Artificial intelligence-based photovoltaic power generation and energy storage cabinet power optimization distribution method

Abstract

The present application relates to the technical field of power dispatch, in particular to a power optimization distribution method for photovoltaic power generation and energy storage electric cabinet based on artificial intelligence. Meteorological prediction is performed to obtain a future meteorological data sequence in a future time window and collect a current running time; according to the future meteorological data sequence and the running time, photovoltaic power generation prediction and power demand prediction are performed to obtain a future power generation data sequence and a future power consumption data sequence; according to the running time, the confidence of the future meteorological data sequence and the future power consumption data sequence is analyzed to obtain a meteorological confidence sequence and a power consumption confidence sequence; according to the future power generation data sequence, the future power consumption data sequence, the meteorological confidence sequence and the power consumption confidence sequence, power distribution parameter configuration and optimization are performed to obtain optimized power distribution parameters, and power distribution power supply and energy storage of the energy storage electric cabinet in the future time window are performed. The stability and economy of the power system are significantly improved.

Description

Technical Field

[0001] This invention relates to the field of power dispatching technology, and in particular to a power optimization allocation method for photovoltaic power generation and energy storage cabinets based on artificial intelligence. Background Technology

[0002] In traditional photovoltaic (PV) power generation and energy storage systems, power allocation is often based on simple rules or experience, lacking accurate prediction of future weather changes and electricity demand. The system can typically only passively adjust based on real-time PV power generation, energy storage capacity, and electricity consumption, unable to plan power allocation strategies in advance. When weather conditions change abruptly, PV power generation drops sharply, and the system may not be able to adjust the discharge strategy of the energy storage cabinets in time, leading to unstable power supply to the load. Furthermore, current technology lacks forward-looking analysis of electricity demand; during peak electricity demand periods, there may be insufficient PV power generation and energy storage capacity, requiring reliance on grid power. If grid electricity prices are high at this time, it will increase electricity costs and reduce economic efficiency. This presents technical problems such as unreasonable allocation strategies, resulting in wasted power resources or supply-demand imbalances. Summary of the Invention

[0003] This invention addresses the technical problem of unreasonable power allocation strategies in existing technologies, which leads to waste of power resources or supply-demand imbalance. It provides a power optimization allocation method based on artificial intelligence for photovoltaic power generation and energy storage cabinets.

[0004] The technical solution of the present invention to solve the above-mentioned technical problems is as follows:

[0005] This invention provides an artificial intelligence-based method for optimizing power allocation in photovoltaic power generation and energy storage cabinets, comprising: acquiring historical meteorological data sequences of the area where the photovoltaic system is located within a historical time window, performing meteorological forecasting, obtaining future meteorological data sequences within a future time window, and collecting the current operating time, wherein the photovoltaic system includes a photovoltaic power station and an energy storage cabinet; based on the future meteorological data sequences and the operating time, performing photovoltaic power generation forecasting and electricity demand forecasting, obtaining future power generation data sequences and future electricity consumption data sequences; based on the operating time, analyzing the confidence levels of the future meteorological data sequences and future electricity consumption data sequences, obtaining meteorological confidence sequences and electricity consumption confidence sequences; based on the future power generation data sequences, future electricity consumption data sequences, meteorological confidence sequences, and electricity consumption confidence sequences, configuring and optimizing power allocation parameters, obtaining optimized power allocation parameters, and distributing power and storing energy in the energy storage cabinet within the future time window.

[0006] Optionally, the process involves acquiring historical meteorological data sequences for the area where the photovoltaic system is located within a historical time window, performing meteorological forecasting, obtaining future meteorological data sequences for the future time window, and collecting the current operating time. This includes: acquiring historical meteorological data sequences for all historical moments within the historical time window for the area where the photovoltaic system is located, wherein the photovoltaic system includes a photovoltaic power station and an energy storage cabinet; inputting the historical meteorological data sequences into a meteorological forecasting agent, outputting the future meteorological data sequences for the future time window; and collecting the current operating time.

[0007] The training steps of the weather forecasting agent include: collecting a set of historical meteorological sequences based on historical meteorological data records of the area where the photovoltaic system is located, and collecting meteorological data sequences within the time window following each historical meteorological sequence, and labeling them to obtain a set of future meteorological data sequences; constructing a weather forecasting agent based on artificial intelligence; and using the set of historical meteorological sequences and the set of future meteorological data sequences to conduct supervised training and parameter adjustment on the weather forecasting agent, and completing the training after verifying that the accuracy meets the threshold.

[0008] Optionally, based on the future meteorological data sequence and operating time, photovoltaic power generation forecasting and electricity demand forecasting are performed to obtain future power generation data sequences and future electricity consumption data sequences. This includes: inputting the future meteorological data sequence into a photovoltaic power generation classification table and outputting a future photovoltaic power generation data sequence, wherein the photovoltaic power generation classification table includes a mapping relationship between sample meteorological datasets and sample photovoltaic power generation datasets; inputting the future meteorological data sequence and operating time into an electricity consumption forecasting agent and outputting a future electricity consumption data sequence, wherein the electricity consumption forecasting agent is trained using a set of sample meteorological data sequences, a set of sample operating times, and a set of sample electricity consumption data sequences.

[0009] Optionally, based on the operating time, the confidence levels of the future meteorological data sequence and the future electricity consumption data sequence are analyzed to obtain a meteorological confidence sequence and an electricity consumption confidence sequence, including: based on the operating time, obtaining historical average meteorological data and historical average electricity consumption data for the same operating time in historical periods; calculating the similarity between the future meteorological data sequence and the future electricity consumption data sequence and the historical average meteorological data and the historical average electricity consumption data, respectively, to obtain a meteorological confidence sequence and an electricity consumption confidence sequence.

[0010] Optionally, based on the future power generation data sequence, future power consumption data sequence, meteorological confidence sequence, and power consumption confidence sequence, power allocation parameters are configured and optimized to obtain optimized power allocation coefficients. This includes: allocating a meteorological weight sequence based on the future power generation data sequence and the meteorological confidence sequence, and calculating a weighted power allocation space; randomly generating a first power allocation parameter within the power allocation space, wherein the first power allocation parameter includes power allocation parameters for power supply; analyzing a first allocation quality parameter of the first power allocation parameter based on the future power generation data sequence, future power consumption data sequence, meteorological confidence sequence, and power consumption confidence sequence; performing iterative optimization of the power allocation parameters until optimization convergence, obtaining the optimized power allocation parameter with the largest allocation quality parameter, and allocating power supply and energy storage in energy storage cabinets within a future time window.

[0011] Specifically, based on the future power generation data sequence, future power consumption data sequence, meteorological confidence sequence, and power consumption confidence sequence, the analysis of the first allocation quality parameter of the first power allocation parameter includes: allocating and obtaining a power consumption weight sequence based on the power consumption confidence sequence; calculating the similarity between the first power allocation parameter and each future power consumption data in the future power consumption data sequence, and performing a weighted calculation based on the power consumption weight sequence to obtain a first power supply quality coefficient; calculating a first stability quality coefficient based on the first power allocation parameter, the future power generation data sequence, and the meteorological weight sequence; and calculating the first allocation quality parameter based on the first power supply quality coefficient and the first stability quality coefficient.

[0012] Specifically, the calculation of the first stability quality coefficient based on the first power allocation parameter, the future power generation data sequence, and the meteorological weight sequence includes: calculating the ratio of each future power generation data in the future power generation data sequence to the first power allocation parameter, and performing a weighted calculation using the meteorological weight sequence to obtain the first stability quality coefficient.

[0013] By implementing this invention, it is possible to acquire historical meteorological data sequences of the area where the photovoltaic system is located within a historical time window, perform meteorological forecasting, obtain future meteorological data sequences within a future time window, and collect the current operating time. The photovoltaic system includes a photovoltaic power station and an energy storage cabinet. Historical meteorological data is a crucial basis for predicting future weather, and the resulting future meteorological data sequences provide reliable input for subsequent processes. Collecting the current operating time synchronizes the entire forecasting and allocation process with actual time, ensuring the timeliness and accuracy of subsequent analysis and laying the foundation for precise power allocation.

[0014] By implementing this invention, it is possible to predict photovoltaic power generation and electricity demand based on the future meteorological data sequence and operating time, thereby obtaining future power generation data sequence and future electricity consumption data sequence. The future power generation data sequence allows the system to know the potential of photovoltaic power generation in advance, while the future electricity consumption data sequence can grasp the changing trend of load electricity demand. The combination of the two enables the system to have a forward-looking understanding of the power supply and demand situation, avoids unreasonable power allocation due to information lag, and provides key data support for optimizing power allocation.

[0015] By implementing this invention, it is possible to analyze the confidence levels of future meteorological data sequences and future electricity consumption data sequences based on the operating time, thereby obtaining meteorological confidence sequences and electricity consumption confidence sequences. The confidence sequences can reflect the reliability of future meteorological and electricity consumption data. High-confidence data can serve as an important basis for power allocation, while low-confidence data reminds the system to refer to them with caution, thereby improving the rationality of power allocation parameter configuration and reducing allocation risks caused by inaccurate data.

[0016] By implementing this invention, it is possible to configure and optimize power allocation parameters based on the future power generation data sequence, future power consumption data sequence, meteorological confidence sequence, and power consumption confidence sequence, thereby obtaining optimized power allocation parameters. Power distribution and energy storage in energy storage cabinets can then be performed within future time windows. By comprehensively optimizing parameters through multi-faceted data, power allocation becomes more scientific and efficient. While meeting the load's power demand, it also enables the rational charging and discharging of energy storage cabinets, effectively interacting with the smart grid, maximizing economic benefits and power resource utilization, and optimizing the overall power allocation pattern.

[0017] In summary, by implementing this invention, it is possible to dynamically adjust the charging and discharging strategy of the energy storage cabinet according to changes in meteorological conditions and electricity demand, and to interact bidirectionally with the smart grid. This maximizes economic benefits while ensuring the electricity demand of the load, significantly improving the stability and economy of the power system. Attached Figure Description

[0018] Figure 1 A flowchart illustrating a power optimization allocation method for a photovoltaic power generation and energy storage cabinet based on artificial intelligence, provided by the present invention.

[0019] Figure 2 This invention provides a training method for a weather forecasting agent. Detailed Implementation

[0020] 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.

[0021] In the description of this invention, the terms "first" and "second" are used for descriptive purposes only and should not be construed as indicating or implying relative importance or implicitly specifying the number of indicated technical features. Thus, a feature defined as "first" or "second" may explicitly or implicitly include one or more features. In the description of this invention, "a plurality of" means two or more, unless otherwise explicitly specified.

[0022] In the description of this invention, the term "for example" is used to mean "used as an example, illustration, or description." Any embodiment described as "for example" in this invention is not necessarily to be construed as being more preferred or advantageous than other embodiments. The following description is provided to enable any person skilled in the art to make and use the invention. Details are set forth in the following description for purposes of explanation. It should be understood that those skilled in the art will recognize that the invention can be made without using these specific details. In other instances, well-known structures and processes will not be described in detail to avoid obscuring the description of the invention with unnecessary detail. Therefore, the invention is not intended to be limited to the embodiments shown, but is consistent with the broadest scope of the principles and features disclosed herein.

[0023] Example 1, as Figure 1 As shown, this embodiment of the invention provides a power optimization allocation method for photovoltaic power generation and energy storage cabinets based on artificial intelligence, including:

[0024] S100: Obtain historical meteorological data sequences of the area where the photovoltaic system is located within a historical time window, perform meteorological forecasting, obtain future meteorological data sequences within a future time window, and collect the current operating time. The photovoltaic system includes a photovoltaic power station and an energy storage cabinet.

[0025] S200: Based on the future meteorological data sequence and operating time, perform photovoltaic power generation forecasting and electricity demand forecasting to obtain future power generation data sequence and future electricity consumption data sequence;

[0026] S300: Based on the operating time, analyze the confidence levels of the future meteorological data sequence and the future electricity consumption data sequence to obtain the meteorological confidence sequence and the electricity consumption confidence sequence;

[0027] S400: Based on the future power generation data sequence, future power consumption data sequence, meteorological confidence sequence, and power consumption confidence sequence, configure and optimize power distribution parameters to obtain optimized power distribution parameters, and distribute power supply and energy storage in energy storage cabinets within the future time window.

[0028] In step S100 of this application embodiment, historical meteorological data sequences of the area where the photovoltaic system is located within a historical time window are obtained, meteorological forecasts are performed, future meteorological data sequences within a future time window are obtained, and the current operating time is collected, including:

[0029] The system acquires historical meteorological data sequences for all historical moments within a historical time window for the area where the photovoltaic system is located, wherein the photovoltaic system includes a photovoltaic power station and an energy storage cabinet.

[0030] The historical meteorological data sequence is input into the meteorological prediction agent, and the future meteorological data sequence within the future time window is output.

[0031] Collect the current running time.

[0032] In this embodiment of the application, the purpose of step S100 is to provide basic data support for the entire power optimization allocation method. By acquiring historical meteorological data sequences, it provides training and input basis for subsequent meteorological forecasts, ensuring the accuracy of the forecasts. Furthermore, it uses a meteorological forecasting agent to generate future meteorological data sequences, providing key meteorological inputs for photovoltaic power generation forecasts and electricity demand forecasts.

[0033] First, it is necessary to obtain the historical meteorological data sequence for the entire historical time window of the area where the photovoltaic system is located. This involves collecting meteorological data recorded at all historical moments within the historical time window for the area containing both the photovoltaic power station and energy storage cabinets, forming a complete historical meteorological data sequence. This meteorological data forms the basis for weather forecasting and includes key meteorological information affecting photovoltaic power generation, such as past temperature, solar radiation intensity, and wind speed.

[0034] Furthermore, the aforementioned historical meteorological data sequence is input into a pre-trained "meteorological prediction agent," which is an artificial intelligence-based model. Through learning and analyzing historical data, the meteorological prediction agent can predict and output future meteorological data sequences within future time windows.

[0035] Finally, it is necessary to collect the current operating time, that is, record the specific operating time of the photovoltaic system. This time will serve as the time reference for subsequent analysis of future meteorological data, electricity consumption confidence, and optimization of power distribution parameters, ensuring that all forecasting and distribution strategies match the actual operating time.

[0036] like Figure 2As shown, in step S100 of this embodiment, the training step of the weather forecasting agent includes:

[0037] Based on historical meteorological data records of the area where the photovoltaic system is located, a set of historical meteorological sequences of samples is collected, and meteorological data sequences within the time window following each historical meteorological sequence are collected. The future meteorological data sequence set of the samples is then labeled and obtained.

[0038] Construct an intelligent weather forecasting agent based on artificial intelligence;

[0039] The meteorological prediction agent is trained and its parameters are adjusted in a supervised manner using the set of historical meteorological sequences and the set of future meteorological data sequences. The training is completed after the accuracy meets the threshold.

[0040] In this embodiment, the purpose of training the weather forecasting agent is to construct an artificial intelligence model capable of accurately predicting future weather data. Through training, the weather forecasting agent acquires the ability to infer future weather changes based on historical weather data, providing reliable meteorological data support for subsequent photovoltaic power generation forecasting, electricity demand forecasting, and power allocation optimization, thereby ensuring the accuracy and effectiveness of the entire power optimization and allocation method.

[0041] First, it is necessary to collect and label sample data. Specifically, it is necessary to extract the "sample historical meteorological sequence set" from the meteorological data records of the area where the photovoltaic system is located. This means the meteorological data sequence within different time windows in the past, such as continuous data on light intensity, temperature, wind speed, etc. for a certain number of days.

[0042] Then, for each sample's historical meteorological sequence, the actual meteorological data sequence within its corresponding subsequent time window is collected as the sample's future meteorological data sequence set, and the actual meteorological data within the corresponding subsequent time window is labeled. For example, if the sample's historical meteorological sequence is "meteorological data from July 1st to July 3rd, 2023", then the sample's future meteorological data sequence is "actual meteorological data from July 4th to July 6th, 2023", and the correspondence between the two is labeled.

[0043] Considering the applications of weather forecasting agents, a Long Short-Term Memory (LSTM) network based on a Recurrent Neural Network (RNN) can be used to build one. LSTM can effectively handle long-term dependencies in time series data and is suitable for forecasting weather data with time-series characteristics.

[0044] The meteorological forecasting agent is constructed using a three-layer structure: an input layer, a hidden layer, and an output layer. The input layer's dimension matches the number of features in the historical meteorological data sequence, including meteorological characteristics such as sunlight intensity, temperature, humidity, and wind speed in the area where the photovoltaic system is located. The specific dimension is determined based on the types of features in the actual collected historical meteorological data sequence. The hidden layer consists of three LSTM layers: 128 neurons in the first layer, 64 in the second, and 32 in the third; each layer uses the tanh activation function to handle nonlinear relationships. The output layer's dimension matches the number of features in the future meteorological data sequence, employing a linear activation function to output the meteorological data sequence within the future time window.

[0045] For the parameter settings of the weather forecasting agent, the batch size is 32, meaning that 32 historical weather sequences are input for each training iteration. The learning rate is 0.001 to control the magnitude of model parameter updates and avoid convergence that is too fast or too slow. The time step is consistent with the length of the historical time window; for example, if the historical time window is 7 days, the time step is set to 7 to ensure that the model can capture the weather change patterns within the corresponding time range. The regularization coefficient is 0.0001, using L2 regularization to prevent overfitting. The dropout rate is 0.2, randomly discarding 20% ​​of neurons during training to enhance the model's generalization ability.

[0046] For training the weather forecasting agent, the training data consists of the aforementioned historical weather sequences. Each historical weather sequence corresponds to a future weather data sequence, which must include at least three years of historical weather data, totaling no fewer than 10,000 data samples. These data samples are divided into training and validation sets in a 7:3 ratio. The initial training rounds are set to 100. If the validation accuracy does not show a significant improvement within 10 consecutive training rounds, training is stopped early.

[0047] The convergence criterion is set as follows: the weather forecasting agent achieves a prediction accuracy of over 90% on the validation set, meaning the error between the predicted future weather data sequence and the actual weather data sequence is within 10%. When the weather forecasting agent's accuracy on the validation set consistently reaches this threshold for three consecutive rounds, the model training is considered converged, and training is stopped.

[0048] In step S200 of this application embodiment, photovoltaic power generation forecasting and electricity demand forecasting are performed based on the future meteorological data sequence and operating time to obtain future power generation data sequence and future electricity consumption data sequence, including:

[0049] The future meteorological data sequence is input into the photovoltaic power generation classification table, and the future photovoltaic power generation data sequence is output. The photovoltaic power generation classification table includes the mapping relationship between the sample meteorological dataset and the sample photovoltaic power generation dataset.

[0050] The future meteorological data sequence and operating time are input into the electricity consumption prediction agent, and the future electricity consumption data sequence is output. The electricity consumption prediction agent is trained using a sample meteorological data sequence set, a sample operating time set, and a sample electricity consumption data sequence set.

[0051] In this embodiment, the purpose of step S200 is to obtain the power generation capacity and electricity demand within a future time window based on future meteorological data sequences and the current operating time, providing key data support for the subsequent optimization of power allocation parameters. By clarifying the future trends of "power generation" and "electricity consumption," the power allocation strategy can be more forward-looking, avoiding resource waste or insufficient power supply due to unclear supply and demand information, and ultimately achieving efficient utilization of power resources.

[0052] First, the future meteorological data sequence needs to be input into a photovoltaic power generation classification table, and the output will be the future photovoltaic power generation data sequence. This photovoltaic power generation classification table pre-establishes a mapping relationship between sample meteorological datasets and sample photovoltaic power generation datasets, that is, records of photovoltaic power generation under different meteorological conditions, for example, when the irradiance is 800 W / m². 2 At a temperature of 25℃, the corresponding photovoltaic power generation is 500kW. h. Input the future meteorological data sequence obtained in step S100, such as the data on sunshine and temperature for a future period, into the photovoltaic power generation classification table. By matching the existing mapping relationship between meteorology and power generation in the photovoltaic power generation classification table, the corresponding future photovoltaic power generation data sequence is directly output, that is, the predicted power generation at each future moment.

[0053] Furthermore, the future meteorological data sequence and operating time need to be input into the electricity consumption prediction agent, and the future electricity consumption data sequence needs to be output.

[0054] This electricity consumption forecasting agent is a model built on artificial intelligence technology. Its training process uses a set of sample meteorological data sequences, a set of sample operating times, and a set of sample electricity consumption data sequences as training data. The training is completed by learning the correlation between the three.

[0055] The future meteorological data sequence obtained from S100 and the current operating time are input into the trained electricity consumption prediction agent. Based on the learned correlation patterns, the agent outputs the future electricity consumption data sequence.

[0056] Specifically, depending on the intended use of the electricity consumption forecasting agent, an electricity consumption forecasting agent can be built using a model based on gated recurrent units (GRUs). GRUs are suitable for processing time series data and can effectively capture the temporal correlation between weather, operating time and electricity demand.

[0057] The electricity consumption forecasting agent comprises a three-layer structure: an input layer, a hidden layer, and an output layer. The input layer's dimension corresponds to the number of input features, including features from future weather data sequences, such as light intensity and temperature, as well as features related to the current time, such as specific time and weekday / holiday identifiers. The specific dimension is determined based on the types of features collected. The hidden layer consists of two GRU layers: the first layer has 64 neurons, and the second layer has 32 neurons. Each layer uses the tanh activation function to handle non-linear relationships. The output layer has a dimension of 1, corresponding to the numerical values ​​of the future electricity consumption data sequence. It uses a linear activation function to output the electricity consumption data within the future time window.

[0058] For the parameter settings of the electricity consumption prediction agent, the batch size is set to 64, meaning 64 sample data are input for each training iteration. The learning rate is set to 0.0005 to control the parameter update amplitude and ensure training stability.

[0059] The time step is consistent with the length of the historical time window. For example, if the historical time window is 14 days, the time step is set to 14 to capture long-term time series patterns. The regularization coefficient is set to 0.0002, using L2 regularization to prevent model overfitting. The dropout rate is 0.15, randomly dropping 15% of neurons during training to enhance the model's generalization ability.

[0060] The training of the electricity consumption prediction agent requires collecting electricity consumption data covering different weather conditions and operating times, forming no fewer than 8,000 sample data points. Each sample contains a set of sample weather data sequences, the corresponding sample operating time, and the sample electricity consumption data sequences, divided into training and validation sets in a 7:3 ratio. The initial training rounds are set to 80 rounds. If the validation metrics show no improvement for 8 consecutive rounds during training, training will be stopped early to avoid ineffective training.

[0061] The convergence criterion for the electricity consumption prediction agent is set as follows: the model's prediction accuracy on the validation set reaches above 85%, meaning the error between the predicted future electricity consumption data sequence and the actual electricity consumption data sequence is within 15%. Training is considered converged when the model's accuracy on the validation set consistently reaches this criterion for three consecutive rounds.

[0062] In step S300 of this application embodiment, based on the running time, the confidence levels of the future meteorological data sequence and the future electricity consumption data sequence are analyzed to obtain a meteorological confidence sequence and an electricity consumption confidence sequence, including:

[0063] Based on the operating time, obtain historical average meteorological data and historical average electricity consumption data for the same operating time in the past.

[0064] The similarity between the future meteorological data sequence and the future electricity consumption data sequence and the historical average meteorological data and the historical average electricity consumption data are calculated respectively to obtain the meteorological confidence sequence and the electricity consumption confidence sequence.

[0065] In this embodiment, the purpose of step S300 is to assess the reliability of future meteorological data sequences and future electricity consumption data sequences, providing a reliable basis for optimizing subsequent power allocation parameters. By quantifying the confidence level of the data, the system can adjust decision weights based on the reliability of the data when allocating power, avoiding the failure of allocation strategies due to low-reliability data, thereby improving the rationality and stability of power allocation.

[0066] First, based on the stated operating time, it is necessary to obtain historical average meteorological data and historical average electricity consumption data for similar operating times. Specifically, based on the current operating time, such as a specific date, time period, or weekday / holiday time attribute, historical data belonging to the same operating time within the historical time window can be filtered out. For example, if the current time is "Monday at 10:00 AM," then the similar operating times are all historical "Monday at 10:00 AM," and the historical average meteorological data for these similar times, such as average light intensity and average temperature, as well as the historical average electricity consumption data, such as average electricity consumption, can be calculated.

[0067] Furthermore, it is necessary to calculate the similarity between the future meteorological data sequence and the future electricity consumption data sequence and the historical average meteorological data and the historical average electricity consumption data, respectively, to obtain the meteorological confidence sequence and the electricity consumption confidence sequence.

[0068] Specifically, for each data point in the future meteorological data sequence, such as the sunshine and temperature at a certain future moment, the similarity between it and the historical average meteorological data is calculated. The higher the similarity, the more the future meteorological data matches the average situation at the same time in history, and the higher the confidence level. Conversely, the lower the similarity, the lower the confidence level, thus forming a meteorological confidence sequence.

[0069] The similarity can be obtained by calculating Euclidean distance. Assuming the current operating time is "10:00 AM on a summer weekday," the historical average meteorological data for similar times is: light intensity 600 W / m². 2 Temperature 28℃; meteorological data at a certain moment in the future meteorological data series is: light intensity 580W / m². 2 Temperature 29℃.

[0070] First, determine the characteristic values ​​of the two data points: the historical average data is (600, 28), and the future data is (580, 29).

[0071] Then, calculate the Euclidean distance between the eigenvalues ​​of the two data points, where the Euclidean distance = = = ≈20.02

[0072] Finally, the Euclidean distance is converted into similarity: based on the fluctuation range of historical data, the maximum possible distance is set to 100, then the similarity = 1 - (actual Euclidean distance / maximum possible Euclidean distance) ≈ 1 - (20.02 / 100) = 0.7998 ≈ 0.8, that is, the confidence level of the meteorological data at this moment is about 80%.

[0073] Using the above method, the similarity between each data point in the future meteorological data sequence and the historical average meteorological data is calculated, thereby forming a meteorological confidence sequence.

[0074] Similarly, for each data point in the future electricity consumption data sequence, calculate its similarity to the historical average electricity consumption data, determine the electricity consumption confidence level at each moment based on the similarity level, and form an electricity consumption confidence level sequence.

[0075] In step S400 of this application embodiment, based on the future power generation data sequence, future power consumption data sequence, meteorological confidence sequence, and power consumption confidence sequence, power allocation parameters are configured and optimized to obtain optimized power allocation coefficients, including:

[0076] Based on the future power generation data sequence, combined with the meteorological confidence sequence, a meteorological weight sequence is assigned, and a weighted calculation is performed to obtain the power allocation space.

[0077] A first power allocation parameter is randomly generated within the power allocation space, wherein the first power allocation parameter includes power allocation parameters for supplying power;

[0078] Based on the future power generation data sequence, future power consumption data sequence, meteorological confidence sequence, and power consumption confidence sequence, analyze the first allocation quality parameter of the first power allocation parameter;

[0079] The power distribution parameters are iteratively optimized until the optimization converges, and the optimized power distribution parameters with the largest distribution quality parameters are obtained. Power distribution and supply and energy storage in energy storage cabinets are carried out within the future time window.

[0080] In this embodiment, the purpose of obtaining the power allocation space in step S400 is to define a reasonable range for the optimization of power allocation parameters and generate initial power allocation parameters within this range, providing a basis for subsequent iterative optimization. By combining future power generation capacity and the reliability of meteorological data to determine the allocation space, it ensures that the initial parameters are within a feasible range, avoids parameter optimization from falling into an invalid region, and improves the efficiency and accuracy of subsequent optimization.

[0081] First, based on the future power generation data sequence, a meteorological weight sequence needs to be assigned according to the meteorological confidence sequence, and a weighted calculation is performed to obtain the power allocation space.

[0082] First, it is necessary to determine the meteorological weight sequence based on the meteorological confidence sequence. Specifically, this can be done by calculating the ratio of the meteorological confidence at each moment to the sum of the meteorological confidence at all moments. This ratio is the meteorological weight at the corresponding moment. For example, if the meteorological confidence at a certain moment is 0.3 and the sum of the confidence at all moments is 1.0, then the meteorological weight at that moment is 0.3.

[0083] Next, it is necessary to determine the power allocation range for each time moment. The power generation data at each time moment in the future power generation data sequence, such as 100kW, 120kW, etc., is the maximum power that can be allocated at that time moment. Therefore, the power allocation range for a single time moment is from 0 to the future power generation data at that time moment, such as 0-100kW.

[0084] Finally, a weighted calculation of the power allocation interval is needed. Using the meteorological weight sequence as the weight, the maximum value of the power allocation interval at all times, i.e., the future power generation data, is weighted and summed. The result is the power allocation space. For example, if the future power generation data at two times are 100kW and 200kW respectively, corresponding to meteorological weights of 0.3 and 0.7 respectively, then the power allocation space is 100×0.3+200×0.7=170kW, meaning the maximum range of the overall distributable power is 0-170kW.

[0085] Furthermore, within the power allocation space calculated above, initial power allocation parameters, i.e., the first power allocation parameters, need to be randomly generated. These parameters include specific power allocation schemes, such as supplying 50kW to the load and charging the energy storage cabinet with 0kW at one time, while supplying only 100kW to the load at another time. All power allocation parameters are within the range of the power allocation space.

[0086] In step S400 of this application embodiment, based on the future power generation data sequence, future power consumption data sequence, meteorological confidence sequence, and power consumption confidence sequence, the first allocation quality parameter of the first power allocation parameter is analyzed, including:

[0087] Based on the electricity consumption confidence sequence, an electricity consumption weight sequence is obtained;

[0088] Calculate the similarity between the first power allocation parameter and each future power consumption data in the future power consumption data sequence, and perform a weighted calculation based on the power consumption weight sequence to obtain the first power quality coefficient;

[0089] The first stability quality coefficient is calculated based on the first power allocation parameter, the future power generation data sequence, and the meteorological weight sequence.

[0090] The first allocation quality parameter is calculated based on the first power supply quality coefficient and the first stability quality coefficient.

[0091] In this embodiment, the purpose of the above steps is to provide a basis for subsequent parameter iterative optimization by quantitatively evaluating the power supply matching degree and system stability of the initial allocation parameters, i.e., the first power allocation parameters. By combining the confidence levels of electricity consumption and weather, the allocation parameters are comprehensively considered to ensure that they meet the electricity demand and match the power generation capacity, thus ensuring that the final optimized allocation parameters can both reliably supply power and maintain system stability.

[0092] First, an electricity consumption weight sequence needs to be obtained based on the electricity consumption confidence sequence. That is, the ratio of the electricity consumption confidence at each time point to the sum of the electricity consumption confidence at all times is calculated. This ratio is the electricity consumption weight at the corresponding time point. For example, if the electricity consumption confidence at a certain time point is 0.4 and the total confidence is 1.0, then the corresponding electricity consumption weight is 0.4.

[0093] Next, it is necessary to calculate the similarity between the first power allocation parameter and each future power consumption data point in the future power consumption data sequence, and then perform a weighted calculation based on the power consumption weight sequence to obtain the first power quality coefficient. Specifically, it is necessary to calculate the similarity between the first power allocation parameter and the data at each time point in the future power consumption data sequence. The specific calculation method is the same as the meteorological confidence calculation method in S300. The higher the similarity, the higher the matching degree between the power allocation parameter and the power demand.

[0094] Then, the similarity scores are weighted according to the power consumption weight sequence. This involves multiplying the similarity score at each time step by the corresponding power consumption weight, and then summing the results to obtain the first power quality coefficient. For example, if the similarity score is 0.9 and the weight is 0.4 at one time step, and the similarity score is 0.8 and the weight is 0.6 at another time step, then the first power quality coefficient = 0.9 × 0.4 + 0.8 × 0.6 = 0.84. This first power quality coefficient reflects the degree to which the power distribution parameters meet the power demand; a higher value is better.

[0095] In step S400 of this application embodiment, the first stability quality coefficient is calculated based on the first power allocation parameters, the future power generation data sequence, and the meteorological weight sequence, including:

[0096] The ratio of each future power generation data point to the first power allocation parameter within the future power generation data sequence is calculated, and the first stable quality coefficient is obtained by weighting the results using the meteorological weight sequence.

[0097] First, it is necessary to calculate the ratio of the power generation data at each moment in the future power generation data sequence to the first power allocation parameter. For example, if the future power generation at a certain moment is 100kW and the power allocation parameter is 80kW, then the ratio = 100 / 80 = 1.25; if the power allocation parameter is 120kW, then the ratio = 100 / 120 ≈ 0.83. The closer this ratio is to 1, the higher the matching degree between the power allocation parameter and the power generation capacity, and the more stable the system.

[0098] Then, the above ratios are weighted using a meteorological weight sequence: the ratio at each time point is multiplied by the corresponding meteorological weight, and then summed to obtain the first stability quality coefficient. For example, if the ratio is 1.25 and the meteorological weight is 0.3 at one time point, and the ratio is 0.83 and the weight is 0.7 at another time point, then the first stability quality coefficient = 1.25 × 0.3 + 0.83 × 0.7 ≈ 0.96. This first stability quality coefficient reflects the matching stability between power distribution parameters and power generation capacity; the closer the value is to 1, the better.

[0099] Finally, the first allocation quality parameter needs to be calculated based on the first power supply quality coefficient and the first stability quality coefficient. Specifically, the first allocation quality parameter can be obtained by combining the first power supply quality coefficient and the first stability quality coefficient through weighted summation. For example, if each has a 50% weight, and the power supply quality coefficient is 0.84 and the stability quality coefficient is 0.96, then the first allocation quality parameter = 0.84 × 0.5 + 0.96 × 0.5 = 0.9. This parameter comprehensively reflects the overall quality of the first power allocation parameter, providing a quantitative standard for subsequent iterative optimization. The higher the first allocation quality parameter, the better the allocation parameter.

[0100] In step S400 of this application embodiment, after completing the above steps, it is also necessary to perform iterative optimization of the power distribution parameters until the optimization converges and the optimized power distribution parameters with the largest distribution quality parameters are obtained, and power distribution and power supply and energy storage of the energy storage cabinet are carried out in the future time window.

[0101] In this embodiment, iterative optimization of power allocation parameters is the final step in power allocation parameter optimization in S400. The core purpose is to select the optimal power allocation parameters with the largest allocation quality parameters from the initial power allocation parameters through continuous iterative optimization, so as to ensure that the power allocation and energy storage strategies in the future time window can meet the electricity demand to the maximum extent and match the power generation capacity, and achieve overall optimization by combining the confidence of meteorological data and electricity consumption data.

[0102] First, the initial power allocation parameters generated in the above steps are used as the initial solution to enter the iterative optimization process. In each iteration, small adjustments are made to the current power allocation parameters. For example, the power supply / storage allocation values ​​at a certain moment are randomly increased or decreased by a certain margin to generate new candidate power allocation parameters. The adjustment range must be limited to the previously determined power allocation space to ensure that the candidate power allocation parameters meet the constraints of future power generation capacity.

[0103] For each round of iteration, the evaluation step in step S400 above is repeated to calculate the allocation quality parameter of the candidate power allocation parameter.

[0104] After each iteration, the allocation quality parameters of the current candidate power allocation parameters are compared with those of the historical best power allocation parameters: if the allocation quality parameters of the candidate power allocation parameters are higher, the historical best power allocation parameters are updated to the candidate power allocation parameters; if the allocation quality parameters of the candidate power allocation parameters are lower, the historical best power allocation parameters are retained.

[0105] The process continues iterating until a convergence condition is met. The convergence condition can be defined as follows: if the improvement in the quality parameter of the historically optimal parameter allocation is less than a preset threshold across multiple iterations, then the optimization is considered converged. For example, a convergence condition could be defined as the improvement in the quality parameter of the historically optimal parameter allocation being less than 0.1% across 10 consecutive iterations.

[0106] Once the optimization converges, the optimal power allocation parameter that maximizes the final allocation quality parameter will be applied to the future time window. That is, the power allocation supply and energy storage of the energy storage cabinet will be dynamically adjusted according to the parameter.

[0107] It should be noted that the descriptions of each embodiment in the above embodiments have different focuses. For parts that are not described in detail in a certain embodiment, please refer to the relevant descriptions in other embodiments.

[0108] Those skilled in the art will understand that embodiments of the present invention can provide methods, systems, or computer program products. Therefore, the present invention can take the form of a completely hardware embodiment, a completely software embodiment, or an embodiment combining software and hardware aspects. Furthermore, the present invention can take the form of a computer program product embodied on one or more computer-usable storage media (including, but not limited to, disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0109] This invention is described with reference to flowchart illustrations and / or block diagrams of methods, apparatus (systems), and computer program products according to embodiments of the invention. It will be understood that each block of the flowchart illustrations and / or block diagrams, as well as combinations of blocks in the flowchart illustrations and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, special-purpose computer, embedded computer, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, generate instructions for implementing the flowchart. Figure 1 One or more processes and / or boxes Figure 1 A device that provides the functions specified in one or more boxes.

[0110] These computer program instructions may also be stored in a computer-readable storage medium that can direct a computer or other programmable data processing device to function in a particular manner, such that the instructions stored in the computer-readable storage medium produce an article of manufacture including instruction means, which are implemented in a process Figure 1 One or more processes and / or boxes Figure 1 The function specified in one or more boxes.

[0111] These computer program instructions may also be loaded onto a computer or other programmable data processing equipment to cause a series of operational steps to be performed on the computer or other programmable equipment to produce a computer-implemented process, thereby providing instructions that execute on the computer or other programmable equipment for implementing the process. Figure 1 One or more processes and / or boxes Figure 1 The steps of the function specified in one or more boxes.

[0112] Although preferred embodiments of the invention have been described, those skilled in the art, once they have learned the basic inventive concept, can make other changes and modifications to these embodiments.

[0113] Obviously, those skilled in the art can make various modifications and variations to this invention without departing from its spirit and scope. Therefore, if these modifications and variations fall within the scope of this invention and its equivalents, this invention also intends to include these modifications and variations.

Claims

1. A method for power optimization distribution of an artificial intelligence-based photovoltaic power generation and energy storage cabinet, characterized in that, The method includes: The system acquires historical meteorological data sequences for the area where the photovoltaic system is located within a historical time window, performs meteorological forecasting, obtains future meteorological data sequences for future time windows, and collects the current operating time. The photovoltaic system includes a photovoltaic power station and an energy storage cabinet. Based on the future meteorological data sequence and operating time, photovoltaic power generation forecast and electricity demand forecast are performed to obtain future power generation data sequence and future electricity consumption data sequence; Based on the stated operating time, the confidence levels of the future meteorological data sequence and the future electricity consumption data sequence are analyzed to obtain the meteorological confidence sequence and the electricity consumption confidence sequence. Based on the future power generation data sequence, future power consumption data sequence, meteorological confidence sequence, and power consumption confidence sequence, power allocation parameters are configured and optimized to obtain optimized power allocation parameters, and power allocation and supply and energy storage in energy storage cabinets are carried out within the future time window; Specifically, based on the future power generation data sequence, future power consumption data sequence, meteorological confidence sequence, and power consumption confidence sequence, power allocation parameters are configured and optimized to obtain optimized power allocation coefficients, including: Based on the future power generation data sequence, combined with the meteorological confidence sequence, a meteorological weight sequence is assigned, and a weighted calculation is performed to obtain the power allocation space. A first power allocation parameter is randomly generated within the power allocation space, wherein the first power allocation parameter includes power allocation parameters for supplying power; Based on the future power generation data sequence, future power consumption data sequence, meteorological confidence sequence, and power consumption confidence sequence, analyze the first allocation quality parameter of the first power allocation parameter; The power distribution parameters are iteratively optimized until the optimization converges, and the optimized power distribution parameters with the largest distribution quality parameters are obtained. Power distribution and supply and energy storage in energy storage cabinets are carried out in the future time window. Specifically, based on the future power generation data sequence, future power consumption data sequence, meteorological confidence sequence, and power consumption confidence sequence, the first allocation quality parameter of the first power allocation parameter is analyzed, including: Based on the electricity consumption confidence sequence, an electricity consumption weight sequence is obtained; Calculate the similarity between the first power allocation parameter and each future power consumption data in the future power consumption data sequence, and perform a weighted calculation based on the power consumption weight sequence to obtain the first power quality coefficient; The first stability quality coefficient is calculated based on the first power allocation parameter, the future power generation data sequence, and the meteorological weight sequence. The first allocation quality parameter is calculated based on the first power supply quality coefficient and the first stability quality coefficient. The first stability quality coefficient is calculated based on the first power allocation parameters, the future power generation data sequence, and the meteorological weight sequence, including: The ratio of each future power generation data point to the first power allocation parameter within the future power generation data sequence is calculated, and the first stable quality coefficient is obtained by weighting the results using the meteorological weight sequence.

2. The power optimization allocation method for photovoltaic power generation and energy storage cabinets based on artificial intelligence according to claim 1, characterized in that, Acquire historical meteorological data sequences for the area where the photovoltaic system is located within a historical time window, perform meteorological forecasting, obtain future meteorological data sequences for future time windows, and collect current operating data, including: The system acquires historical meteorological data sequences for all historical moments within a historical time window for the area where the photovoltaic system is located, wherein the photovoltaic system includes a photovoltaic power station and an energy storage cabinet. The historical meteorological data sequence is input into the meteorological prediction agent, and the future meteorological data sequence within the future time window is output. Collect the current runtime data.

3. The power optimization allocation method for photovoltaic power generation and energy storage cabinets based on artificial intelligence according to claim 2, characterized in that, The training steps for the weather forecasting agent include: Based on historical meteorological data records of the area where the photovoltaic system is located, a set of historical meteorological sequences of samples is collected, and meteorological data sequences within the time window following each historical meteorological sequence are collected. The future meteorological data sequence set of the samples is then labeled and obtained. Construct an intelligent weather forecasting agent based on artificial intelligence; The meteorological prediction agent is trained and its parameters are adjusted in a supervised manner using the set of historical meteorological sequences and the set of future meteorological data sequences. The training is completed after the accuracy meets the threshold.

4. The power optimization allocation method for photovoltaic power generation and energy storage cabinets based on artificial intelligence according to claim 1, characterized in that, Based on the aforementioned future meteorological data sequence and operating time, photovoltaic power generation forecasting and electricity demand forecasting are performed to obtain future power generation data sequences and future electricity consumption data sequences, including: The future meteorological data sequence is input into the photovoltaic power generation classification table, and the future photovoltaic power generation data sequence is output. The photovoltaic power generation classification table includes the mapping relationship between the sample meteorological dataset and the sample photovoltaic power generation dataset. The future meteorological data sequence and operating time are input into the electricity consumption prediction agent, and the future electricity consumption data sequence is output. The electricity consumption prediction agent is trained using a set of sample meteorological data sequences, a set of sample operating times, and a set of sample electricity consumption data sequences.

5. The power optimization allocation method for photovoltaic power generation and energy storage cabinets based on artificial intelligence according to claim 1, characterized in that, Based on the stated operating time, the confidence levels of the future meteorological data sequence and the future electricity consumption data sequence are analyzed to obtain meteorological confidence sequences and electricity consumption confidence sequences, including: Based on the operating time, obtain historical average meteorological data and historical average electricity consumption data for the same operating time in the past. The similarity between the future meteorological data sequence and the future electricity consumption data sequence and the historical average meteorological data and the historical average electricity consumption data are calculated respectively to obtain the meteorological confidence sequence and the electricity consumption confidence sequence.

Images