A method for predicting the output of multi-energy complementary power plants based on dynamic group cooperation optimization

Abstract

This invention proposes a power output prediction method for multi-energy complementary power plants based on dynamic grouping and cooperative optimization to address the accuracy and reliability issues in power output prediction for such plants. First, it analyzes in detail the characteristics and coupling relationships of various energy generation methods within a multi-energy complementary power plant, as well as the uncertainties arising from environmental factors affecting each energy generation. Then, through a dynamic grouping strategy, based on the correlation and complementarity of energy generation, different energy generation units within the multi-energy complementary power plant are rationally grouped to form dynamically changing group structures. This technical solution more accurately captures the power output change trends of multi-energy complementary power plants, effectively reducing prediction errors. It provides a more reliable basis for power output prediction for the operation and scheduling of multi-energy complementary power plants and electricity market transactions, thereby improving the operation and management level and economic benefits of multi-energy complementary power plants.

Description

Technical Field

[0001] This invention relates to the field of multi-energy complementary power output prediction, and in particular to a method for predicting the power output of multi-energy complementary power plants based on dynamic group cooperation optimization. Background Technology

[0002] Multi-energy complementary power plants, by integrating various energy forms such as solar, wind, hydro, and biomass energy, can effectively mitigate the output fluctuations of single renewable energy sources, becoming a key carrier for improving energy supply stability and comprehensive utilization efficiency. Output forecasting for such power plants is a core prerequisite for operation scheduling, electricity market transactions, and energy storage configuration optimization. Its forecasting accuracy directly determines the economic benefits of the power plant and the safe and stable operation of the power grid—accurate output forecasting can reduce wind and solar curtailment rates, decrease reserve capacity configuration costs, and provide a reliable basis for power grid dispatching departments to formulate inter-provincial and inter-regional power balance plans.

[0003] However, power output prediction for multi-energy complementary power plants faces multiple technical challenges: on the one hand, the output of renewable energy sources such as solar and wind power is significantly affected by the strong randomness of environmental factors such as solar irradiance, wind speed, and temperature, exhibiting significant intermittent and fluctuating characteristics; on the other hand, there are complex coupling relationships between different energy types (e.g., increased hydropower output during the rainy season can compensate for insufficient wind power output, and solar and wind power output may have a complementary relationship of waxing and waning in sunny weather). Traditional prediction methods often ignore this dynamic coupling mechanism or use static grouping strategies to predict each energy unit in isolation, making it difficult to fully explore the synergistic effects between energy sources. In existing research, prediction methods based on single machine learning models (such as BP neural networks and LSTM) can capture some nonlinear laws, but their ability to characterize multi-energy coupling relationships is limited; while collaborative prediction methods based on fixed grouping cannot adapt to the dynamic evolution of energy correlation caused by changes in environmental parameters, and the prediction error increases significantly under complex operating conditions.

[0004] To address the aforementioned issues, this paper proposes a power output prediction method for multi-energy complementary power plants based on dynamic grouping cooperative optimization. This method first systematically analyzes the characteristics and coupling relationships of different energy sources, quantifying the impact of environmental factors on power output uncertainty. Then, through a dynamic grouping strategy, it adjusts the group structure based on the real-time correlation and complementarity between energy sources, breaking the limitations of traditional static grouping. Within each group, a cooperative optimization algorithm is used to integrate historical power output data, real-time operating status, and environmental parameters to achieve accurate power output prediction. Simultaneously, an information exchange mechanism is established between groups to further optimize overall prediction accuracy through dynamic information sharing. Through case studies of actual power plants and simulation experiments, the advantages of this method in capturing power output change trends and reducing prediction errors are verified, providing technical support for the refined management of multi-energy complementary power plants. Summary of the Invention

[0005] The purpose of this invention is to provide a method for predicting the output of multi-energy complementary power plants based on dynamic group cooperative optimization, so as to solve the problems mentioned in the background art.

[0006] To achieve the above objectives, the present invention adopts the following technical solution: a method for predicting the output of a multi-energy complementary power station based on dynamic grouping cooperation optimization, comprising the following steps to implement the proposed method for predicting the output of a multi-energy complementary power station based on dynamic grouping cooperation optimization:

[0007] S1: Construct a mathematical model of the single energy output in a multi-energy complementary power station;

[0008] S2: Defines the coupling coefficient between different single-energy output models;

[0009] S3: Establish an objective function with the goal of maximizing energy complementarity within groups and minimizing intergroup correlation;

[0010] S4: Dynamic group partitioning is achieved using the K-means clustering algorithm;

[0011] S5: Establish an objective function with the goal of minimizing the prediction error of the output within the group;

[0012] S6: The particle swarm optimization algorithm is used to predict the output within the group;

[0013] S7: Define the inter-group information sharing coefficient and correct the overall predicted output of multi-energy complementary power plants;

[0014] S8: Evaluation model based on root mean square error and mean absolute percentage error;

[0015] As a further improvement to this technical solution, the following steps are included to construct a mathematical model for the output of a single energy source in a multi-energy complementary power station:

[0016] Solar energy in multi-energy complementary power plants Wind energy Hydropower Biomass energy For different energy types, establish output mathematical models based on environmental factors;

[0017] The solar power output model is represented as follows:

[0018] (1)

[0019] in, for The solar power unit is constantly generating power. For photovoltaic module conversion efficiency, For the area of ​​the photovoltaic array, for Solar irradiance at any time The temperature coefficient of photovoltaic modules. for ambient temperature at all times Standard test temperature;

[0020] The wind energy output model is represented as follows:

[0021] (2)

[0022] in, for Wind power unit output at all times Let t be the wind speed. To cut into wind speed, To cut off the wind speed, Rated wind speed, Rated output of the wind power unit;

[0023] The hydropower output model is represented as follows:

[0024] (3)

[0025] in, for The hydroelectric unit outputs power at all times. The density of water, It is the acceleration due to gravity. for Real-time machine traffic flow for Always clean the water head, For hydropower unit efficiency;

[0026] The biomass energy output model is expressed as follows:

[0027] (4)

[0028] in, for The output of the biomass power generation unit is constantly being monitored. For biomass energy conversion efficiency, for Real-time biomass fuel consumption It provides the lower heating value of biomass fuel;

[0029] As a further improvement to this technical solution, the following steps are included to define the coupling coefficient between different single energy output models:

[0030] Define the coupling coefficient Describing energy With energy exist The coupling strength at time t is calculated using the following formula:

[0031] (5)

[0032] in, For energy With energy Covariance of output , Energy , Variance of output; when At that time, the power outputs of the two energy sources are positively coupled; They exhibit negative coupling (complementarity); There is no coupling at that time.

[0033] As a further improvement to this technical solution, the following steps are included to establish an objective function with the goals of maximizing intra-group energy complementarity and minimizing inter-group correlation:

[0034] With the objectives of maximizing intra-group energy complementarity and minimizing inter-group correlation, the objective function is established as follows:

[0035] (6)

[0036] in, For the number of groups, For the first A group is a collection of energy units.

[0037] As a further improvement to this technical solution, the following steps are included for using the K-means clustering algorithm to achieve dynamic group partitioning:

[0038] The K-means clustering algorithm is used to achieve dynamic group partitioning. The specific steps are as follows:

[0039] 1) Random selection Each energy unit serves as the initial cluster center. ;

[0040] 2) Calculate each energy unit To each cluster center The distance then the energy unit Assign to the group with the smallest distance ,

[0041] (7)

[0042] 3) Recalculate each group Cluster centers

[0043] (8)

[0044] 4) Repeat steps 2-3 until the cluster centers no longer change or the iteration threshold is met, to obtain the dynamic group structure at time t. ;

[0045] As a further improvement to this technical solution, the following steps are included to establish an objective function with the goal of minimizing the in-group output prediction error:

[0046] With the goal of minimizing the power output prediction error within the group, and considering historical power output data, real-time operating status, and environmental parameters, the objective function is established as follows:

[0047] (9)

[0048] in, For the first The root mean square error of prediction for the group. For historical data sample size, For groups exist The actual total output at any given moment For groups exist Total output predicted at any given moment;

[0049] As a further improvement to this technical solution, the following steps are included to use the particle swarm optimization algorithm to predict the output within a group:

[0050] First, the particles are encoded, with each particle representing a set of prediction model parameters. ,in Define the model parameter dimension; define the fitness function, where a larger value indicates higher prediction accuracy.

[0051] (10)

[0052] Update the particle's velocity and position.

[0053] (11)

[0054] in, For inertial weights, , As a learning factor, , for Random numbers, For particles The best historical position This is the globally optimal position for the entire population;

[0055] Determine the optimal parameters: Iterate until the fitness function converges to obtain the optimal parameters. Substitute the values ​​into the prediction model to complete the power output prediction within the group.

[0056] (12)

[0057] in, For groups The input feature vector includes environmental parameters, historical output, etc.

[0058] As a further improvement to this technical solution, the following steps are included to define the inter-group information sharing coefficient and correct the overall predicted output of the multi-energy complementary power station:

[0059] Define the inter-group information sharing coefficient , indicating the first Groups and the first The degree of information sharing within the group:

[0060] (13)

[0061] based on , group The formula for correcting the predicted output is:

[0062] (14)

[0063] in, For groups Historical average output For the corrected group Predicted output;

[0064] The overall predicted output of the multi-energy complementary power station is the sum of the corrected outputs of each group:

[0065] (15)

[0066] As a further improvement to this technical solution, the following steps are included for evaluating the model based on root mean square error and mean absolute percentage error:

[0067] Define root mean square error for:

[0068] (16)

[0069] Define mean absolute percentage error for:

[0070] (17)

[0071] in, To determine the size of the validation set, For power station The actual total output at all times.

[0072] Compared with existing technologies, the beneficial effects of this invention are as follows: This paper proposes a method for predicting the output of multi-energy complementary power plants based on dynamic grouping cooperative optimization, to solve the problems of accuracy and reliability in the output prediction of multi-energy complementary power plants. First, the characteristics of various energy generation methods in a multi-energy complementary power plant and their coupling relationships, as well as the uncertainties of each energy generation method affected by environmental factors, are analyzed in detail. Through a dynamic grouping strategy, based on the correlation and complementarity of energy generation, different energy generation units in the multi-energy complementary power plant are rationally grouped to form a dynamically changing group structure. Within each group, a cooperative optimization algorithm is used to fully explore the synergistic effect between each energy generation unit, comprehensively considering historical output data, real-time operating status, and environmental parameters of each unit to optimize and predict the group output. Simultaneously, an information exchange mechanism between groups is established to achieve dynamic cooperation and information sharing between different groups, further improving the accuracy of the overall output prediction of the multi-energy complementary power plant. Through case studies and simulation experiments of actual multi-energy complementary power plants, the results show that this method, compared with traditional prediction methods, can more accurately capture the output change trend of multi-energy complementary power plants, effectively reduce prediction errors, provide more reliable output prediction basis for the operation and scheduling of multi-energy complementary power plants and electricity market transactions, and improve the operation and management level and economic benefits of multi-energy complementary power plants.

[0073] The above description is merely an overview of the technical solution of the present invention. In order to better understand the technical means of the present invention and to implement it according to the contents of the specification, the preferred embodiments of the present invention are described in detail below with reference to the accompanying drawings. Specific embodiments of the present invention are given in detail below with reference to the accompanying drawings. Attached Figure Description

[0074] Figure 1 This is a flowchart of the multi-energy complementary power plant output prediction method based on dynamic group cooperation optimization, which is the subject of this invention patent.

[0075] Figure 2 This is a comparison chart of the actual output of the training set and the predicted output of the test set in this invention patent.

[0076] Figure 3 This is a comparison chart of the prediction errors of the two methods described in this invention patent. Detailed Implementation

[0077] The principles and features of the present invention are described below with reference to the accompanying drawings. The examples given are for illustrative purposes only and are not intended to limit the scope of the invention. The invention is described more specifically in the following paragraphs by way of example with reference to the accompanying drawings. The advantages and features of the invention will become clearer from the following description and claims. It should be noted that the drawings are in a very simplified form and use non-precise proportions, and are only used to facilitate and clarify the illustration of the embodiments of the invention.

[0078] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this invention pertains. The terminology used herein in the description of the invention is for the purpose of describing particular embodiments only and is not intended to be limiting of the invention.

[0079] This embodiment is based on Figure 1 The flowchart shown is a method for predicting the output of a multi-energy complementary power plant based on dynamic grouping cooperation optimization. The method for predicting the output of a multi-energy complementary power plant based on dynamic grouping cooperation optimization is analyzed.

[0080] Please see the appendix Figure 1 A method for predicting the output of a multi-energy complementary power plant based on dynamic grouping cooperation optimization includes the following steps to implement the proposed method:

[0081] S1: Construct a mathematical model of the single energy output in a multi-energy complementary power station;

[0082] S2: Defines the coupling coefficient between different single-energy output models;

[0083] S3: Establish an objective function with the goal of maximizing energy complementarity within groups and minimizing intergroup correlation;

[0084] S4: Dynamic group partitioning is achieved using the K-means clustering algorithm;

[0085] S5: Establish an objective function with the goal of minimizing the prediction error of the output within the group;

[0086] S6: The particle swarm optimization algorithm is used to predict the output within the group;

[0087] S7: Define the inter-group information sharing coefficient and correct the overall predicted output of multi-energy complementary power plants;

[0088] S8: Evaluation model based on root mean square error and mean absolute percentage error;

[0089] Furthermore, the following steps are included for constructing a mathematical model of the single energy source output in a multi-energy complementary power plant:

[0090] Solar energy in multi-energy complementary power plants Wind energy Hydropower Biomass energy For different energy types, establish output mathematical models based on environmental factors;

[0091] The solar power output model is represented as follows:

[0092] (1)

[0093] in, for The solar power unit is constantly generating power. For photovoltaic module conversion efficiency, For the area of ​​the photovoltaic array, for Solar irradiance at any time The temperature coefficient of photovoltaic modules. for ambient temperature at all times Standard test temperature;

[0094] The wind energy output model is represented as follows:

[0095] (2)

[0096] in, for Wind power unit output at all times for Wind speed at all times To cut into wind speed, To cut off the wind speed, Rated wind speed, Rated output of the wind power unit;

[0097] The hydropower output model is represented as follows:

[0098] (3)

[0099] in, for The hydroelectric unit outputs power at all times. The density of water, It is the acceleration due to gravity. for Real-time machine traffic flow The head of the clean water at time t. For hydropower unit efficiency;

[0100] The biomass energy output model is expressed as follows:

[0101] (4)

[0102] in, for The output of the biomass power generation unit is constantly being monitored. For biomass energy conversion efficiency, for Real-time biomass fuel consumption It provides the lower heating value of biomass fuel;

[0103] Furthermore, the following steps are included to define the coupling coefficients between different single-energy output models:

[0104] Define the coupling coefficient Describing energy With energy exist The coupling strength at time t is calculated using the following formula:

[0105] (5)

[0106] in, For energy With energy Covariance of output , Energy , The variance of output. When At that time, the power outputs of the two energy sources are positively coupled; They exhibit negative coupling (complementarity); There is no coupling at that time.

[0107] Furthermore, the steps include establishing an objective function that aims to maximize intra-group energy complementarity and minimize inter-group correlation:

[0108] With the objectives of maximizing intra-group energy complementarity and minimizing inter-group correlation, the objective function is established as follows:

[0109] (6)

[0110] in, For the number of groups, For the first A group is a collection of energy units.

[0111] Furthermore, the following steps are included for implementing dynamic group partitioning using the K-means clustering algorithm:

[0112] The K-means clustering algorithm is used to achieve dynamic group partitioning. The specific steps are as follows:

[0113] 1) Random selection Each energy unit serves as the initial cluster center. ;

[0114] 2) Calculate the distance from each energy unit i to each cluster center. The distance is then used to allocate energy unit i to the group with the smallest distance. ,

[0115] (7)

[0116] 3) Recalculate each group Cluster centers

[0117] (8)

[0118] 4) Repeat steps 2-3 until the cluster centers no longer change or the iteration threshold is met, to obtain the dynamic group structure at time t. ;

[0119] Furthermore, the steps include establishing an objective function with the goal of minimizing the within-group output prediction error:

[0120] With the goal of minimizing the power output prediction error within the group, and considering historical power output data, real-time operating status, and environmental parameters, the objective function is established as follows:

[0121] (9)

[0122] in, For the first The root mean square error of prediction for the group. For historical data sample size, For groups exist The actual total output at any given moment For groups exist Total output predicted at any given moment;

[0123] Furthermore, the method includes the following steps for predicting intra-group output using a particle swarm optimization algorithm:

[0124] First, the particles are encoded, with each particle representing a set of prediction model parameters. ,in Define the model parameter dimension; define the fitness function, where a larger value indicates higher prediction accuracy.

[0125] (10)

[0126] Update the particle's velocity and position.

[0127] (11)

[0128] in, For inertial weights, , As a learning factor, , for Random numbers, For particles The best historical position This is the globally optimal position for the entire population;

[0129] Determine the optimal parameters: Iterate until the fitness function converges to obtain the optimal parameters. Substitute the values ​​into the prediction model to complete the power output prediction within the group.

[0130] (12)

[0131] in, For groups The input feature vector includes environmental parameters, historical output, etc.

[0132] Furthermore, the following steps are included to define the inter-group information sharing coefficient and correct the overall predicted output of the multi-energy complementary power plant:

[0133] Define the inter-group information sharing coefficient , indicating the degree of information sharing between the k-th group and the l-th group:

[0134] (13)

[0135] based on , group The formula for correcting the predicted output is:

[0136] (14)

[0137] in, For groups Historical average output For the corrected group Predicted output;

[0138] The overall predicted output of the multi-energy complementary power station is the sum of the corrected outputs of each group:

[0139] (15)

[0140] Furthermore, the following steps are included for evaluating the model based on root mean square error and mean absolute percentage error:

[0141] Define root mean square error for:

[0142] (16)

[0143] Define mean absolute percentage error for:

[0144] (17)

[0145] in, To determine the size of the validation set, This represents the actual total output of the power station at time t.

[0146] This embodiment takes a small multi-energy complementary power station that includes photovoltaic (PV), wind power (WT), and small hydropower (HYD) as the research object, with a time scale of 48 hours (1 hour / step). The core objective is to verify the predictive advantages of the "dynamic group cooperative optimization" method compared with the traditional linear regression method.

[0147] The case design follows a technical path of "data simulation → dynamic grouping → group prediction → fusion evaluation": Data simulation: Based on the physical characteristics of energy, output data that conforms to actual laws is generated (e.g., photovoltaic power varies with diurnal irradiance, wind power fluctuates with the cubic relationship of wind speed, and hydropower outputs stably with water inflow), while random noise is added to simulate environmental uncertainty. Dynamic grouping: Based on "energy output correlation" (judged by Pearson correlation coefficient), photovoltaic and wind power are merged into one group when the correlation is >0.5; otherwise, the three are grouped independently to simulate the dynamic changes in energy coupling relationships in actual power plants. Prediction and fusion: Linear regression is used within groups to mine the output patterns of individual groups, and groups are merged through fixed weights (photovoltaics 0.35, wind power 0.4, hydropower 0.25) to balance the output proportion of different energy sources; traditional methods directly perform linear regression on the total output as a comparison benchmark.

[0148] Figure 1 This is a comparison chart of the actual output of the training set and the predicted output of the test set. The upper subplot shows the actual total output of the power station in the training set (1-32 hours); the blue solid line shows the change in total output in the first 32 hours of the 48-hour period, exhibiting a "double-peak fluctuation" characteristic: 6-18 hours: total output remains in the range of 15-25MW, with the peak concentrated in the range of 10-14 hours (PV output reaches 8-10MW due to the peak irradiance; wind power is stable at 6-8MW, and hydropower is stable at 4-6MW, with the three superimposed to form a peak); 19-5 hours (the next day): total output drops to the range of 5-12MW, supported only by wind power (3-5MW) and hydropower (4-6MW), while PV output is close to 0 due to the lack of irradiance, consistent with the nighttime energy supply pattern. The training set data realistically simulates the "day-night output difference" and "energy complementarity characteristics" of multi-energy complementary power stations, providing training samples that conform to physical laws for subsequent prediction models, ensuring that the model learns the essential laws of energy output rather than noise.

[0149] The subplot below shows a comparison of the prediction results for the test set (33-48 hours). The blue solid line (actual total output), the red solid line (dynamic cluster prediction value), and the green dashed line (traditional method prediction value) correspond to the time axis of 33-48 hours (i.e., from 17:00 on the next day to 16:00 on the third day).

[0150] Figure 2 The chart compares the prediction errors of the two methods (bar chart). The horizontal axis represents the two prediction methods (dynamic swarm method and traditional method), and the vertical axis represents the mean absolute error (MAE, unit: MW). The blue bars represent the dynamic swarm method, and the gray bars represent the typical operating results of the traditional method (the error value fluctuates slightly due to random noise): Dynamic swarm method MAE: 1.8-2.2MW; Traditional method MAE: 3.5-4.0MW. The dynamic swarm cooperative optimization method reduces the mean absolute error by approximately 45%-50% compared to the traditional method within a 48-hour prediction period. Its prediction accuracy advantage is particularly significant in key scenarios such as "photovoltaic sunrise / sunset" and "sudden changes in wind speed," validating the technical value of "grouping to mine synergistic effects."

[0151] The above description is merely a preferred embodiment of the present invention and is not intended to limit the present invention in any way. Those skilled in the art can readily implement the present invention based on the accompanying drawings and the above description. However, any modifications, alterations, or variations made by those skilled in the art without departing from the scope of the present invention, utilizing the disclosed technical content, are equivalent embodiments of the present invention. Furthermore, any modifications, alterations, or variations made to the above embodiments based on the essential technology of the present invention are still within the protection scope of the present invention.

Claims

1. A method for predicting the output of a multi-energy complementary power plant based on dynamic group cooperation optimization, characterized in that, Includes the following steps: S1: Construct a mathematical model of the single energy output in a multi-energy complementary power station; The following steps are used to construct a mathematical model for the output of a single energy source in a multi-energy complementary power plant: Solar energy in multi-energy complementary power plants Wind energy Hydropower Biomass energy For each energy type, establish an output mathematical model based on environmental factors; The solar power output model is represented as follows: (1) in, for The solar power unit is constantly generating power. For photovoltaic module conversion efficiency, For the area of ​​the photovoltaic array, for Solar irradiance at any time The temperature coefficient of photovoltaic modules. for ambient temperature at all times Standard test temperature; The wind energy output model is represented as follows: (2) in, for Wind power unit output at all times Let t be the wind speed. To cut into wind speed, To cut off the wind speed, Rated wind speed, Rated output of the wind power unit; The hydropower output model is represented as follows: (3) in, for The hydroelectric unit outputs power at all times. The density of water, It is the acceleration due to gravity. for Real-time machine traffic flow The head of the clean water at time t. For hydropower unit efficiency; The biomass energy output model is expressed as: (4) in, for The output of the biomass power generation unit at all times For biomass energy conversion efficiency, for Real-time biomass fuel consumption It provides the lower heating value for biomass fuel; S2: Defines the coupling coefficient between different single-energy output models; S3: Establish an objective function with the goal of maximizing energy complementarity within groups and minimizing intergroup correlation; The objective function is established by following these steps, with the goal of maximizing intra-group energy complementarity and minimizing inter-group correlation: With the objectives of maximizing intra-group energy complementarity and minimizing inter-group correlation, the objective function is established as follows: (6) in, For the number of groups, For the first Each group contains a set of energy units; S4: Dynamic grouping is achieved using the K-means clustering algorithm; S5: Establish an objective function with the goal of minimizing the prediction error of the output within the group; The objective function for minimizing the within-group output prediction error is established through the following steps: With the goal of minimizing the power output prediction error within the group, and considering historical power output data, real-time operating status, and environmental parameters, the objective function is established as follows: (9) in, For the first The root mean square error of prediction for the group. For historical data sample size, For groups exist The actual total output at any given moment For groups exist Total output predicted at any given moment; S6: The particle swarm optimization algorithm is used to predict the output within the group; S7: Define the inter-group information sharing coefficient and correct the overall predicted output of multi-energy complementary power plants; S8: Evaluation model based on root mean square error and mean absolute percentage error.

2. The method for predicting the output of a multi-energy complementary power station based on dynamic group cooperation optimization as described in claim 1, characterized in that, The following steps are included to define the coupling coefficients between different single-energy output models: Define the coupling coefficient Describe energy With energy exist The coupling strength at time t is calculated using the following formula: (5) in, For energy With energy Covariance of output , Energy , Variance of output; when At that time, the power outputs of the two energy sources are positively coupled; They exhibit negative coupling (complementarity); There is no coupling at that time.

3. The method for predicting the output of a multi-energy complementary power station based on dynamic group cooperation optimization according to claim 1, characterized in that, The following steps are included for dynamic group partitioning using the K-means clustering algorithm: The K-means clustering algorithm is used to achieve dynamic group partitioning. The specific steps are as follows: 1) Random selection Each energy unit serves as the initial cluster center. ; 2) Calculate each energy unit To each cluster center The distance then the energy unit Assign to the group with the smallest distance , (7) 3) Recalculate each group Cluster centers (8) 4) Repeat steps 2-3 until the cluster centers no longer change or the iteration threshold is met, thus obtaining... Dynamic group structure at time .

4. The method for predicting the output of a multi-energy complementary power station based on dynamic group cooperation optimization according to claim 1, characterized in that, The following steps are included for using the particle swarm optimization algorithm to predict the output within a group: First, the particles are encoded, with each particle representing a set of prediction model parameters. ,in Define the model parameter dimension; define the fitness function, where a larger value indicates higher prediction accuracy. (10) Update the particle's velocity and position. (11) in, For inertial weights, , As a learning factor, , for Random numbers, For particles The best historical position This is the globally optimal position for the entire population; Determine the optimal parameters: Iterate until the fitness function converges to obtain the optimal parameters. Substitute the values ​​into the prediction model to complete the power output prediction within the group. (12) in, For groups The input feature vector includes environmental parameters and historical output.

5. The method for predicting the output of a multi-energy complementary power station based on dynamic group cooperation optimization according to claim 1, characterized in that, The following steps are included to define the inter-group information sharing coefficient: Define the inter-group information sharing coefficient , indicating the first Groups and the first The degree of information sharing within the group: (13)。 6. The method for predicting the output of a multi-energy complementary power station based on dynamic group cooperation optimization according to claim 1, characterized in that, The following steps are included to correct the overall projected output of a multi-energy complementary power plant: Based on inter-group information sharing coefficient , group The formula for correcting the predicted output is: (14) in, For groups Historical average output For the corrected group Predicted output; The overall predicted output of the multi-energy complementary power station is the sum of the corrected outputs of each group: (15)。 7. The method for predicting the output of a multi-energy complementary power station based on dynamic group cooperation optimization according to claim 1, characterized in that, The following steps are included for evaluating models based on root mean square error and mean absolute percentage error: Define root mean square error for: (16) Define mean absolute percentage error for: (17) in, To determine the size of the validation set, This represents the actual total output of the power station at time t.

Images