which is used to control the number of steps of the step function;

[0043] h is the adhesion force, which is a random number uniformly distributed between [0,1] and is used to control the distance between the centers of two rime particles;

[0044] Ub ij and Lb ij are the upper and lower bounds of the escape space, respectively limiting the effective area of ​​particle movement; E is the attachment coefficient, which affects the condensation probability of the aerosol and increases with the number of iterations. The formula is as follows:

[0045]

[0046] r2 is a random number within the range, and together with it controls whether the particles condense, that is, whether the particle positions are updated;

[0047] (3.4) Adopt frost puncture mechanism to achieve effective information exchange between individuals;

[0048] This mechanism can be used to update the algorithm, thereby realizing the exchange of algorithm particles, improving the convergence of the algorithm and the ability to escape from the local optimum; the particle position update formula is as follows:

[0049]

[0050] In the formula, F normr (S i ) is the normalized value of the current fog fitness value, indicating the chance of the i-th fog being selected; r3 is a random number uniformly distributed between [0,1];

[0051] (3.5) Adopt the positive greedy selection mechanism to retain the best solution and eliminate the inferior solution;

[0052] Compare the fitness value of the fog agent after the update with the fitness value of the fog agent before the update. If the fitness value after the update is better than the fitness value before the update, a replacement occurs, and the solutions of the two fog agents are replaced at the same time;

[0053] (3.6) Determine whether the termination condition is met and obtain the required optimal weight matrix and bias constant.

[0054] Furthermore, in order to prevent the rime population from falling into the local optimum when selecting the search space, an improvement is proposed. The improvement of the rime algorithm is to improve (3.2) by adopting a learning strategy based on chaos reverse to enhance the diversity of the population;

[0055] The improved formula is as follows:

[0056] R0=Ub ij +Lb ij -R

[0057] In the formula, R0 is the reverse solution corresponding to R.

[0058] Further, a smart grid dispatch optimization system includes a central control module, a regulator module, a sensor module, a data processing module and a final AI system module;

[0059] Among them, the regulator module adjusts the required parameters according to the instructions of the central control module, provides the data required for the project, and optimizes the power operation of the equipment to improve power supply efficiency, improve power supply quality, reduce equipment costs, and increase equipment safety;

[0060] The sensor module is installed in the device and can monitor the stability of three key parameters: voltage, frequency and power angle in real time;

[0061] The training data set works together with the data collected by the sensor module, and is analyzed and processed in the data processing module through rime algorithm optimization and LSTM prediction model, and the data credibility is tested. The optimal AI control strategy is obtained by detecting the data credibility of voltage stability, frequency stability, and power angle stability of the AI-assisted dispatching power grid.

[0062] The sensor module includes a voltmeter, a frequency meter and a power angle meter; the sensor module is installed inside the device to detect environmental data.

[0063] The regulator module of the device includes: a generator set switch control device for controlling the generator set to start or stop running to adjust the power supply;

[0064] A camera switch control device is used to control the operating state of the camera to maintain the voltage stability of the power system;

[0065] A power supply controller for controlling the amount of power generated;

[0066] The grid switch control device is used to control the switching operation of the grid to ensure the safe operation of the power system.

[0067] Furthermore, the construction process of the training data set is as follows:

[0068] 1) Determine your needs: First, you need to clarify your machine learning task (such as classification, regression, clustering, etc.) and the goals you want the model to achieve. This will help you determine the data types and features required;

[0069] 2) Data collection: Collect relevant data according to task requirements; these data can come from public data sets, existing data warehouses, web crawlers, sensors, experimental data, etc.

[0070] 3) Data cleaning: Raw data often contains noise, errors and irrelevant information. Data cleaning includes removing duplicate records, correcting errors, processing missing values, removing irrelevant data, etc. to improve data quality;

[0071] 4) Data transformation: converting data into a format suitable for model input, which may include data standardization, normalization, encoding categorical variables, creating new feature-derived variables, etc.

[0072] 5) Data labeling: For tasks that require supervised learning, such as classification or regression, it is necessary to provide labels or target values ​​for the dataset; data labeling can be completed through expert review, automatic labeling algorithms, or crowdsourcing;

[0073] 6) Data partitioning: Divide the data set into training set and test set; the training set is used to train the model, while the test set is used to evaluate the model performance; common data partitioning methods are to use 70-30, 80-20 or other ratios for partitioning;

[0074] 7) Data enhancement: In order to improve the generalization ability of the model, the diversity of training data can be increased through data enhancement techniques; for example, transformations such as rotation, scaling, and cropping can be applied to image data;

[0075] 8) Data sampling: As needed, the data set may need to be sampled to reduce the amount of data or improve the representativeness of certain categories;

[0076] 9) Data storage and management: Store the processed data in an appropriate database or file system and ensure the integrity and accessibility of the data;

[0077] 10) Data verification: Before starting model training, verify the data set to ensure that the data quality meets the requirements of model training.

[0078] The beneficial effects of the present invention are: 1. This strategy can reduce environmental pollution and conform to the development trend of green energy; 2. This plan uses the relatively new rime optimization algorithm to obtain the best environmental parameters, adjust the equipment, effectively reduce the cost loss, and provide a feasible solution for the sustainable development of the product; 3. The real-time data monitoring of the sensor module and the intelligent control of the central control module ensure the power supply efficiency of the generator and reduce the fluctuations caused by human and external factors. At the same time, the system can automatically adjust according to different environmental parameters, so that the power grid can function normally, reduce costs, increase safety, and improve quality. BRIEF DESCRIPTION OF THE DRAWINGS

[0079] Figure 1 It is the structural diagram of this system;

[0080] Figure 2 It is a flow chart of the rime algorithm in the present invention. DETAILED DESCRIPTION

[0081] The specific technical scheme of the present invention is further described in detail below with reference to specific examples.

[0082] As shown in the figure, a smart grid dispatch optimization system described in the present invention includes a central controller module, a regulator module, a sensor module, a data processing module and a final AI system module;

[0083] Among them, the regulator module adjusts the required parameters according to the instructions of the central control module, provides the data required for the project, and optimizes the power operation of the equipment to improve power supply efficiency, improve power supply quality, reduce equipment costs, and increase equipment safety;

[0084] The sensor module is installed in the device and can monitor the stability of three key parameters: voltage, frequency and power angle in real time;

[0085] The training data set works together with the data collected by the sensor module, and is analyzed and processed in the data processing module through rime algorithm optimization and LSTM prediction model, and the data credibility is tested. The optimal AI control strategy is obtained by detecting the data credibility of voltage stability, frequency stability, and power angle stability of the AI-assisted dispatching power grid.

[0086] The sensor module includes a voltmeter, a frequency meter and a power angle meter; the sensor module is installed inside the device to detect environmental data; AI collects these relevant data for LSTM prediction, conducts training data sets and performs precise analysis through the rime optimization algorithm, and gives the optimal equipment operation strategy based on the real-time data measured by the sensor module and the data pre-stored in the CPU to calculate the optimal environmental factor index with the smallest current regulation error, and then performs reasonable regulation and intervention through the central control module and adjustment equipment, ultimately realizing AI-optimized power grid scheduling.

[0087] The regulator module of the equipment includes: a generator set switch control device, which is used to control the generator set to start or stop running to adjust the power supply;

[0088] A camera switch control device is used to control the operating state of the camera to maintain the voltage stability of the power system; a power supply controller is used to control the power generation;

[0089] The grid switch control device is used to control the switching operation of the grid to ensure the safe operation of the power system.

[0090] Among them, the construction process of the training data set is:

[0091] 1) Determine your needs: First, you need to clarify your machine learning task (such as classification, regression, clustering, etc.) and the goals you want the model to achieve; this will help you determine the type of data and features you need;

[0092] 2) Data collection: Collect relevant data according to task requirements; these data can come from public data sets, existing data warehouses, web crawlers, sensors, experimental data, etc.

[0093] 3) Data cleaning: Raw data often contains noise, errors and irrelevant information; data cleaning includes removing duplicate records, correcting errors, processing missing values, removing irrelevant data, etc. to improve data quality;

[0094] 4) Data transformation: converting data into a format suitable for model input, which may include data standardization, normalization, encoding categorical variables, creating new feature-derived variables, etc.

[0095] 5) Data labeling: For tasks that require supervised learning, such as classification or regression, it is necessary to provide labels or target values ​​for the dataset; data labeling can be completed through expert review, automatic labeling algorithms, or crowdsourcing;

[0096] 6) Data partitioning: Divide the data set into training set and test set; the training set is used to train the model, while the test set is used to evaluate the model performance; common data partitioning methods are to use 70-30, 80-20 or other ratios for partitioning;

[0097] 7) Data enhancement: In order to improve the generalization ability of the model, the diversity of training data can be increased through data enhancement techniques; for example, transformations such as rotation, scaling, and cropping can be applied to image data;

[0098] 8) Data sampling: As needed, the data set may need to be sampled to reduce the amount of data or improve the representativeness of certain categories;

[0099] 9) Data storage and management: Store the processed data in an appropriate database or file system and ensure the integrity and accessibility of the data;

[0100] 10) Data verification: Before starting model training, verify the data set to ensure that the data quality meets the requirements of model training.

[0101] Furthermore, a smart grid dispatch optimization method comprises the following steps:

[0102] Step (1): The regulator module adjusts the required parameters according to the instructions of the central control module and provides the data required for the project;

[0103] Step (2): Obtain preliminary internal environmental data (voltage, frequency, power angle, etc.) of the power grid dispatch device and pre-process the environmental data;

[0104] Step (3): Build an LSTM prediction model based on the acquired environmental data; use the rime algorithm, use the training data set, and take the minimum error as the optimization goal to obtain the optimal AI control strategy;

[0105] The establishment process of the LSTM prediction model is as follows:

[0106] Use LSTM to model each modal component separately;

[0107] The LSTM model is mainly composed of a forget gate, an input gate, a memory gate, and an output gate; the forget gate is used to selectively forget the AI ​​power grid data input by the previous node, the input gate is used to input the AI ​​power grid data, the memory gate is used to selectively memorize the input AI power grid data, and the output gate is used to output the AI ​​power grid data;

[0108] The calculation formula is as follows:

[0109] f t =σ(W f ·[x t ,h t-1 ]+b f )

[0110] i t =σ(W i ·[x t ,h t-1 ]+b i )

[0111] C t =tanh(W c ·[x t ,h t-1 ]+b c )

[0112] C t =f t ·C t-1 +i t ·C t

[0113] O t =σ(W o ·[x t ,h t-1 ]+b o )

[0114] h t =O t tanh(C t )

[0115] In the formula, f t 、i t , C t , O t is the information update process of the forget gate, input gate, memory gate, and output gate, W f , W i , W c , W o Its corresponding weight matrix, b f , b i , bc , b o is the corresponding bias constant; C t is a temporary state; σ is the sigmoid function; tanh is the hyperbolic tangent function; x t It is the voltage, frequency and power angle data of the current state; h t-1 is the voltage, frequency, and power angle data of the previous node; h t The output voltage, frequency, and power angle data are obtained through the LSTM model, and the accurate prediction of the AI-controlled power grid dispatching quantity within a certain period of time is obtained;

[0116] Step (4): Based on the obtained optimized AI control strategy, the equipment adjustment is controlled again to achieve control of the AI ​​power grid dispatch.

[0117] The objective function in step (3) is:

[0118]

[0119] Where n is the number of samples; X obs,i is the true value; X model,i is the predicted value.

[0120] like Figure 2 As shown, the specific steps of the rime optimization algorithm adopted by the present invention are:

[0121] (3.1), initialization parameters and population;

[0122] The rime population R consists of n sprays S i Each fog agent consists of d rime particles x ij composition;

[0123] Therefore, the rime population can also be directly used x ij Represented; the formula is as follows:

[0124]

[0125] In the formula, i is the ordinal number of the aerosol, and j is the ordinal number of the aerosol particle;

[0126] (3.2), bring the initial population into the objective function for evaluation;

[0127] Each rime population is made to generate a corresponding set of optimal weight matrices and bias constants. The objective function formula is as follows:

[0128]

[0129] Where n is the number of samples; X obs,i is the true value; X model,i is the predicted value;

[0130] (3.3), using the step-by-step rime search strategy to update the population;

[0131] The formula for the position of rime particles is as follows:

[0132]

[0133] In the formula, is the new position of the particle after update, that is, the optimal weight matrix and bias constant; P best,j is the jth particle of the best fogger in the rime population; r1 is a random number uniformly distributed between [0,1]; cosθ controls the direction of particle movement; as the number of iterations changes, the formula is as follows:

[0134]

[0135] In the formula, t is the current number of iterations, and T is the maximum number of iterations of the algorithm;

[0136] β is the environmental factor, which can simulate the influence of the external environment as the number of iterations changes to ensure the convergence of the algorithm. The formula is as follows:

[0137]

[0138] Where β is the step function; [·] indicates rounding; the default value of ω is 5, which is used to control the number of steps of the step function;

[0139] h is the adhesion force, which is a random number uniformly distributed between [0,1] and is used to control the distance between the centers of two rime particles;

[0140] Ub ij and Lb ij are the upper and lower bounds of the escape space, respectively limiting the effective area of ​​particle movement; E is the attachment coefficient, which affects the condensation probability of the aerosol and increases with the number of iterations. The formula is as follows:

[0141]

[0142] r2 is a random number within the range, and together with it controls whether the particles condense, that is, whether the particle positions are updated;

[0143] (3.4) Adopt frost puncture mechanism to achieve effective information exchange between individuals;

[0144] This mechanism can be used to update the algorithm, thereby realizing the exchange of algorithm particles, improving the algorithm's convergence and ability to escape from local optimality;

[0145] The particle position update formula is as follows:

[0146]

[0147] In the formula, F normr (S i ) is the normalized value of the current fog fitness value, indicating the chance of the i-th fog being selected; r3 is a random number uniformly distributed between [0,1];

[0148] (3.5) Adopt the positive greedy selection mechanism to retain the best solution and eliminate the inferior solution;

[0149] Compare the fitness value of the fog agent after the update with the fitness value of the fog agent before the update. If the fitness value after the update is better than the fitness value before the update, a replacement occurs, and the solutions of the two fog agents are replaced at the same time;

[0150] (3.6) Determine whether the termination condition is met and obtain the required optimal weight matrix and bias constant.

[0151] Furthermore, in order to prevent the rime population from falling into the local optimum when selecting the search space, an improvement is proposed; the improvement of the rime algorithm is to improve (3.2) by adopting a learning strategy based on chaos reverse to enhance the diversity of the population. The improved formula is as follows:

[0152] R0=Ub ij +Lb ij -R

[0153] Among them, R0 is the reverse solution corresponding to R.

[0154] Implementation scenario: The present invention is mainly used to evaluate the power supply quality of power grids. In the actual data processing of evaluating the power supply quality of power grids, this technology can monitor the status of the power grid system in real time and capture subtle signs of external and internal factors by integrating an efficient data acquisition and analysis system. Through precise data analysis and model prediction, this technology can provide timely and accurate power grid dispatching information to power grid staff, helping power grid dispatching staff to control the power grid more easily and intelligently, thereby achieving efficient operation and quality optimization of the power grid, and avoiding potential safety accidents and energy losses.

[0155] Implementation case: The application of the present invention is not limited to evaluating the power supply quality of the power grid, it also has the potential for flexible application in a variety of scenarios; for example, when home smart appliances are working, the operating conditions of home smart appliances are directly related to the safety and stability of home smart appliances, and even more so to the safety and comfort of family members. By applying this technology to home smart appliances, it is possible to monitor the temperature, humidity, air quality and other parameters of the home in real time, and predict the efficient operation of home smart appliances by analyzing the changing trends of these parameters. Once it is predicted that the environment exceeds the preset value or the appliance is damaged, the system can automatically trigger the alarm mechanism to notify the management personnel to take maintenance measures in time to ensure the normal operation of home smart appliances. In addition, this technology can also be applied to other similar devices with AI intelligent control, such as trams, substations, etc., to provide strong support for the safe and efficient operation of the power system.

Claims

1. An optimization method for a smart grid dispatching optimization system, It is characterized in that The optimization steps are as follows: Step (1): The regulator module adjusts the required parameters according to the instructions of the central control module and provides the data required for the project; Step (2): obtaining preliminary internal environmental data of the device for power grid dispatching and preprocessing the environmental data; Step (3): Build an LSTM prediction model based on the acquired environmental data; use the rime algorithm, use the training data set, and take the minimum error as the optimization objective function to obtain the optimal AI control strategy; Step (4): Based on the obtained optimized AI control strategy, the equipment adjustment is controlled again to achieve control of the AI ​​power grid dispatch.

2. The optimization method of a smart grid dispatch optimization system according to claim 1, It is characterized in that In step (2), the environmental data includes voltage, frequency, and power angle.

3. The optimization method of a smart grid dispatch optimization system according to claim 1, It is characterized in that In step (3), the establishment process of the LSTM prediction model is as follows: Use LSTM to model each modal component separately; The LSTM model includes a forget gate, an input gate, a memory gate, and an output gate; Among them, the forget gate is used to selectively forget the AI ​​power grid data input by the previous node. The input gate is used to input AI power grid data; The memory gate is used to selectively memorize the input AI power grid data; The output gate is used to output AI power grid data; The calculation formula is as follows: f t =σ(W f ·[x t ,h t-1 ]+b f ) i t =σ(W i ·[x t ,h t-1 ]+b i ) C t =tanh(W c ·[x t ,h t-1 ]+b c ) C t =f t ·C t-1 +i t ·C t The t =σ(W o ·[x t ,h t-1 ]+b o ) h t =O t ·tanh(C t ) In the formula, f t 、i t , C t , O t is the information update process of the forget gate, input gate, memory gate, and output gate, W f , W i , W c , W o Its corresponding weight matrix, b f , b i , b c , b o is the corresponding bias constant; C t is a temporary state; σ is the sigmoid function; tanh is the hyperbolic tangent function; x t It is the voltage, frequency and power angle data of the current state; h t-1 is the voltage, frequency, and power angle data of the previous node; h t The output voltage, frequency, and power angle data are obtained; through the LSTM model, an accurate prediction of the AI-controlled power grid dispatching volume within a certain period of time is obtained.

4. The optimization method of a smart grid dispatch optimization system according to claim 1, It is characterized in that In step (3), the objective function is: Where n is the number of samples; X obs,i is the true value; X model,i is the predicted value.

5. The optimization method of a smart grid dispatch optimization system according to claim 1, It is characterized in that In step (3), the specific steps of adopting the rime optimization algorithm are: (3.1), initialization parameters and population; The rime population R consists of n sprays S i Each fog agent consists of d rime particles x ij Therefore, the rime population can be directly used ij express; The formula is as follows: In the formula, i is the ordinal number of the aerosol, and j is the ordinal number of the aerosol particle; (3.2), bring the initial population into the objective function for evaluation; Each rime population is made to generate a corresponding set of optimal weight matrices and bias constants. The objective function formula is as follows: Where n is the number of samples; X obs,i is the true value; X model,i is the predicted value; (3.3), using the step-by-step rime search strategy to update the population; The formula for the position of rime particles is as follows: In the formula, is the new position of the particle after update, that is, the optimal weight matrix and bias constant; P best,j is the jth particle of the best fogging agent in the rime population; r 1 is a random number uniformly distributed between [0,1]; cosθ controls the direction of particle movement; As the number of iterations changes, the formula becomes as follows: In the formula, t is the current number of iterations, and T is the maximum number of iterations of the algorithm; β is the environmental factor, which can simulate the influence of the external environment as the number of iterations changes to ensure the convergence of the algorithm. Its formula is as follows: Where β is the step function; [·] indicates rounding; the default value of ω is 5, which is used to control the number of steps of the step function; h is the adhesion force, which is a random number uniformly distributed between [0,1] and is used to control the distance between the centers of two rime particles; Ub ij and Lb ij are the upper and lower bounds of the escape space, respectively limiting the effective area of ​​particle movement; E is the attachment coefficient, which affects the condensation probability of the aerosol and increases with the number of iterations. The formula is as follows: r 2 is a random number within the range, and controls whether the particles condense, that is, whether the particle positions are updated; (3.4) Adopt frost puncture mechanism to achieve effective information exchange between individuals; This mechanism can be used to update the algorithm, realize the exchange of algorithm particles, and improve the convergence of the algorithm and the ability to escape from the local optimum. The particle position update formula is as follows: In the formula, F normr (S i ) is the normalized value of the current fog fitness value, indicating the chance of the i-th fog being selected; r 3 is a random number uniformly distributed between [0,1]; (3.5) Adopt the positive greedy selection mechanism to retain the best solution and eliminate the inferior solution; Compare the fitness value of the fog agent after the update with the fitness value of the fog agent before the update. If the fitness value after the update is better than the fitness value before the update, a replacement occurs, and the solutions of the two fog agents are replaced at the same time; (3.6) Determine whether the termination condition is met and obtain the required optimal weight matrix and bias constant.

6. The optimization method of a smart grid dispatch optimization system according to claim 1, Features ,To avoid the rime population from falling into the local optimum when selecting the search space; the improvement of the rime algorithm is to improve (3.2) and enhance the diversity of the population by adopting a learning strategy based on chaos reverse; The improved formula is as follows: R 0 =Ub ij +Lb ij -R In the formula, R 0 is the reverse solution corresponding to R.

7. A smart grid dispatch optimization system, It is characterized in that It includes interconnected central control module, regulator module, sensor module, data processing module and final AI system module; The regulator module includes: a generator set switch control device, a camera switch control device, a power supply controller and a grid switch control device; The sensor module includes a voltmeter, a frequency meter and a power angle meter; The data processing module includes an LSTM prediction model.

8. A smart grid dispatch optimization system according to claim 7, It is characterized in that A training data set is also connected to the sensor module.

9. A smart grid dispatch optimization system according to claim 8, It is characterized in that The construction process of the training data set is as follows: (9.1) Determine the requirements; (9.2), data collection; (9.3), data cleaning; (9.4), data conversion; (9.5), data annotation; (9.6), data partitioning; (9.7), data enhancement; (9.8), data sampling; (9.9), Data storage and management; (9.10), data verification.

10. A smart grid dispatch optimization system according to claim 9, It is characterized in that In step (9.1), the requirements are specifically determined as follows: first, clarify the machine learning tasks and the goals to be achieved by the model; In step (9.2), the data collection specifically includes: collecting corresponding data according to task requirements, the data comes from public data sets, existing data warehouses, web crawlers, sensors and experimental data; In step (9.3), the data cleaning specifically includes: the original data contains noise, errors and irrelevant information; the data cleaning includes removing duplicate records, correcting errors, processing missing values ​​and removing irrelevant data; In step (9.4), the data conversion specifically includes: converting the data into a format suitable for model input, including data standardization, normalization, encoding categorical variables and creating new feature-derived variables; In step (9.5), the data annotation specifically includes: providing a label or target value for the data set for tasks that require supervised learning; In step (9.6), the data division is specifically: dividing the data set into a training set and a test set; the training set is used to train the model, and the test set is used to evaluate the model performance; a common data division method is to use a 70-30, 80-20 or other ratio for division; In step (9.7), the data enhancement specifically includes: increasing the diversity of training data through data enhancement technology to improve the generalization ability of the model; In step (9.8), the data sampling specifically includes: sampling the data set as needed to reduce the amount of data or improve the representativeness of certain categories; In step (9.9), the data storage and management specifically includes: storing the processed data in an appropriate database or file system, and ensuring the integrity and accessibility of the data; In step (9.10), the data verification specifically includes: before starting the model training, verifying the data set to ensure that the data quality meets the requirements of the model training.

Citation Information

Patent Citations

  • Visual analysis system and method for electric power big data

    CN117114449A

  • SCADA (supervisory control and data acquisition)-based auxiliary decision-making method and system for loop closing reverse power supply control conditions of main and distribution networks

    CN117394529A

  • Short-term wind power prediction method and system considering available installed capacity

    CN118172047A

  • Short-term power load prediction method

    CN118174294A

  • Short-term load interval prediction method

    CN118336705A