Distributed energy optimization decision system based on cloud computing

Abstract

The application belongs to the technical field of energy optimization, and discloses a distributed energy optimization decision system based on cloud computing; comprising: a data receiving module, which is used for receiving historical power generation data collected by a resource end and historical load data collected by a user end by a cloud platform; a region division module, which is used for dividing a distributed power grid into regions according to the historical power generation data and the historical load data by the cloud platform; a data prediction module, which is used for predicting future power generation data and future load data according to the historical power generation data and the historical load data by the cloud platform; and an energy optimization module, which is used for formulating an energy dispatching optimization strategy by using an optimized fish optimization algorithm according to the future power generation data and the future load data by the cloud platform; the application fully gives play to the advantages of cloud computing, realizes intelligent management and automatic operation of the distributed power grid, reduces energy consumption and cost in the energy dispatching process, and improves the stable operation and sustainable development capacity of the distributed power grid.

Description

Technical Field

[0001] This invention relates to the field of energy optimization technology, and more specifically, to a cloud-based distributed energy optimization decision system. Background Technology

[0002] With the rapid development of renewable energy and the increasing number of distributed energy systems on the user side, traditional centralized power grids are no longer able to meet the complex needs of the current power system. Distributed power grids, with their advantages of flexibility and local consumption of renewable energy, are becoming an important part of the power system. However, the complexity of distributed power grid systems is also becoming increasingly prominent. A large number of distributed generation units, energy storage devices, load response components and other components are interconnected, generating massive amounts of operational data. How to effectively integrate and utilize this data to achieve optimized scheduling and energy management of distributed power grids has become the key to the intelligent development of the power system.

[0003] Chinese patent application CN117394418A discloses a multi-objective optimization method for dispatching distributed renewable energy distribution networks. The method includes: data acquisition: acquiring power data of the distributed renewable energy distribution network; data storage: a data concentrator uploads the power data acquired during the collection phase to the application layer, i.e., data on-chain operation; objective optimization: after referencing and analyzing the actual situation of the distribution network and the distribution network zoning principles, the distributed renewable energy distribution network is divided into regions; constraints are defined; and a solution method is developed. This invention can optimize the photovoltaic power absorption rate and voltage deviation of distributed renewable energy distribution networks, significantly improving the absorption rate of distributed renewable energy and reducing voltage deviation.

[0004] While the aforementioned technologies can optimize the dispatch of distributed renewable energy sources, the regional division is mainly based on preset zoning principles, which cannot adapt to the dynamic changes in the actual situation of distributed power grids. Furthermore, while introducing blockchain technology improves data security and transparency, it also increases the complexity and operating costs of the system. In addition, the dispatch optimization mainly considers the output absorption rate of distributed renewable energy sources and the voltage deviation of the distribution network, without taking into account the costs and energy losses during the dispatch process. This leads to a series of hidden dangers such as poor economic benefits, reduced energy efficiency, and reduced system stability during energy dispatch.

[0005] In view of this, the present invention proposes a distributed energy optimization decision system based on cloud computing to solve the above problems. Summary of the Invention

[0006] To overcome the aforementioned shortcomings of the prior art and achieve the above objectives, the present invention provides the following technical solution: a cloud computing-based distributed energy optimization decision-making system, applied in a cloud platform, comprising:

[0007] The data receiving module is used to receive historical power generation data collected by the resource end and historical load data collected by the user end. The historical load data is the load data of m consecutive historical time points corresponding to each user end. The load data is the power demand of the user end on the distributed grid within a preset time interval. The load data is obtained by: obtaining the power consumption, which is the power consumption of the user end within the preset time interval; obtaining the user power consumption, which is the power generation of the user end within the preset time interval; and subtracting the user power consumption from the power consumption to obtain the load data.

[0008] The region division module is used to divide the distributed power grid into regions based on historical power generation data and historical load data. The steps for dividing the distributed power grid into regions include:

[0009] Step a: Obtain the topology of the distributed power grid and generate the power grid coordinate system;

[0010] Step b: Pre-divide the distributed power grid into A pre-divided areas;

[0011] Step c: Obtain the region center coordinates of each pre-divided region and the resource coordinates of each resource endpoint;

[0012] Step d: Assign all resource terminals to A pre-allocated regions;

[0013] Step e: Divide the distributed power grid into regions, defining... Distribution areas, ;

[0014] In step b, the distributed power grid is pre-divided into regions according to the administrative districts under the jurisdiction of different cities. The pre-divided regions correspond one-to-one with the administrative districts, and A is an integer greater than 1.

[0015] The data prediction module is used to predict future power generation and load data based on historical power generation and load data.

[0016] The energy optimization module is used to formulate energy dispatch optimization strategies based on future power generation data and future load data using an optimized biomimetic optimization algorithm.

[0017] Furthermore, the historical power generation data refers to the power generation data of each resource terminal corresponding to m consecutive historical time points. The power generation data is the power generation of the resource terminal within a preset time interval, where m is an integer greater than 1. The power generation data and load data are collected according to the preset time interval, and the time difference between each two time points is the preset time interval.

[0018] Further, in step c, the coordinates of B vertices in each pre-divided region are obtained; the coordinates of all vertices in each pre-divided region are sorted using radial or angular sorting; the area of ​​each pre-divided region is calculated; the expression for the area of ​​the pre-divided region is: In the formula, The area of ​​the pre-divided region, Let x be the x-coordinate of the i-th vertex. Let be the ordinate of the i-th vertex. ,vertex Consider as vertex ; Calculate the center coordinates of each pre-divided region based on its area; the expression for the center coordinates is: , In the formula, With the x-axis as the center, The central ordinate;

[0019] In step d, the resource allocation steps include:

[0020] Step d1: Count the number of resource terminals according to the power grid coordinate system. Preset A subset of resources Each resource subset corresponds one-to-one with a pre-divided region; Each resource terminal is randomly assigned to Within each of the resource subsets, there are different resource endpoints; Each subset of resources is considered as a single resource set;

[0021] Step d2: Repeat step d1 until M distinct resource sets are obtained, then the loop ends and proceeds to step d3;

[0022] Step d3: Calculate the total distance and total load difference for each of the M resource sets; select the optimal resource set from the M resource sets. One resource terminal is allocated to A pre-allocated regions;

[0023] In step e, Each resource end is integrated into its corresponding pre-allocated region, and then re-divided. Distribution areas.

[0024] Furthermore, the methods for calculating the total distance and total load difference include:

[0025] Calculate the Eulerian distance between the resource coordinates of each resource endpoint in each resource set and the coordinates of the region center corresponding to the corresponding resource subset; sum the Eulerian distances of all resource endpoints in each resource set to obtain the total distance value for each resource set; calculate the average power generation of each resource endpoint based on historical power generation data; calculate the average load of each user endpoint based on historical load data; sum the average power generation of each resource endpoint within each resource subset to obtain the total power generation value for each resource subset; sum the average load of each user endpoint within each resource subset to obtain the total load value for each resource subset; subtract the corresponding total power generation value from the total load value of each resource subset and take the absolute value to obtain the load difference for each resource subset; sum the load differences for each resource subset to obtain the total load difference for each resource set.

[0026] Methods for selecting the optimal resource set from M resource sets include:

[0027] Get the maximum total distance and total load difference from the M resource sets, and mark them as maximum distance and maximum difference, respectively; get the minimum total distance and total load difference from the M resource sets, and mark them as minimum distance and minimum difference, respectively; subtract the minimum distance from the maximum distance to get the distance difference; subtract the minimum difference from the maximum difference to get the difference value.

[0028] The standardized distance for each resource set is obtained by subtracting the minimum distance from the total distance of each resource set and then dividing by the distance difference; the standardized load difference for each resource set is obtained by subtracting the minimum difference from the total load difference and then dividing by the difference value.

[0029] The standardized distance of each resource set is added to the standardized load difference to obtain the standardized data of each resource set; the standardized data of each resource set is sorted from largest to smallest, and the resource set corresponding to the standardized data at the top of the list is taken as the optimal resource set.

[0030] Furthermore, methods for predicting future power generation data include:

[0031] Future power generation data corresponds to the power generation data at n future time points for each resource end; future meteorological data is obtained, which corresponds to the meteorological data at n future time points for each distribution area;

[0032] Based on historical power generation data, a power generation data time series is constructed for each resource end. The power generation data time series for each resource end and the corresponding predicted meteorological data are input into the trained power prediction model to predict the power generation data for each resource end after a time interval, and these are marked as predicted power. The predicted meteorological data is the meteorological data corresponding to each resource end in the future meteorological data, which is the same as the time point corresponding to the predicted power. Then, each predicted power replaces the earliest power generation data in the corresponding power generation data time series to obtain a new power generation data time series for each resource end. The new power generation data time series for each resource end and the corresponding predicted meteorological data are then input into the trained power prediction model to predict the predicted power again. A total of n predicted power are predicted.

[0033] Furthermore, methods for predicting future load data include:

[0034] Future load data corresponds to the load data at n future time points for each user terminal; the load data of each user terminal in the historical load data is input into the trained load prediction model to predict the load data at n future time points for each user terminal; the training process of the load prediction model includes:

[0035] Collecting continuous Load data from historical time points are used to construct a load training set. Based on the load training set, a load prediction model is trained to predict load data for future time points. ;

[0036] The system presets a sliding step size L and a sliding window length; it transforms the load data in the load training set into multiple training samples using the sliding window method, uses these training samples as input to a recurrent neural network model, and outputs the predicted load data after the sliding step size L. The subsequent load data for each training sample is used as the prediction target, with prediction accuracy as the training objective. The model accuracy is evaluated using the mean absolute percentage error (MAPE). When the calculated MAPE is less than the preset MAPE, the load prediction model training is complete. A load prediction model is then generated to predict future load data based on the existing load data. This load prediction model is an RNN neural network model.

[0037] Furthermore, the optimized biomimetic optimization algorithm is the optimized remora optimization algorithm, and the steps for formulating an energy scheduling optimization strategy include:

[0038] Step 1: Identify the load gap area and energy dispatch area;

[0039] Step 2: Preset the population size Z and the frequency threshold T;

[0040] Step 3: Initialize the population. The position of the remora in the population is defined in a one-dimensional search space, and the position of the remora corresponds one-to-one with the set label.

[0041] Step 4: Determine the fitness function;

[0042] Step 5: Select a host for each remora and update the location of each remora;

[0043] Step 6: Calculate the fitness of each remora after its location is updated, and determine whether to switch hosts;

[0044] Step 7: Update the location of the remora after switching hosts;

[0045] Step 8: Each remora enters the foraging phase, and the location of each remora is updated;

[0046] Step 9: Calculate the dynamic boundary, generate the reverse solution corresponding to each remora and add it to the population; also treat the reverse solutions of the remoras as remoras, filter the remoras in the population, and retain the Z remoras with the highest fitness.

[0047] Step 10: If the number of iterations t is less than the threshold T, return to step 5 and set the number of iterations... The iteration count is incremented by one and then assigned the iteration count; if the iteration count t is greater than or equal to the threshold T, proceed to step 11.

[0048] Step 11: Calculate the fitness corresponding to the position of each remora, sort all fitness from largest to smallest, obtain the position of the remora corresponding to the fitness at the top, and obtain the energy set corresponding to the set tag based on the obtained remora position.

[0049] In step 1, the future load data of the user terminals corresponding to each distribution area are added together sequentially to obtain the distributed load value of each distribution area; the future power generation data of the resource terminals corresponding to each distribution area are added together sequentially to obtain the distributed power generation value of each distribution area; the distribution area with the distributed power generation value greater than the distributed load value is marked as an energy dispatch area; the distribution area with the distributed power generation value less than the distributed load value is marked as a load gap area; the distribution area with the distributed power generation value equal to the distributed load value is not marked.

[0050] Furthermore, distributed energy storage units within the energy dispatch area are designated as dispatch units; a dispatch value is randomly assigned to each dispatch unit corresponding to each energy dispatch area, forming a dispatch set, for a total of W dispatch sets; wherein, the range of the dispatch value corresponding to each dispatch unit is... Y represents the difference between the distributed load value and the corresponding distributed generation value in the load gap region; the sum of the dispatch values ​​corresponding to each dispatch unit is Y; the sum of the dispatch values ​​of the dispatch units corresponding to each energy dispatch region is less than or equal to the difference between the corresponding distributed load value and the distributed generation value; different numerical labels are set for different dispatch sets and marked as set labels, and the range of the set label and the one-dimensional search space is... The expression for the position of each remora is: In the formula, Let be the initial position of the i-th remora. for Random numbers are generated between the population, and the corresponding iteration number t is initialized to 0.

[0051] In step 5, the hosts include sailfish and whales; a random number is obtained for each remora. The random number is rounded up. If the rounded random number is 0, the host of the remora is a whale. If the rounded random number is 1, the host of the remora is a sailfish.

[0052] In step 6, when the remora parasitizes the host, it will make tentative moves based on the remora's final position in the previous iteration and its current position; calculate the fitness corresponding to the remora's tentative move and mark it as the tentative fitness; mark the fitness corresponding to the remora's current position as the current fitness; compare the tentative fitness with the current fitness; if the tentative fitness is greater than or equal to the current fitness, the host is not switched; if the tentative fitness is less than the current fitness, the host is switched.

[0053] In step 7, the method for updating the location of the remora after switching hosts is the same as the method for updating the location of each remora in step 5.

[0054] Furthermore, in step 4, the expression for the fitness function is: In the formula, For fitness, This refers to the scheduling cost value; methods for obtaining the scheduling cost value include:

[0055] Based on the location of the remora, obtain the scheduling set corresponding to the set label; based on the power grid coordinate system, obtain the coordinates of each scheduling unit in the scheduling set and the corresponding scheduling value of each scheduling unit; calculate the Eulerian distance between the coordinates of each scheduling unit and the coordinates of the regional center of the load gap area, and mark it as the scheduling distance; sum the scheduling distances of each scheduling unit sequentially to obtain the total scheduling distance; preset the unit power generation cost, which includes the cost required for each type of distributed generation unit to produce one unit of electricity; based on the resource end corresponding to each scheduling unit, obtain the unit power generation cost corresponding to each scheduling unit; multiply the unit power generation cost of each scheduling unit by the corresponding scheduling value, and sum them sequentially to obtain the total cost value; obtain the line loss value of each transmission line in the distributed power grid; based on the topology of the distributed power grid, obtain the transmission lines between each scheduling unit and the load gap area, and mark them as scheduling lines; sum the line loss values ​​of each scheduling line sequentially to obtain the total line loss value; use the total scheduling distance, total cost value, and total line loss value as analysis data, input the analysis data into the trained loss prediction model, and predict the corresponding scheduling loss value.

[0056] Furthermore, the method for updating the position of each remora is as follows: calculate the fitness corresponding to the position of each remora, and mark the position of the remora with the highest fitness value as [position name missing]. ;

[0057] If the host of the parasite is a whale, then the updated remora's location is calculated as follows: ; In the formula, This represents the position of the i-th remora after the update. This represents the position of the i-th remora before the update. Adjust the coefficients to broaden the search range. for Random numbers between;

[0058] If the host fish is a sailfish, then the updated remora's location is calculated as follows: In the formula, This represents the location of a random remora before the update.

[0059] The technical effects and advantages of the cloud computing-based distributed energy optimization decision-making system of this invention are as follows:

[0060] Based on historical power generation and load data, the distributed power grid is accurately divided into regions to adapt to its dynamic changes. Furthermore, by predicting future power generation and load data, an improved Yuyu optimization algorithm is used for regional analysis, comprehensively considering factors such as resource location and load differences to formulate energy dispatch optimization strategies. By fully leveraging the advantages of cloud computing, intelligent management and automated operation of the distributed power grid are achieved. This effectively consolidates distributed energy resources, reduces energy consumption and costs during energy dispatch, improves the energy dispatch level of the distributed power grid, and enhances its stable operation and sustainable development capabilities. Attached Figure Description

[0061] Figure 1 This is a schematic diagram of a cloud computing-based distributed energy optimization decision-making system according to Embodiment 1 of the present invention;

[0062] Figure 2 This is a flowchart of the distributed energy optimization decision-making method based on cloud computing according to Embodiment 2 of the present invention. Detailed Implementation

[0063] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0064] Example 1

[0065] Please see Figure 1 As shown in the figure, the cloud computing-based distributed energy optimization decision-making system described in this embodiment includes a data receiving module, a region division module, a data prediction module, and an energy optimization module; each module is connected via wired and / or wireless means to realize data transmission between modules;

[0066] The data receiving module is used by the cloud platform to receive historical power generation data collected from the resource end and historical load data collected from the user end.

[0067] The cloud platform is a platform that provides cloud computing services; the resource end is the distributed generation unit in the distributed power grid, such as hydropower unit, wind power unit, solar power unit, etc., and one generation unit is one resource end; the user end is the electricity consumer of the distributed power grid, such as residential users, commercial users, industrial users, etc., and one electricity consumer is one user end.

[0068] Historical power generation data includes power generation data for each resource terminal corresponding to m consecutive historical time points. The power generation data represents the power generation of the resource terminal within a preset time interval, where m is an integer greater than 1. The power generation data is acquired by the energy meter built into the resource terminal. Historical load data includes load data for each user terminal corresponding to m consecutive historical time points. The load data represents the power demand of the user terminal on the distributed grid within a preset time interval. The power generation data and load data are collected according to the preset time interval. The time difference between any two time points is the preset time interval. The time interval is preset by those skilled in the art based on the actual power generation characteristics of the power generation unit (such as the variability of wind or solar power) and the fluctuation of power demand of electricity consumers.

[0069] The load data is obtained as follows: Electricity usage is obtained, which is the amount of electricity used by the user within a preset time interval. Electricity usage is obtained through a smart meter installed at the user's end; User electricity is obtained, which is the amount of electricity generated by the user within a preset time interval. User electricity is obtained from the energy meter inside a distributed generation device (such as a rooftop solar panel, a small wind turbine, etc.) installed at the user's end; Electricity usage is subtracted from user electricity to obtain load data.

[0070] It should be noted that in distributed power grids, users generally have distributed generation equipment. This is because distributed generation equipment, located at the user end, can meet the user's electricity demand locally, reduce transmission line losses, and improve overall energy efficiency. Furthermore, it can provide backup power during grid failures, enhancing the reliability of power supply at the user end. In addition, it can encourage user participation. With their own generation equipment, users can participate more actively in the operation and management of the power system, which helps cultivate users' energy awareness and promotes the development of the power system towards a more flexible and intelligent direction.

[0071] The region division module is used by the cloud platform to divide the distributed power grid into regions based on historical power generation data and historical load data.

[0072] The steps for dividing a distributed power grid into regions include:

[0073] Step a: Obtain the topology of the distributed power grid and generate the power grid coordinate system;

[0074] Step b: Pre-divide the distributed power grid into A pre-divided areas;

[0075] Step c: Obtain the region center coordinates of each pre-divided region and the resource coordinates of each resource endpoint;

[0076] Step d: Assign all resource terminals to A pre-allocated regions;

[0077] Step e: Divide the distributed power grid into regions, defining... Distribution areas, .

[0078] In step a above, the topology of the distributed power grid is obtained by those skilled in the art from the power grid company. The topology of the distributed power grid includes nodes (such as substations, distribution cabinets, generation units, etc.) and edges (such as transmission lines, distribution lines, etc.). The topology of the distributed power grid is imported into GIS software (such as ArcGIS, QGIS, etc.) to generate a power grid coordinate system.

[0079] In step b above, the distributed power grid is pre-divided into regions according to the administrative districts under the jurisdiction of different cities. The pre-divided regions correspond one-to-one with the administrative districts, and A is an integer greater than 1.

[0080] In step c above, the coordinates of B vertices in each pre-divided region are obtained. The vertices of the pre-divided region are the coordinate points describing the boundary of the pre-divided region. The coordinates of all vertices in each pre-divided region are sorted using radial or angular sorting. Radial or angular sorting is existing technology and will not be elaborated upon here. Based on the coordinates of B vertices in each pre-divided region, the area of ​​each pre-divided region is calculated. The expression for the area of ​​the pre-divided region is: In the formula, The area of ​​the pre-divided region, Let x be the x-coordinate of the i-th vertex. Let be the ordinate of the i-th vertex. ,vertex Consider as vertex ; Calculate the center coordinates of each pre-divided region based on its area; the expression for the center coordinates is: , In the formula, With the x-axis as the center, The center ordinate is the vertical coordinate; the center horizontal coordinate and the center vertical coordinate are the center coordinates; the resource coordinates of each resource end are obtained from the power grid coordinate system.

[0081] In step d above, the step of allocating all resource terminals to A pre-allocated regions includes:

[0082] Step d1: Count the number of resource terminals according to the power grid coordinate system. Preset A subset of resources Each resource subset corresponds one-to-one with a pre-divided region; Each resource terminal is randomly assigned to Within each of the resource subsets, there are different resource endpoints; Each subset of resources is considered as a single resource set;

[0083] Step d2: Repeat step d1 until M distinct resource sets are obtained, then the loop ends and proceeds to step d3; ,in, , ;

[0084] Step d3: Calculate the total distance and total load difference for each of the M resource sets; select the optimal resource set from the M resource sets, and then... One resource terminal is allocated to A pre-allocated regions.

[0085] The methods for calculating the total distance and total load difference include:

[0086] Calculate the Eulerian distance between the resource coordinates of each resource endpoint in each resource set and the center coordinates of the corresponding region of the resource subset. The Eulerian distance is calculated as follows: subtract the x-coordinate of the corresponding region center coordinate from the x-coordinate of the resource endpoint to obtain the x-coordinate difference; subtract the y-coordinate of the corresponding region center coordinate from the y-coordinate of the resource endpoint to obtain the y-coordinate difference; take the square root of the square of the sum of the squares of the x-coordinate differences and the square root of the square ...

[0087] To obtain the average power generation value for each resource, sum the power generation data for each resource at m consecutive historical time points in the historical power generation data, then divide by m. To obtain the average load value for each resource, sum the load data for each resource at m consecutive historical time points in the historical load data, then divide by m. To obtain the average load value for each resource, sum the power generation average values ​​for each resource at each resource subset. To obtain the total power generation value for each resource subset, sum the average load values ​​for each user at each resource subset. To obtain the total load value for each resource subset, subtract the corresponding total power generation value from the total load value and take the absolute value. To obtain the load difference for each resource subset, sum the load differences for each resource subset. To obtain the total load difference for each resource subset.

[0088] Methods for selecting the optimal resource set from M resource sets include:

[0089] Get the maximum total distance and total load difference from the M resource sets, and mark them as maximum distance and maximum difference, respectively; get the minimum total distance and total load difference from the M resource sets, and mark them as minimum distance and minimum difference, respectively; subtract the minimum distance from the maximum distance to get the distance difference; subtract the minimum difference from the maximum difference to get the difference value.

[0090] The standardized distance for each resource set is obtained by subtracting the minimum distance from the total distance of each resource set and then dividing by the distance difference; the standardized load difference for each resource set is obtained by subtracting the minimum difference from the total load difference and then dividing by the difference value.

[0091] The standardized distance of each resource set is added to the standardized load difference to obtain the standardized data of each resource set; the standardized data of each resource set is sorted from largest to smallest, and the resource set corresponding to the standardized data at the top of the list is taken as the optimal resource set.

[0092] In step e above, Each resource end is integrated into its corresponding pre-allocated region, and then re-divided. Distribution areas.

[0093] The data prediction module is used by the cloud platform to predict future power generation and load data based on historical power generation and load data.

[0094] Methods for predicting future power generation data include:

[0095] Future power generation data corresponds to the power generation data at n future time points for each resource, where n is an integer greater than 1; future meteorological data is obtained, which corresponds to the meteorological data at n future time points for each distribution area; the future meteorological data is obtained from the website of the local meteorological department corresponding to each distribution area; the meteorological data is the data that affects the power generation of the distributed power generation unit, such as rainfall, wind speed, wind direction, air pressure, sunshine duration, cloud coverage, etc.

[0096] Based on historical power generation data, a power generation data time series is constructed for each resource end. The power generation data time series for each resource end and the corresponding predicted meteorological data are input into the trained power prediction model to predict the power generation data for each resource end after a time interval, and these are marked as predicted power. The predicted meteorological data is the meteorological data for each resource end in the future meteorological data that corresponds to the same time point as the predicted power. For example, if the current prediction is for the power generation data at the second time point in the future, then the predicted meteorological data is the meteorological data at the second time point in the future meteorological data.

[0097] Next, replace the earliest power generation data in the corresponding power generation data time series with each predicted power generation data to obtain a new power generation data time series for each resource end. Then, input the new power generation data time series for each resource end and the corresponding predicted meteorological data into the trained power generation prediction model to predict the predicted power generation again; a total of n predicted power generation data are predicted.

[0098] The training process for the power prediction model includes:

[0099] Define the architecture of the power prediction model, including the input state, hidden state, number of output units, and hyperparameters such as learning rate and number of iterations. Using a deep learning framework (e.g., TensorFlow, PyTorch), create the network layers for the power prediction model. Each network layer consists of an input layer, multiple LSTM layers, and an output layer. Each LSTM layer comprises an input gate, a forget gate, an output gate, and a cell state. Connect the created network layers to form the complete power prediction model, where the number of units in the input layer is... The output layer has 1 unit.

[0100] Pre-collection A set of continuous power generation data and A set of continuous meteorological data, , The power generation data time series is constructed by taking the first m sets of power generation data in chronological order. This time series, along with a set of meteorological data, is used as input to a power generation prediction model. The model outputs power generation data after the specified time intervals, using actual power generation data as the prediction target. The model is then trained, and the mean absolute percentage error (MAPE) is used to evaluate its accuracy. The training is considered complete when the calculated MAPE is less than the preset MAPE. The formula for calculating MAPE is... Where p represents p time intervals, The actual power generation data after p time intervals. Let R be the predicted power generation data after the p-th time interval, and R be the number of predicted power generation data.

[0101] It should be noted that the preset MAPE is pre-set by those skilled in the art based on the required accuracy of the model.

[0102] Methods for predicting future load data include:

[0103] Future load data corresponds to the load data at n future time points for each user terminal; the load data of each user terminal in the historical load data is input into the trained load prediction model to predict the load data at n future time points for each user terminal; the training process of the load prediction model includes:

[0104] Collecting continuous Load data from historical time points are used to construct a load training set. Based on the load training set, a load prediction model is trained to predict load data for future time points. ;

[0105] Based on the practical experience of those skilled in the art, the sliding step size L and sliding window length are preset; the load data in the load training set is transformed into multiple training samples using the sliding window method. These training samples are used as input to the recurrent neural network model, and the load data after the sliding step size L is used as the output. Subsequent load data for each training sample is used as the prediction target, with prediction accuracy as the training objective. The mean absolute percentage error (MAPE) is used to evaluate the model accuracy. When the calculated MAPE is less than the preset MAPE, the load prediction model training is complete. The formula for calculating the mean absolute percentage error (MAPE) is as follows: ,in, For the forecast target corresponding to the f-th predicted load data, Let f be the predicted load data and H be the number of predicted load data. Generate a load prediction model based on the load data to predict the load data at future time points. The load prediction model is an RNN neural network model.

[0106] For example, suppose the load training set F contains 10 sets of load data. Multiple training samples are constructed using a sliding window. The length of the sliding window is defined as 3, and the sliding step size L is 1. Each training sample contains three consecutive load data points, and the next load data point in the three consecutive load data points is used as the prediction target. For example:

[0107] As training data, The corresponding prediction target is ;

[0108] As training data, The corresponding prediction target is Similarly, this is used to train the load prediction model.

[0109] It should be noted that the purpose of predicting future power generation and load data is to more accurately grasp the operating status of the distributed power grid, provide a basis for decision-making in formulating subsequent energy dispatch optimization strategies, and improve the accuracy and reliability of energy dispatch.

[0110] The energy optimization module is used by the cloud platform to formulate energy dispatch optimization strategies based on future power generation data and future load data, using the optimized Yanyu optimization algorithm.

[0111] The steps for formulating an energy scheduling optimization strategy using the optimized Remora algorithm include:

[0112] Step 1: Identify the load gap area and energy dispatch area;

[0113] Step 2: Preset the population size Z and the frequency threshold T;

[0114] Step 3: Initialize the population. The position of the remora in the population is defined in a one-dimensional search space, and the position of the remora corresponds one-to-one with the set label.

[0115] Step 4: Determine the fitness function;

[0116] Step 5: Select a host for each remora and update the location of each remora;

[0117] Step 6: Calculate the fitness of each remora after its location is updated, and determine whether to switch hosts;

[0118] Step 7: Update the location of the remora after switching hosts;

[0119] Step 8: Each remora enters the foraging phase, and the location of each remora is updated;

[0120] Step 9: Calculate the dynamic boundary, generate the reverse solution corresponding to each remora and add it to the population; also treat the reverse solutions of the remoras as remoras, filter the remoras in the population, and retain the Z remoras with the highest fitness.

[0121] Step 10: If the number of iterations t is less than the threshold T, return to step 5 and set the number of iterations... The iteration count is incremented by one and then assigned the iteration count; if the iteration count t is greater than or equal to the threshold T, proceed to step 11.

[0122] Step 11: Calculate the fitness corresponding to the position of each remora, sort all fitness values ​​from largest to smallest, obtain the position of the remora corresponding to the highest fitness value, and obtain the energy set corresponding to the set tag based on the obtained remora position.

[0123] In step 1 above, the future load data of the user terminals corresponding to each distribution area are added together sequentially to obtain the distributed load value of each distribution area; the future power generation data of the resource terminals corresponding to each distribution area are added together sequentially to obtain the distributed power generation value of each distribution area; the distribution area with the distributed power generation value greater than the distributed load value is marked as an energy dispatch area; the distribution area with the distributed power generation value less than the distributed load value is marked as a load gap area; the distribution area with the distributed power generation value equal to the distributed load value is not marked.

[0124] In step 2 above, the population size Z is determined by those skilled in the art during the historical distributed power grid energy dispatch optimization process. Under multiple sets of different future power generation data and future load data, multiple different population sizes are set for the same future power generation data and future load data, and the optimization algorithm is performed multiple times. After the same number of iterations, the corresponding set label is obtained. The population size with the same set label as the actual set label is taken as the population size corresponding to that set of future power generation data and future load data. The actual set label is the digital label of the energy set that best matches that set of future power generation data and future load data. The actual set label is obtained by those skilled in the art based on practical experience. The population size corresponding to each set of future power generation data and future load data is obtained in this way, and the average of multiple population sizes is taken as the preset population size Z.

[0125] The number of iterations threshold T is determined by those skilled in the art during the distributed power grid energy dispatch optimization process. Under multiple different sets of future power generation data and future load data, the same future power generation data and future load data are subjected to multiple iterations of the Threading algorithm to obtain multiple set labels. The number of iterations of the Threading algorithm is different for each iteration, but the population size is the same and is Z. The iteration number corresponding to the set label that is closest to the actual set label is taken as the iteration number corresponding to that set of future power generation data and future load data. This process is repeated to obtain the iteration number corresponding to each set of future power generation data and future load data, and the average of the multiple iteration numbers is taken as the number of iterations threshold T.

[0126] It should be understood that the population size Z determines the breadth of the search; a larger population size allows for the exploration of more solution spaces. The number of iterations threshold T determines the termination condition of the algorithm and controls the speed of convergence.

[0127] In step 3 above, the population is initialized. This includes Z remoras; distributed energy storage units in the energy dispatch area are marked as dispatch units; a dispatch value is randomly assigned to each dispatch unit corresponding to each energy dispatch area, and this is used as a dispatch set, with a total of W dispatch sets, where W is an integer greater than 1; the range of the dispatch value corresponding to each dispatch unit is... Y is the difference between the distributed load value and the corresponding distributed power generation value in the load gap area; the sum of the dispatch values ​​corresponding to each dispatch unit is Y; the sum of the dispatch values ​​of the dispatch units corresponding to each energy dispatch area is less than or equal to the difference between the corresponding distributed load value and the distributed power generation value, that is, the total amount of energy dispatched from the energy dispatch area to the load gap area should be less than its own excess power generation, so as to avoid the energy dispatch area's own power generation data being less than the load data.

[0128] Different numerical tags are assigned to different scheduling sets and marked as set tags. The range of set tags is... That is, the range of the one-dimensional search space is The expression for the position of each remora is: In the formula, Let be the initial position of the i-th remora. for The random number between the two elements is used to initialize the population with the corresponding iteration number t as 0.

[0129] It should be noted that in a distributed power grid, there are corresponding distributed energy storage units and distributed generation units. When a distributed generation unit generates electricity, the generated electrical energy is stored in the corresponding distributed energy storage unit.

[0130] In step 4 above, the expression for the fitness function is: In the formula, For fitness, This represents the scheduling overhead.

[0131] Methods for obtaining scheduling overhead values ​​include:

[0132] Based on the location of the remora, obtain the scheduling set corresponding to the set label; based on the power grid coordinate system, obtain the coordinates of each scheduling unit in the scheduling set and the scheduling value corresponding to each scheduling unit; calculate the Eulerian distance between the coordinates of each scheduling unit and the regional center coordinates of the load gap area, and mark it as the scheduling distance; the calculation method of the scheduling distance is the same as the calculation method of the Eulerian distance between the resource coordinates of the resource end and the regional center coordinates of the corresponding pre-divided area; add the scheduling distances of each scheduling unit in sequence to obtain the total scheduling distance.

[0133] The unit power generation cost is preset, which includes the cost required for each distributed power generation unit to produce one unit of electricity, where electricity is usually expressed in kilowatt-hours (kWh). The unit power generation cost corresponding to each scheduling unit is obtained based on the resource terminal corresponding to each scheduling unit. The unit power generation cost of each scheduling unit is multiplied by the corresponding scheduling value and then added together to obtain the total cost. The power generation cost is preset by those skilled in the art based on the actual situation.

[0134] The line loss value of each transmission line in the distributed power grid is obtained; based on the topology of the distributed power grid, the transmission lines between each dispatching unit and the load gap area are obtained and marked as dispatching lines; the line loss values ​​of each dispatching line are added together to obtain the total line loss value; the line loss value of each transmission line in the distributed power grid is obtained by those skilled in the art through the historical distributed power grid energy dispatch optimization process, by installing power sensors at both ends of each transmission line, collecting the power values ​​at both ends of each transmission line multiple times, subtracting the power value at the input end from the power value at the output end of the transmission line to obtain the power difference, and taking the average of the multiple power differences of each transmission line as the line loss value.

[0135] The total scheduling distance, total cost, and total line loss are used as analytical data. This data is then input into a trained loss prediction model to predict the corresponding scheduling loss value. The training process of the loss prediction model includes:

[0136] G sets of analysis data are collected in advance, and corresponding scheduling loss values ​​are set for each of the g sets of analysis data, where g is an integer greater than 1. The analysis data and the corresponding scheduling loss values ​​are converted into a set of corresponding feature vectors. The scheduling loss values ​​corresponding to the analysis data are collected by those skilled in the art during the historical distributed power grid energy dispatch optimization process. G sets of analysis data are collected, and corresponding scheduling loss values ​​are set based on actual experience under the conditions of each set of analysis data. The scheduling loss values ​​reflect the overall loss situation in the current energy dispatch process. The corresponding scheduling loss values ​​are set for the g sets of analysis data in sequence.

[0137] Each set of feature vectors is used as input to the loss prediction model. The model outputs a set of predicted scheduling loss values ​​corresponding to each set of analyzed data, and uses the actual scheduling loss value corresponding to each set of analyzed data as the prediction target. The actual scheduling loss value is the pre-collected scheduling loss value corresponding to the analyzed data. The training objective is to minimize the sum of prediction errors for all analyzed data. The formula for calculating the prediction error is as follows: ,in The prediction error is represented by K, where K is the group number of the feature vector corresponding to the analyzed data. Let K be the predicted scheduling cost value corresponding to the Kth set of analysis data. Let K be the actual scheduling loss value corresponding to the Kth set of analysis data; train the loss prediction model until the sum of prediction errors converges and then stop training.

[0138] The aforementioned loss prediction model is specifically a deep neural network model; it includes an input layer, hidden layers, and an output layer; each hidden layer contains multiple neurons, and each neuron is connected to the neurons in the next layer. The connections contain weights that determine the importance and impact of data transmission in the neural network; each neuron between the hidden layer and the output layer applies an activation function, which introduces non-linearity, allowing the network to learn more complex patterns and features.

[0139] In step 5 above, the hosts include sailfish and whales; obtain the random number corresponding to each remora. The system performs a rounding operation. If the rounded random number is 0, the host of the remora is a whale; if the rounded random number is 1, the host of the remora is a sailfish. The remora uses its parasitic host to effectively guide its search direction.

[0140] Calculate the fitness of each remora's location, and mark the location of the remora with the highest fitness value as... .

[0141] If the host of the parasite is a whale, the updated method for calculating the remora's location includes:

[0142] ;

[0143] ;

[0144] In the formula, This represents the position of the i-th remora after the update. This represents the position of the i-th remora before the update. Adjust the coefficients to broaden the search range. for Random numbers between intervals.

[0145] If the host fish is a sailfish, the updated method for calculating the remora's location includes:

[0146] ;

[0147] In the formula, This represents the location of a random remora before the update.

[0148] In step 6 above, when the remora parasitizes the host, it will make a tentative move based on the remora's final position in the previous iteration and its current position; calculate the fitness corresponding to the remora's tentative move and mark it as the tentative fitness; mark the fitness corresponding to the remora's current position as the current fitness; compare the tentative fitness with the current fitness; if the tentative fitness is greater than or equal to the current fitness, the host will not be switched; if the tentative fitness is less than the current fitness, the host will be switched.

[0149] It should be noted that the purpose of switching hosts is to indicate that when the fitness of the probe is less than the current fitness, it means that after the remora moves within a small area around the host, the fitness of the new location is low. This means that the current host location may be close to the local optimum, and continuing to explore around the host may not bring significant improvement. Therefore, by switching hosts, the remora can jump out of the current local area, explore a wider solution space, increase the chance of finding the global optimum, and avoid all individuals gathering in the same area, increasing the diversity of the population and making the algorithm more adaptable and robust.

[0150] In step 7 above, the method for updating the location of the remora after switching hosts is the same as the method for updating the location of each remora in step 5.

[0151] In step 8 above, the method for updating the location of the remora when it enters the foraging phase includes:

[0152] ;

[0153] ;

[0154] ;

[0155] ;

[0156] In the formula, Let be the position of the i-th remora after the foraging phase. The updated position after the i-th remora switches hosts. Let be the distance traveled by the i-th remora during the foraging phase. Let be the volume of the host on which the i-th remora parasitizes. This is a remora factor used to restrict the remora's location. , Let be the volume of the i-th remora. for Random numbers between intervals.

[0157] In step 9 above, the dynamic boundary is , The set label with the smallest numerical value in the population. The set label is the one with the largest numerical value in the population.

[0158] Methods for generating inverse solutions include:

[0159] ;

[0160] In the formula, Let be the reverse solution for the i-th remora. for Random numbers in the array.

[0161] If the generated inverse solution is greater than or less If the value is not found, it is marked as a transcendental solution. A value within the dynamic boundary is randomly generated using a random number function, and the randomly generated value is assigned to the transcendental solution.

[0162] It should be noted that the optimized remora optimization algorithm is used to formulate the energy scheduling optimization strategy because the remora optimization algorithm is highly flexible and adaptable. It can define the fitness function according to specific performance requirements and flexibly deal with different energy scheduling optimization problems. It also has excellent global search capabilities. By continuously updating the position of the remora parasitic host, it can explore a wide search space and increase the probability of finding the global optimum. At the same time, it also adopts host switching and foraging mechanisms, which can effectively overcome the problem of local optima and accelerate the convergence speed of the algorithm. In addition, the optimized remora optimization algorithm has low computational resource requirements, strong interpretability of results, and wide applicability.

[0163] This embodiment accurately divides the distributed power grid into regions based on historical power generation and load data to adapt to the dynamic changes of the distributed power grid. Furthermore, by predicting future power generation and load data, it uses an improved Yuyu optimization algorithm to perform regional analysis, comprehensively considering factors such as resource location and load differences to formulate energy dispatch optimization strategies. By fully leveraging the advantages of cloud computing, it achieves intelligent management and automated operation of the distributed power grid, effectively consolidating distributed energy resources, reducing energy consumption and costs in the energy dispatch process, improving the energy dispatch level of the distributed power grid, and enhancing the stable operation and sustainable development capabilities of the distributed power grid.

[0164] Example 2

[0165] Please see Figure 2 As shown, parts not described in detail in this embodiment are described in Embodiment 1. A distributed energy optimization decision-making method based on cloud computing is provided and applied in a cloud platform. The method includes:

[0166] Receive historical power generation data collected from the resource end and historical load data collected from the user end;

[0167] The distributed power grid is divided into regions based on historical power generation data and historical load data;

[0168] Based on historical power generation and load data, predict future power generation and load data;

[0169] Based on future power generation and load data, an optimized energy dispatching strategy is formulated using the optimized Yanyu optimization algorithm.

[0170] The above description is merely a specific embodiment of the present invention, but the scope of protection of the present invention is not limited thereto. Any variations or substitutions that can be easily conceived by those skilled in the art within the technical scope disclosed in the present invention should be included within the scope of protection of the present invention. Therefore, the scope of protection of the present invention should be determined by the scope of protection.

[0171] In conclusion, the above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the protection scope of the present invention.

Claims

1. A cloud computing based distributed energy optimization decision system, characterized in that, Be applied to a cloud platform, comprising: A data receiving module for receiving historical power generation data collected by a resource end and historical load data collected by a user end; the historical load data is load data of m consecutive historical time points corresponding to each user end, and the load data is the amount of electrical energy required by the user end for the distributed power grid within a preset time interval; the method for obtaining the load data is: obtaining the amount of electricity used, which is the amount of electrical energy used by the user end within the preset time interval; obtaining the user electricity, which is the amount of electrical energy generated by the user end within the preset time interval; subtracting the user electricity from the amount of electricity used to obtain the load data; A region division module for dividing the distributed power grid into regions according to the historical power generation data and the historical load data; the steps for dividing the distributed power grid into regions include: Step a: obtaining the topological structure of the distributed power grid and generating a power grid coordinate system; Step b: pre-dividing the distributed power grid into A pre-division regions; Step c: obtaining the region center coordinates of each pre-division region and the resource coordinates of each resource end; Step d: distributing all resource ends into A pre-division regions; Step e: integrating all resource ends into the corresponding pre-division regions respectively to re-divide A distribution regions; In step b, the distributed power grid is pre-divided according to the administrative regions under different cities, the pre-division regions correspond one-to-one to the administrative regions, and A is an integer greater than 1; A data prediction module for predicting future power generation data and future load data according to the historical power generation data and the historical load data; An energy optimization module for formulating an energy dispatching optimization strategy by using an optimized angelfish optimization algorithm according to the future power generation data and the future load data, and the specific steps include: Step 1: determining the load gap region and the energy dispatching region; Step 2: presetting the population size Z and the number threshold T; Step 3: initializing the population, the position of each angelfish in the initialized population is defined in a one-dimensional search space, and the position of each angelfish corresponds one-to-one to a set tag; Step 4: determining the fitness function; Step 5: selecting the host of each angelfish and updating the position of each angelfish; Step 6: calculating the fitness corresponding to the updated position of each angelfish and determining whether to switch the host; Step 7: updating the position of the angelfish that switches the host again; Step 8: each angelfish enters the foraging stage to update the position of each angelfish; Step 9: calculating the dynamic boundary, generating the reverse solution corresponding to each angelfish and adding it to the population; the reverse solution of the angelfish is also regarded as an angelfish, the angelfishes in the population are screened, and the Z angelfishes with the largest fitness are retained; Step 10: If the iteration number t is less than the number threshold T, return to Step 5 and let the iteration number t be t = t + 1. Step 11: If the iteration number t is greater than or equal to the number threshold T, enter Step 12. Step 12: Output the iteration number t. Step 11: calculating the fitness corresponding to the position of each angelfish, sorting all the fitnesses from large to small, obtaining the position of the angelfish corresponding to the largest fitness, and obtaining the energy set corresponding to the set tag according to the obtained angelfish position; In step 6, when the remora fish is parasitic on the host, the remora fish makes a tentative movement according to the final position of the remora fish in the last iteration and the current position of the remora fish; the fitness corresponding to the tentative movement of the remora fish is calculated and marked as tentative fitness; the fitness corresponding to the current position of the remora fish is marked as current fitness; the tentative fitness is compared with the current fitness; if the tentative fitness is greater than or equal to the current fitness, the host is not switched; if the tentative fitness is less than the current fitness, the host is switched; In step 7, the method for updating the position of the remora fish after switching the host is consistent with the method for updating the position of each remora fish in step 5.

2. The cloud computing based distributed energy optimization decision system according to claim 1, wherein, The historical power generation data is power generation data of each resource end corresponding to m continuous historical time points, the power generation data is the power generation amount of the resource end in a preset time interval, m is an integer greater than 1; the power generation data and the load data are collected according to the preset time interval, and the time difference between every two time points is the preset time interval. 3.The cloud-computing-based distributed energy optimization decision system according to claim 1, wherein, In step c, the coordinates of the B vertices in each pre-division region are obtained; radial sorting or angle sorting is used to sort all vertex coordinates of each pre-division region; The area of each pre-division region is calculated; the expression of the area of the pre-division region is: ; wherein, is the area of the pre-division region, is the horizontal coordinate of the Ith vertex, is the vertical coordinate of the Ith vertex, , vertex is regarded as the vertex ; the center coordinate of each pre-division region is calculated according to the area of each pre-division region; the expression of the center coordinate is: , ; wherein, is the center horizontal coordinate, is the center vertical coordinate.

4. The cloud computing based distributed energy optimization decision system according to claim 1, wherein, In step d, the steps of resource end allocation include: Step d1: according to the power grid coordinate system, count the number of resource ends ; a preset resource subset, , the resource subset corresponds to the pre-division area one by one; randomly allocate resource ends to resource subsets, and each resource subset is different resource end; take resource subsets as a resource set; Step d2: loop step d1 until M different resource sets are obtained, and the loop ends and step d3 is entered; Step d3: Calculate the distance total value and the load total difference of M resource sets respectively; select the best resource set from M resource sets, and allocate the resource end to A pre-regional areas. Step d3: Calculate the distance total value and the load total difference of M resource sets respectively; select the best resource set from M resource sets, and allocate the resource end to A pre-regional areas.

5. The cloud computing based distributed energy optimization decision system according to claim 4, wherein, The calculation method of the distance total value and the load total difference includes: The Euler distance between the resource coordinates of each resource end in each resource set and the region center coordinates corresponding to the corresponding resource sub-set is calculated; the Euler distances corresponding to all resource ends in each resource set are sequentially added to obtain the distance total value corresponding to each resource set; the average power generation of each resource end is calculated according to the historical power generation data; the average load of each user end is calculated according to the historical load data; the average power generation corresponding to each resource end in each resource sub-set is sequentially added to obtain the total power generation of each resource sub-set; the average load of each user end in each resource sub-set is sequentially added to obtain the total load of each resource sub-set; the total load of each resource sub-set is subtracted from the corresponding total power generation and the absolute value is taken to obtain the load difference corresponding to each resource sub-set; the load differences corresponding to each resource sub-set are sequentially added to obtain the load total difference corresponding to each resource set.

6. The cloud computing based distributed energy optimization decision system, as recited in claim 1, wherein, The method for predicting future power generation data includes: The future power generation data is power generation data of each resource end corresponding to n future time points; future meteorological data is obtained, and the future meteorological data is meteorological data of each distribution region corresponding to n future time points; The future power generation data is power generation data of each resource end corresponding to n future time points; future meteorological data is obtained, and the future meteorological data is meteorological data of each distribution region corresponding to n future time points; According to historical power generation data, the power generation data time series of each resource end is constructed; the power generation data time series of each resource end and the corresponding predicted meteorological data are respectively input into the trained power prediction model to predict the power generation data of each resource end at a time interval and marked as predicted power; the predicted meteorological data is the meteorological data corresponding to each resource end at the same time point as the predicted power; then each predicted power is replaced with the power generation data at the earliest time in the corresponding power generation data time series to obtain a new power generation data time series of each resource end, and the new power generation data time series of each resource end and the corresponding predicted meteorological data are input into the trained power prediction model to predict the predicted power again; n predicted powers are predicted.

7. The cloud computing based distributed energy optimization decision system, as claimed in claim 1, wherein, In step 1, the future load data of each distribution area corresponding to the user end is added in sequence to obtain the distribution load value of each distribution area; the future power generation data of each distribution area corresponding to the resource end is added in sequence to obtain the distribution power generation value of each distribution area; the distribution area with a distribution power generation value greater than the distribution load value is marked as an energy scheduling area; the distribution area with a distribution power generation value less than the distribution load value is marked as a load gap area; and the distribution area with a distribution power generation value equal to the distribution load value is not marked. 8.The cloud-computing-based distributed energy optimization decision system according to claim 7, wherein, The distributed energy storage unit in the energy scheduling area is marked as a scheduling unit; a scheduling value is randomly set for each scheduling unit corresponding to each energy scheduling area, and serves as a scheduling set, and W scheduling sets are set in total; wherein the scheduling value corresponding to each scheduling unit ranges from 0 to Y , Y is the difference between the distributed load value of the load gap area and the corresponding distributed power generation value; the sum of the scheduling values corresponding to each scheduling unit is Y; the sum of the scheduling values of the scheduling units corresponding to each energy scheduling area is less than or equal to the difference between the corresponding distributed load value and the distributed power generation value; different scheduling sets are set with different digital labels and are marked as set labels, and the range of the set labels and the one-dimensional search space is ; the expression of the position of each fish is: ; wherein, is the initial position of the ith fish, is a random number between 0 and 1, the iteration number t corresponding to the initialization population is 0; The step 5, the host includes sailfish and whale; obtain the random number corresponding to each oarfish And rounding operation, if the random number after rounding is 0, the host of oarfish parasitism is whale, if the random number after rounding is 1, the host of oarfish parasitism is sailfish. 9.The cloud-computing-based distributed energy optimization decision system of claim 8, wherein, In step 4, the expression of the fitness function is: wherein, is the fitness, is the scheduling penalty value; The method for obtaining the scheduling loss value comprises: According to the position of the fish, a scheduling set corresponding to the set label is obtained; according to the power grid coordinate system, the coordinates of each scheduling unit in the scheduling set and the scheduling value corresponding to each scheduling unit are obtained; the Euler distance between the coordinates of each scheduling unit and the regional center coordinates of the load gap area is calculated and marked as a scheduling distance; the scheduling distances of each scheduling unit are added in sequence to obtain a total scheduling distance; a preset unit power generation cost is obtained, which includes the cost required by each distributed power generation unit to generate one unit of power; according to the resource end corresponding to each scheduling unit, the unit power generation cost corresponding to each scheduling unit is obtained; the unit power generation cost of each scheduling unit is multiplied by the corresponding scheduling value and added in sequence to obtain a total cost value; the line loss value of each transmission line in the distributed power grid is obtained; according to the topological structure of the distributed power grid, the transmission line between each scheduling unit and the load gap area is obtained and marked as a scheduling line; the line loss values of each scheduling line are added in sequence to obtain a total line loss value; the total scheduling distance, the total cost value and the total line loss value are taken as analysis data, and the analysis data is input into the trained loss prediction model to predict the corresponding scheduling loss value.

Images