An optimized dispatching method for microgrid grid-connected operation based on intelligent optimization algorithm

Abstract

The present invention relates to the field of power grid dispatching technology, and specifically to a microgrid grid-connected operation optimization dispatching method based on an intelligent optimization algorithm. The method comprises the following steps: obtaining target power demand information through historical power demand of an industrial park, establishing a power quality demand analysis model, analyzing characteristic data and calculating corresponding characteristic intervals to obtain first power demand data and second power demand data, so as to obtain power quality requirements in different time intervals; obtaining the power supply capacity of a main power grid and different microgrids, and obtaining corresponding power dispatching parameters; and using a particle swarm algorithm dispatching model with the power quality requirements in different time intervals as the target, and using the power dispatching parameters of the microgrid as the particle swarm to perform optimization calculations to obtain an optimal dispatching solution. The method can perform grid-connected optimization dispatching for different microgrid energy supply conditions while meeting the power quality requirements of the industrial park, thereby improving the accuracy of microgrid dispatching.

Description

Technical Field

[0001] The present invention relates to the technical field of power grid dispatching, and in particular to a microgrid grid-connected operation optimization dispatching method based on an intelligent optimization algorithm. Background Art

[0002] A microgrid is a small power grid facility independent of the main power grid, mainly composed of distributed power sources, distribution facilities, energy storage systems and protection devices; it can generate electricity and store electricity on its own; the microgrid includes a grid-connected operation mode parallel to the main grid and an independent island operation mode; through the self-regulation of the microgrid, when the main grid's power supply is tight, it can cooperate with the main grid for power supply or independently supply power; at the same time, because microgrids mainly use clean energy and renewable energy, they play an important role in economic benefits, environmental protection and energy saving, and emergency power supply.

[0003] Since microgrids are used in a variety of scenarios such as industrial parks, hospitals, and residential areas, the scheduling strategies of microgrids vary according to the actual conditions of different scenarios. For industrial parks with more complex power systems, in order to achieve reasonable scheduling of grid-connected operations between microgrids and the main grid, current technologies use intelligent optimization algorithms to achieve optimal scheduling strategies.

[0004] Current intelligent optimization algorithms, such as particle swarm optimization and genetic optimization, all use the power parameters and cost parameters in the microgrid system as the main influencing parameters to optimize the scheduling of the microgrid. However, the problem is that the power quality in industrial scenarios has high requirements for industrial scenarios, and the microgrid contains a variety of distributed energy sources. If only unified power parameters and cost parameters are used as targets for optimization and scheduling, poor power transmission quality is likely to occur, resulting in reduced accuracy in microgrid optimization and scheduling.

[0005] Therefore, a microgrid grid-connected operation optimization scheduling method based on intelligent optimization algorithm is proposed. Summary of the Invention

[0006] The purpose of the present invention is to provide a microgrid grid-connected operation optimization scheduling method based on an intelligent optimization algorithm, which obtains target power demand information through the historical power demand of the industrial park, establishes a power quality demand analysis model, analyzes characteristic data and calculates the corresponding characteristic interval to obtain first power demand data and second power demand data, so as to obtain power quality requirements in different time intervals; obtains the power supply capacity of the main grid and different microgrids, and obtains corresponding power scheduling parameters, and uses the particle swarm algorithm scheduling model to take the power quality requirements in different time intervals as the target, and uses the power scheduling parameters of the microgrid as the particle swarm to perform optimization calculation to obtain the best scheduling solution; through this method, while meeting the power quality requirements of the industrial park, grid-connected optimization scheduling can be performed for different microgrid energy supply conditions, thereby improving the accuracy of microgrid scheduling.

[0007] To achieve the above object, the present invention provides the following technical solutions:

[0008] A microgrid grid-connected operation optimization scheduling method based on an intelligent optimization algorithm comprises:

[0009] Obtaining historical power demand of the industrial park, and generating target power demand information based on the historical power demand;

[0010] Establishing a power quality demand analysis model to analyze the power quality characteristic data and abnormal power quality characteristic data of the target power demand information;

[0011] Furthermore, the power quality demand analysis model includes a demand data processing unit, a demand data feature analysis unit, a demand data anomaly marking unit, a demand data similarity calculation unit and a demand anomaly data merging and outputting unit;

[0012] The target power demand information includes target voltage demand information and target power demand information;

[0013] The target power demand information is input into the demand data processing unit for preprocessing to generate preprocessed demand data;

[0014] The demand data characteristic analysis unit analyzes the characteristics of the pre-processed demand data to generate power quality characteristic data;

[0015] The demand data feature analysis unit includes a Transformer encoding layer, a residual convolution group, a long short-term memory neural network and a fully connected layer; the Transformer encoding layer is respectively connected to the residual convolution group and the long short-term memory neural network; the results of the residual convolution group and the long short-term memory neural network are input into the fully connected layer;

[0016] The residual convolution group includes a first residual convolution group and a second residual convolution group, the first residual convolution group includes 3 convolution layers and 1 relu activation layer; the second residual convolution group includes 3 convolution layers and 1 relu activation layer; the number of the long short-term memory neural networks is 6;

[0017] Furthermore, the first residual convolution group is connected to the second residual convolution group, and the output of the first residual convolution group is input to the first residual convolution group; the first residual convolution group RC1 and the second residual convolution group RC2 are constructed as follows:

[0018]

[0019] in, The convolution layer has 64 channels and a convolution kernel size of 1*3. is the convolution layer with 64 channels and a convolution kernel size of 1*1, and relu is the relu activation layer. Indicates series connection, Indicates parallel connection;

[0020] The demand data abnormality marking unit performs feature comparison based on the features of the power quality feature data, marks and generates abnormal power quality feature data;

[0021] The demand data similarity calculation unit performs similarity calculation on the abnormal power quality feature data and the non-abnormal power quality feature data to generate abnormal demand similarity data and non-abnormal demand similarity data;

[0022] κ=α1·f c +α2·f σ +α3·f v ;

[0023] Among them, κ is the similarity calculation result, f c is the cluster analysis result, f σ is the standard deviation analysis result, f v is the voting analysis result, α1, α2 and α3 are similarity weights, α1+α2+α3=1;

[0024] Acquire characteristic intervals based on the power quality characteristic data and generate a plurality of first power demand data; the first power demand data is the target power demand information acquired based on the characteristic intervals of the non-abnormal power quality characteristics;

[0025] Obtaining an abnormal characteristic interval based on the abnormal power quality characteristic data and generating a plurality of second power demand data; the second power demand data is the target power demand information obtained based on the characteristic interval of the demand abnormality similarity data;

[0026] Wherein, the second power demand data includes independent second power demand data and interval second power demand data;

[0027] Performing data interval calculation on the demand anomaly similarity data that is not less than a similarity threshold to obtain the target power demand information of the characteristic interval, wherein the second power demand data of the interval is the target power demand information;

[0028] The independent second power demand data is the demand abnormality similarity data less than the similarity threshold, and the target power demand information of the characteristic interval is obtained;

[0029] The abnormal demand data merging and outputting unit performs data interval calculation on the target power demand information according to the abnormal demand similarity data and the non-abnormal demand similarity data according to a similarity threshold, and outputs the characteristic interval;

[0030] Acquire first power supply information based on the main power grid system of the industrial park, and acquire corresponding second power supply information and microgrid power dispatch parameters based on each microgrid system of the industrial park;

[0031] Establishing a particle swarm algorithm scheduling model, wherein the particle swarm algorithm scheduling model schedules a plurality of microgrid power scheduling parameters according to a plurality of the first power demand data, the first power supply information, and a plurality of the second power supply information, and generates a first microgrid scheduling strategy to schedule each of the microgrid systems;

[0032] Furthermore, the corresponding second power supply information and a plurality of microgrid power dispatching parameters are obtained according to each microgrid system; the power dispatching parameters and the microgrid power dispatching parameters include voltage and power;

[0033] Obtaining a priority based on the second power supply information of each microgrid system; setting a corresponding priority weight for a plurality of the microgrid power dispatching parameters of each microgrid system according to the priority; initializing the speed and position of the particle swarm using the microgrid power dispatching parameters, using the priority weight and cost as a penalty coefficient for iterative updating of the particle swarm, using the difference between the first power demand data and the first power supply information as a target condition for the particle swarm, and iterating the position and speed of the particle swarm multiple times until the particle swarm reaches the first power demand data;

[0034] generating the first microgrid dispatching strategy by taking the microgrid power dispatching parameter of the last iteration as the target parameter;

[0035] The particle swarm optimization scheduling model schedules the plurality of microgrid power scheduling parameters through the plurality of second power demand data, the first power supply information and the plurality of second power supply information, and generates a second microgrid scheduling strategy to schedule each of the microgrid systems;

[0036] Further, the corresponding second power supply information and a plurality of microgrid power dispatch parameters are obtained according to each microgrid system;

[0037] Obtaining a priority based on the second power supply information of each of the microgrid systems; setting a corresponding priority weight for a plurality of the microgrid power dispatching parameters of each of the microgrid systems according to the priority; initializing the speed and position of the particle swarm using the microgrid power dispatching parameters, and using the priority weight and cost as a penalty coefficient for iterative updating of the particle swarm;

[0038] using a difference between the second power demand data of the interval of the second power demand data and the first power supply information as a target condition for the particle swarm, and iterating the position and the speed of the particle swarm multiple times until the particle swarm reaches the second power demand data of the interval;

[0039] using a difference between independent second power demand data of the second power demand data and the first power supply information as a target condition for the particle swarm, and iterating the position and the speed of the particle swarm multiple times until the particle swarm reaches the independent second power demand data;

[0040] The second microgrid scheduling strategy is generated by taking the microgrid power scheduling parameters of the two particle swarms of the last iteration as target parameters.

[0041] Compared with the prior art, the present invention has the following beneficial effects:

[0042] 1. The present invention obtains target power demand information through the historical power demand of the industrial park, analyzes the abnormal power demand data and abnormal power demand data of the target power demand information based on the power quality demand analysis model, and calculates the characteristic intervals of each target power demand information according to the similarity between abnormal data and non-abnormal data, so as to obtain power quality requirement data of different demands at different times; compared with other methods, this method can ensure that the microgrid of the industrial park meets different power quality requirements at different times during grid-connected optimization scheduling, so as to ensure the power quality requirements of the industrial park and improve the accuracy of microgrid optimization scheduling.

[0043] 2. In order to improve the accuracy of similarity calculation between abnormal data and non-abnormal data, the present invention obtains the marked abnormal power quality characteristic data and non-abnormal power quality characteristic data, and then performs weighted calculation through clustering algorithm, standard deviation method and voting method, and calculates the similarity between the abnormal power quality characteristic data and the non-abnormal power quality characteristic data through multiple dimensions to ensure the accuracy of the similarity calculation results. The high-accuracy similarity can accurately divide different feature intervals and obtain accurate power demand data in different intervals, providing an accurate data basis for subsequent microgrid optimization scheduling to improve the accuracy of microgrid optimization scheduling.

[0044] 3. The present invention uses the difference between the power demand information and the main power grid power supply data in different characteristic intervals as the adjustment target data, and the current supply weights of different microgrid systems as the penalty coefficient, and combines the particle swarm optimization algorithm to optimize the power dispatching parameters of the microgrid, and uses the final result after multiple iterations of the target as the power dispatching strategy; compared with other methods, the present invention can dispatch the grid-connected parameters of the microgrid system on the basis of ensuring the power quality requirements of the industrial park, and can ensure the quality requirements of power in addition to the cost while optimizing, and can also perform reasonable dispatch according to the power supply quality of different microgrid systems to improve the accuracy of microgrid optimization dispatching. BRIEF DESCRIPTION OF THE DRAWINGS

[0045] Figure 1 Schematic diagram of the method flow of the present invention;

[0046] Figure 2 A flow chart of the power quality demand analysis model of the present invention;

[0047] Figure 3 A schematic flow chart of a demand data feature analysis unit of the power quality demand analysis model of the present invention;

[0048] Figure 4 Schematic diagram of the flow of the residual convolution group of the power quality demand analysis model of the present invention;

[0049] Figure 5 This is the microgrid power result diagram after the power quality demand analysis model of the present invention and particle swarm algorithm optimization. DETAILED DESCRIPTION

[0050] The following will provide a clear and complete description of the technical solutions in the embodiments of the present invention in conjunction with the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. All other embodiments obtained by ordinary technicians in this field based on the embodiments of the present invention without making any creative efforts shall fall within the scope of protection of the present invention.

[0051] Reference Figure 1 As shown, the microgrid system has good advantages in economy, energy saving, environmental protection and emergency response, and is therefore widely used in many fields. For the microgrid system of the industrial park, since the power supply system is relatively complex, in order to reduce the pressure of the main power grid, the microgrid system needs to be dispatched. However, since the power quality requirements of the industrial park in different time intervals are also different, the current optimization scheduling method is to optimize the scheduling by using unified power parameters and cost parameters as targets, resulting in poor accuracy of microgrid optimization scheduling. Therefore, the present invention provides a microgrid grid-connected operation optimization scheduling method based on an intelligent optimization algorithm, and the technical solution is as follows:

[0052] Obtaining historical power demand of the industrial park, and generating target power demand information based on the historical power demand;

[0053] Establishing a power quality demand analysis model to analyze the power quality characteristic data and abnormal power quality characteristic data of the target power demand information; obtaining characteristic intervals based on the power quality characteristic data and generating a plurality of first power demand data; obtaining abnormal characteristic intervals based on the abnormal power quality characteristic data and generating a plurality of second power demand data;

[0054] Acquire first power supply information based on the main power grid system of the industrial park, and acquire corresponding second power supply information and microgrid power dispatch parameters based on each microgrid system of the industrial park;

[0055] Establishing a particle swarm algorithm scheduling model, wherein the particle swarm algorithm scheduling model schedules a plurality of microgrid power scheduling parameters according to a plurality of the first power demand data, the first power supply information, and a plurality of the second power supply information, and generates a first microgrid scheduling strategy to schedule each of the microgrid systems;

[0056] The particle swarm optimization scheduling model schedules multiple microgrid power scheduling parameters through multiple second power demand data, the first power supply information and multiple second power supply information, and generates a second microgrid scheduling strategy to schedule each microgrid system.

[0057] To obtain more accurate power quality requirements for different time intervals, the target power demand information is generated from the industrial park's historical power demand as a reference for analysis. A power quality demand analysis model is established, and situations where power quality requirements are significantly increased are treated as abnormal data. This data is then marked through abnormal data analysis to achieve the purpose of marking data with high power quality requirements. Similarly, similarity calculation is used to merge data with the same quality requirements to obtain the corresponding target power demand information and the corresponding time interval.

[0058] This method can reliably understand the power quality requirements of each industrial park at different time intervals. This data can provide an accurate data basis for the subsequent optimization and scheduling of the microgrid system, thereby improving the accuracy of subsequent microgrid optimization and scheduling.

[0059] Afterwards, the power supply parameter information of the main grid system and each microgrid system is obtained to obtain the power quality of different microgrid systems and the power quality of the main grid system. Combined with the scheduling parameters of each microgrid system, the difference between the power quality requirements of different time intervals and the power supply of the main grid system is used as the target parameter, and the cost and the power quality of the microgrid power supply are used as the penalty coefficient. The relevant parameters of the microgrid system are optimized based on the particle swarm algorithm to generate the corresponding microgrid grid-connected optimization scheduling strategy;

[0060] This method can optimize the grid connection and scheduling of microgrids while ensuring the power quality of the industrial park in different time periods. While ensuring the power cost and power quality requirements of the industrial park, it can also achieve reasonable power output based on the microgrid's own power supply capacity, thereby improving the accuracy of microgrid optimization scheduling.

[0061] Example 1

[0062] Obtaining historical power demand of the industrial park, and generating target power demand information based on the historical power demand;

[0063] Establishing a power quality demand analysis model to analyze the power quality characteristic data and abnormal power quality characteristic data of the target power demand information;

[0064] Furthermore, the power quality demand analysis model includes a demand data processing unit, a demand data feature analysis unit, a demand data anomaly marking unit, a demand data similarity calculation unit and a demand anomaly data merging and output unit. Figure 2 As shown in the figure, since the demand for power quality in the industrial park varies in different time periods and seasons, failure to provide power supply of corresponding quality will have a huge impact on the work of the industrial park. Therefore, the historical power demand data of the industrial park is used as a reference, and the increase or decrease in voltage demand is analyzed as abnormal data. Finally, the similarity of these abnormal data is classified and calculated and mapped to the corresponding time. The time intervals corresponding to different power demands can be analyzed to facilitate the subsequent grid-connected optimization scheduling of the microgrid and the main grid. Adjustments are made based on actual power demand to improve the accuracy of the grid-connected scheduling of the microgrid system.

[0065] The target power demand information includes target voltage demand information and target power demand information; the target voltage demand information and target power demand information are time series data recorded by time;

[0066] The target power demand information is input into the demand data processing unit for preprocessing to generate preprocessed demand data;

[0067] The preprocessing of the demand data processing unit mainly includes denoising, standardization, missing value filling and value screening, and data segmentation and splicing. For example, a year's data is divided into weeks, that is, 7 days of data as units, forming data with a dimension of 7*(365 / 7). Then, it is rounded down to form data with a dimension of 7*51. In order to ensure the integrity of the preprocessing demand data, the remaining data that cannot be divided evenly is discarded. In the specific implementation process, corresponding processing can be carried out according to the specific situation.

[0068] The demand data characteristic analysis unit analyzes the characteristics of the pre-processed demand data to generate power quality characteristic data;

[0069] The demand data feature analysis unit includes a Transformer encoding layer, a residual convolution group, a long short-term memory neural network and a fully connected layer; the Transformer encoding layer is respectively connected to the residual convolution group and the long short-term memory neural network; the results of the residual convolution group and the long short-term memory neural network are input to the fully connected layer; Figure 3 As shown;

[0070] The residual convolution group includes a first residual convolution group and a second residual convolution group, the first residual convolution group includes 3 convolution layers and 1 relu activation layer; the second residual convolution group includes 3 convolution layers and 1 relu activation layer; the number of the long short-term memory neural networks is 6;

[0071] In order to accurately identify abnormal data such as high and low power demand in power demand data, the attention mechanism of the Transformer encoding layer can analyze the long-range dependency characteristics between each data in the time series data formed by power data. The residual convolution group captures the spatial morphological change characteristics, and the long-term and short-term neural network captures the temporal continuity change characteristics. Through the combined analysis of multiple models, the change characteristics of power data can be accurately analyzed to accurately identify abnormal data in power demand, provide an accurate abnormal data time interval data basis, facilitate subsequent microgrid grid connection adjustment and optimization, and improve the accuracy of optimization;

[0072] Furthermore, the first residual convolution group is connected to the second residual convolution group, and the output of the first residual convolution group is input to the first residual convolution group; the first residual convolution group RC1 and the second residual convolution group RC2 are constructed as follows:

[0073]

[0074] in, The convolution layer has 64 channels and a convolution kernel size of 1*3. is the convolution layer with 64 channels and 1*1 convolution kernel size, and relu is the relu activation layer. Indicates series connection, Indicates parallel connection, refer to Figure 4 As shown in the figure, by adopting a parallel analysis method, the accuracy of spatial features can be further improved in the process of extracting spatial features. By adjusting the convolution kernel size in different dimensions, the target power demand information can be analyzed from multiple aspects, and the accuracy of spatial information can be improved to ensure the time intervals of different power quality requirements and improve the accuracy of subsequent microgrid grid connection optimization and scheduling adjustments.

[0075] The demand data abnormality marking unit performs feature comparison based on the features of the power quality feature data, marks and generates abnormal power quality feature data;

[0076] The demand data similarity calculation unit performs similarity calculation on the abnormal power quality feature data and the non-abnormal power quality feature data to generate abnormal demand similarity data and non-abnormal demand similarity data;

[0077] κ=α1·f c +α2·f σ +α3·f v ;

[0078] Among them, κ is the similarity calculation result, f c is the cluster analysis result, f σ is the standard deviation analysis result, f v is the voting analysis result, α1, α2 and α3 are similarity weights, α1+α2+α3=1; α1 defaults to 0.4, α2 defaults to 0.4, and α3 defaults to 0.2. The specific weight values ​​can be changed according to actual conditions;

[0079] The clustering model uses the k-means model, and the value of k is set based on historical data and the experience of relevant personnel;

[0080] The standard deviation and voting method traverses all feature data, subtracts a certain data from all other data to obtain the change fluctuation data, and after traversing all data, calculates the similarity based on the change fluctuation data of each data using the standard deviation and voting method;

[0081] The standard deviation is calculated by taking the maximum fluctuation value between a large number of non-abnormal historical data and the maximum fluctuation value between abnormal historical data. The result is obtained based on the ratio of the maximum fluctuation values. The specific calculation is as follows:

[0082]

[0083] Among them, fl is the fluctuation data after subtracting the current data from other data, fl max is the maximum fluctuation value;

[0084] The voting rule is to set multiple similarity standard values ​​based on expert experience, and determine whether similarity is based on the maximum ratio of the similarity standard values; the specific calculation is as follows:

[0085]

[0086] Among them, max is the maximum value, flt n is the nth similarity standard value;

[0087] In order to ensure the accuracy of similarity calculation of marked abnormal data, this embodiment clusters all data through cluster analysis. The clustering results can ensure the accuracy of similarity calculation. At the same time, in order to further improve the accuracy of similarity results, standard deviation analysis and voting method analysis are also added. In combination with statistical methods, the similarity of data is calculated from multiple aspects to ensure the accuracy of data similarity calculation and the accuracy of subsequent time interval division, so as to improve the accuracy of subsequent microgrid grid connection optimization scheduling adjustment.

[0088] Acquire characteristic intervals based on the power quality characteristic data and generate a plurality of first power demand data; the first power demand data is the target power demand information acquired based on the characteristic intervals of the non-abnormal power quality characteristics;

[0089] To improve the accuracy of subsequent microgrid optimization and scheduling, the high and low power demand conditions in industrial parks are prioritized as power quality requirements under abnormal circumstances. Therefore, multiple time intervals are calculated based on historical data and the data is stored. After the time intervals of high and low power data are confirmed, the time intervals under non-abnormal circumstances are confirmed, and the average power quality data of each interval is calculated. The power demand value under the non-abnormal interval is generated as the target power demand for the optimized scheduling of the microgrid and the main grid.

[0090] The corresponding power quality data under non-abnormal conditions is obtained through the divided time interval as the target power demand information, so as to facilitate the subsequent optimization and scheduling of the microgrid grid connection and improve the accuracy of the microgrid system grid connection optimization scheduling;

[0091] Obtaining an abnormal characteristic interval based on the abnormal power quality characteristic data and generating a plurality of second power demand data; the second power demand data is the target power demand information obtained based on the characteristic interval of the demand abnormality similarity data;

[0092] After obtaining abnormal high power demand data and low power demand data, they are mapped to the time interval of the target power demand information and the corresponding time feature interval is obtained. In order to ensure that accurate power quality data can be supplied in the time feature interval, after calculating the similarity, the minimum time span of the abnormal time feature interval in all historical data is determined, and the maximum time span in the time feature interval containing the minimum time span is determined. The average of the power demand data of the maximum time span is used as the target power demand for the optimized scheduling of the microgrid and the main grid.

[0093] The divided abnormal data can accurately determine the time intervals of power quality requirements for industrial parks under high and low voltage conditions. Using the power quality requirement data within this time interval as the target data, the microgrid is dispatched and optimized, providing the industrial park with power quality that meets the requirements in different time periods, thereby improving the accuracy of the microgrid system's grid-connected optimization scheduling.

[0094] Wherein, the second power demand data includes independent second power demand data and interval second power demand data;

[0095] Performing data interval calculation on the demand anomaly similarity data that is not less than a similarity threshold to obtain the target power demand information of the characteristic interval, wherein the second power demand data of the interval is the target power demand information;

[0096] The independent second power demand data is the demand abnormality similarity data less than the similarity threshold, and the target power demand information of the characteristic interval is obtained;

[0097] The similarity threshold is set to 0.95 by default and can be modified according to actual conditions;

[0098] Since data fluctuations or additional power demand increases or decreases may occur in the time characteristic interval of abnormal data calculation, after the similarity calculation, these data with low similarity are classified and stored to generate independent time intervals. Based on the previous interval division method, the frequency of occurrence of these time intervals, if the frequency is high, it is saved as a special time interval, if the frequency is low, these data are stored as special cases and not recorded as adjusted time intervals; this method can grasp the high-quality power demand and low-quality power demand of the industrial park in different time periods, and obtain some characteristic abnormal time characteristic intervals, further improving the accuracy of the acquisition of different power quality requirement data of the industrial park at different times, facilitating the subsequent optimization and adjustment of the microgrid grid-connected data, and improving the accuracy of the microgrid system grid-connected optimization scheduling;

[0099] Frequency judgment is mainly carried out by comparing data from the same time interval of each year. For example, the data from July 2017 is compared with data from July of different years, such as July 2016 and July 2015. The upper and lower thresholds for frequency judgment are 3. If it is not less than 3 times, it is judged as a high frequency and recorded as a special time feature interval. If it is less than 3 times, it is judged as a characteristic situation and is not recorded as a special time feature interval. The data will be stored. The specific frequency number can be adjusted based on expert experience and actual conditions.

[0100] Furthermore, if the time interval in which the frequency exceeds the threshold is smaller than the actual interval, for example, the time interval of the abnormal comparison data is 10 days, and after comparison it is found that only 5 days of data exceed the threshold frequency, in this case, the data of these 5 days will be regarded as the abnormal time interval, and the other 5 days will be regarded as the quasi-abnormal time interval and stored as alternative data; in the process of grid-connected tuning of microgrid parameters, the 5-day data will be used as the main tuning target data to generate the main strategy plan, and the 10-day data will be used as the alternative tuning target data to generate the alternative strategy plan. By generating multiple cases, the accuracy of the target data for microgrid tuning can be guaranteed from multiple aspects, thereby improving the accuracy of the grid-connected optimization scheduling of the microgrid system;

[0101] The abnormal demand data merging and outputting unit performs data interval calculation on the target power demand information according to the abnormal demand similarity data and the non-abnormal demand similarity data according to a similarity threshold, and outputs the characteristic interval;

[0102] Acquire first power supply information based on the main power grid system of the industrial park, and acquire corresponding second power supply information and microgrid power dispatch parameters based on each microgrid system of the industrial park;

[0103] Establishing a particle swarm algorithm scheduling model, wherein the particle swarm algorithm scheduling model schedules a plurality of microgrid power scheduling parameters according to a plurality of the first power demand data, the first power supply information, and a plurality of the second power supply information, and generates a first microgrid scheduling strategy to schedule each of the microgrid systems;

[0104] Furthermore, the second power supply information corresponding to each microgrid system is obtained and a plurality of microgrid power dispatch parameters are obtained; the power dispatch parameters and the microgrid power dispatch parameters include voltage and power; the first power supply information is the average power supply of the main power grid system, and the second power supply information is the stored power amount and average supply amount of each microgrid system;

[0105] Obtaining a priority based on the second power supply information of each microgrid system; setting a corresponding priority weight for a plurality of the microgrid power dispatching parameters of each microgrid system according to the priority; initializing the speed and position of the particle swarm using the microgrid power dispatching parameters, using the priority weight and cost as a penalty coefficient for iterative updating of the particle swarm, using the difference between the first power demand data and the first power supply information as a target condition for the particle swarm, and iterating the position and speed of the particle swarm multiple times until the particle swarm reaches the first power demand data;

[0106] generating the first microgrid dispatching strategy by taking the microgrid power dispatching parameter of the last iteration as the target parameter;

[0107] By taking the difference between the power quality requirements of the industrial park and the power supply of the main grid in different time intervals under non-abnormal conditions and using these data as target data, the power quality requirements of the industrial park can be met during the microgrid grid optimization process; at the same time, in order to ensure the accuracy of the parameter adjustment optimization process, the cost data of the power data is used as a penalty coefficient, and the actual supply and actual storage capacity of each microgrid system are used as another penalty coefficient; this method can make adaptive adjustments based on the specific parameters of the microgrid system during the optimization process while ensuring that the cost is not too high. This method ensures that the microgrid system will not exceed the parameters of the microgrid system supply while ensuring that the power quality requirements of the industrial park are met, thereby improving the accuracy of the microgrid system grid optimization scheduling;

[0108] The particle swarm optimization scheduling model schedules the plurality of microgrid power scheduling parameters through the plurality of second power demand data, the first power supply information and the plurality of second power supply information, and generates a second microgrid scheduling strategy to schedule each of the microgrid systems;

[0109] Further, the corresponding second power supply information and a plurality of microgrid power dispatch parameters are obtained according to each microgrid system;

[0110] Obtaining a priority based on the second power supply information of each of the microgrid systems; setting a corresponding priority weight for a plurality of the microgrid power dispatching parameters of each of the microgrid systems according to the priority; initializing the speed and position of the particle swarm using the microgrid power dispatching parameters, and using the priority weight and cost as a penalty coefficient for iterative updating of the particle swarm;

[0111] using a difference between the second power demand data of the interval of the second power demand data and the first power supply information as a target condition for the particle swarm, and iterating the position and the speed of the particle swarm multiple times until the particle swarm reaches the second power demand data of the interval;

[0112] using a difference between independent second power demand data of the second power demand data and the first power supply information as a target condition for the particle swarm, and iterating the position and the speed of the particle swarm multiple times until the particle swarm reaches the independent second power demand data;

[0113] generating the second microgrid dispatching strategy by taking the microgrid power dispatching parameters of the two particle swarms of the last iteration as target parameters;

[0114] By using high power quality requirements or low power quality requirements in different time characteristic intervals in the industrial park as optimization target data, the microgrid optimization adjustment can meet the power quality requirements of the actual industrial park; at the same time, based on the actual power supply situation of each microgrid, multiple priority weights are divided as penalty coefficients. During the optimization and adjustment process, appropriate optimization adjustments are made to each microgrid system to prevent the situation of exceeding the microgrid system power supply data, thereby improving the accuracy of the microgrid system grid-connected optimization scheduling.

[0115] Example 2

[0116] The present invention collects 6 years of data from Industrial Park A, selects 360 days of data each year, pre-processes all the data, divides the data into 30-day units, collects data every half an hour every day, and generates data with a dimension of (48*30)*60. The data of the first 5 years are used as training data, and the data of the last year are used as test data for abnormal training and identification. The model with the highest training accuracy is selected as the test model to identify the test data of the 6th year. The data of the 6th year is divided into the same dimension size, i.e. 48*30, and 6 groups of data are selected from the 60 groups of data for abnormal data identification and marking test. 3 groups in the array contain abnormal data, including high power demand data and low power demand data, and 3 groups are non-abnormal data. Table 1 shows the recognition accuracy results:

[0117] Table 1. Accuracy of abnormal and normal identification marks for 6 test data sets in the 6th year

[0118] Data group label Whether it contains abnormal data Anomaly recognition accuracy Normal recognition accuracy Data 1 yes 94.13％ 93.28％ Data 2 no none 94.15％ Data 3 no none 93.66％ Data 4 yes 93.22％ 92.94％ Data 5 yes 93.79％ 92.86％ Data 6 no none 93.47％

[0119] From the results in Table 1, we can see that the accuracy of abnormal data recognition is above 93%, which is a good recognition result. The recognition of normal data is also above 92.5%, which shows that the model has good recognition and labeling results for different data.

[0120] To further test the recognition accuracy of the model, this example also collected one year of data from another industrial park, B. Using the same preprocessing and partitioning methods, four data sets were selected for testing. Two of the data sets contained abnormal data, and two did not. The test results are shown in Table 2.

[0121] Table 2. Accuracy of abnormal and normal identification marks for another industrial park test data

[0122] Data group label Whether it contains abnormal data Anomaly recognition accuracy Normal recognition accuracy Data 1 yes 91.13％ 92.34％ Data 2 no none 91.98％ Data 3 yes 90.68％ 91.72％ Data 4 no none 92.56％

[0123] From the results in Table 2, we can see that although the recognition accuracy of the data has decreased because they are not in the same industrial park, it still remains above 90%, proving that the recognition effect of this method is good;

[0124] This example optimizes the 24-hour data of Industrial Park A on a certain day. The optimization results refer to Figure 5 As shown in the figure, the diamond block line is the expected power data of the initially set microgrid, and the square block line is the dispatch power data optimized after the power quality demand analysis model and particle swarm algorithm. Figure 5 It can be seen that the data is optimized according to the high power demand and low power demand in different time periods.

[0125] While embodiments of the present invention have been shown and described, it will be appreciated by those skilled in the art that various changes, modifications, substitutions, and variations may be made to these embodiments without departing from the principles and spirit of the invention, and that the scope of the invention is defined by the appended claims and their equivalents.

Claims

1. A microgrid grid-connected operation optimization scheduling method based on intelligent optimization algorithm, characterized in that: include: Obtaining historical power demand of the industrial park, and generating target power demand information based on the historical power demand; Establishing a power quality demand analysis model to analyze the power quality characteristic data and abnormal power quality characteristic data of the target power demand information; obtaining characteristic intervals based on the power quality characteristic data and generating a plurality of first power demand data; Obtaining an abnormal characteristic interval according to the abnormal power quality characteristic data and generating a plurality of second power demand data; Acquire first power supply information based on the main power grid system of the industrial park, and acquire corresponding second power supply information and microgrid power dispatch parameters based on each microgrid system of the industrial park; Establishing a particle swarm algorithm scheduling model, wherein the particle swarm algorithm scheduling model schedules a plurality of microgrid power scheduling parameters according to a plurality of the first power demand data, the first power supply information, and a plurality of the second power supply information, and generates a first microgrid scheduling strategy to schedule each of the microgrid systems, including: Acquire the corresponding second power supply information and a plurality of microgrid power dispatching parameters according to each microgrid system; the power dispatching parameters and the microgrid power dispatching parameters include voltage and power; Obtaining a priority based on the second power supply information of each microgrid system; setting a corresponding priority weight for a plurality of the microgrid power dispatching parameters of each microgrid system according to the priority; initializing the speed and position of the particle swarm using the microgrid power dispatching parameters, using the priority weight and cost as a penalty coefficient for iterative updating of the particle swarm, using the difference between the first power demand data and the first power supply information as a target condition for the particle swarm, and iterating the position and speed of the particle swarm multiple times until the particle swarm reaches the first power demand data; generating the first microgrid dispatching strategy by taking the microgrid power dispatching parameter of the last iteration as the target parameter; The particle swarm algorithm scheduling model schedules the plurality of microgrid power scheduling parameters by using the plurality of second power demand data, the first power supply information, and the plurality of second power supply information, and generates a second microgrid scheduling strategy to schedule each of the microgrid systems, including: Acquiring the corresponding second power supply information of each microgrid system to obtain a plurality of the microgrid power dispatching parameters; acquiring a priority according to the second power supply information of each microgrid system; setting a corresponding priority weight for the plurality of the microgrid power dispatching parameters of each microgrid system according to the priority; initializing the speed and position of the particle swarm with the microgrid power dispatching parameters, and using the priority weight and cost as a penalty coefficient for iterative updating of the particle swarm; Using the difference between the interval second power demand data of the second power demand data and the first power supply information as the target condition of the particle swarm, iterating the position and the speed of the particle swarm multiple times until the particle swarm reaches the interval second power demand data; using the difference between the independent second power demand data of the second power demand data and the first power supply information as the target condition of the particle swarm, iterating the position and the speed of the particle swarm multiple times until the particle swarm reaches the independent second power demand data; The second microgrid scheduling strategy is generated by taking the microgrid power scheduling parameters of the two particle swarms of the last iteration as target parameters.

2. A microgrid grid-connected operation optimization scheduling method based on an intelligent optimization algorithm according to claim 1, characterized in that: Establishing a power quality demand analysis model to analyze the power quality characteristic data and abnormal power quality characteristic data of the target power demand information includes: The power quality demand analysis model includes a demand data processing unit, a demand data feature analysis unit, a demand data anomaly marking unit, a demand data similarity calculation unit and a demand anomaly data merging and output unit; The target power demand information includes target voltage demand information and target power demand information; The target power demand information is input into the demand data processing unit for preprocessing to generate preprocessed demand data; The demand data characteristic analysis unit analyzes the characteristics of the pre-processed demand data to generate power quality characteristic data; The demand data abnormality marking unit performs feature comparison based on the features of the power quality feature data, marks and generates abnormal power quality feature data; The demand data similarity calculation unit performs similarity calculation on the abnormal power quality feature data and the non-abnormal power quality feature data to generate abnormal demand similarity data and non-abnormal demand similarity data; The abnormal demand data merging and outputting unit calculates a data interval for the target power demand information according to the abnormal demand similarity data and the non-abnormal demand similarity data and a similarity threshold, and outputs the characteristic interval.

3. The microgrid grid-connected operation optimization scheduling method based on intelligent optimization algorithm according to claim 2 is characterized in that: The demand data feature analysis unit analyzes the pre-processed demand data features, including: The demand data feature analysis unit includes a Transformer encoding layer, a residual convolution group, a long short-term memory neural network and a fully connected layer; the Transformer encoding layer is respectively connected to the residual convolution group and the long short-term memory neural network; the results of the residual convolution group and the long short-term memory neural network are input into the fully connected layer; The residual convolution group includes a first residual convolution group and a second residual convolution group, the first residual convolution group includes 3 convolution layers and 1 relu activation layer; the second residual convolution group includes 3 convolution layers and 1 relu activation layer; the number of the long short-term memory neural networks is 6.

4. A microgrid grid-connected operation optimization scheduling method based on an intelligent optimization algorithm according to claim 3, characterized in that: The first residual convolution group and the second residual convolution group include: The first residual convolution group is connected to the second residual convolution group, and the output of the first residual convolution group is input to the first residual convolution group; the first residual convolution group And the second residual convolution group The composition is as follows: ; ;in, The convolution layer has 64 channels and a convolution kernel size of 1*3. is the convolution layer with 64 channels and a convolution kernel size of 1*1, and relu is the relu activation layer. Indicates series connection, Indicates parallel connection.

5. The microgrid grid-connected operation optimization scheduling method based on intelligent optimization algorithm according to claim 2 is characterized in that: The demand data similarity calculation unit performs similarity calculation on the abnormal power quality characteristic data and the non-abnormal power quality characteristic data, including: ;in, is the similarity calculation result, is the cluster analysis result, is the standard deviation analysis result, For voting method analysis results, 、 and is the similarity weight, .

6. The microgrid grid-connected operation optimization scheduling method based on intelligent optimization algorithm according to claim 1 is characterized in that: A characteristic interval is acquired according to the power quality characteristic data and a plurality of first power demand data are generated. The first power demand data is the target power demand information acquired according to the characteristic interval of the non-abnormal power quality characteristic.

7. The microgrid grid-connected operation optimization scheduling method based on intelligent optimization algorithm according to claim 1 is characterized in that: An abnormal characteristic interval is obtained through the abnormal power quality characteristic data and a plurality of second power demand data are generated. The second power demand data is the target power demand information obtained according to the characteristic interval of the demand abnormality similarity data.

8. The microgrid grid-connected operation optimization scheduling method based on intelligent optimization algorithm according to claim 7 is characterized in that: The second power demand data includes independent second power demand data and interval second power demand data: Performing data interval calculation on the demand anomaly similarity data that is not less than a similarity threshold to obtain the target power demand information of the characteristic interval, wherein the second power demand data of the interval is the target power demand information; The independent second power demand data is the demand abnormality similarity data less than the similarity threshold, and data interval calculation is performed to obtain the target power demand information of the characteristic interval.

Images