A method for frequency modulation, peak regulation and power distribution of an intelligent micro-grid based on photovoltaic power generation

Abstract

The application relates to a kind of intelligent micro-grid frequency modulation peak shaving and power distribution methods based on photovoltaic power generation, and relates to the field of intelligent micro-grid, comprising: obtaining predicted fluctuation degree and energy supply deviation amplitude by analysis, obtaining fluctuation degree by combining predicted fluctuation degree and historical fluctuation degree, obtaining frequency modulation coefficient and peak shaving distribution coefficient according to fluctuation degree and energy supply deviation amplitude configuration; randomly generate power distribution scheme, according to predicted photovoltaic power generation power sequence and predicted demand load power sequence, carry out peak shaving and frequency modulation analysis, calculate distribution fitness according to frequency modulation coefficient and peak shaving distribution coefficient, carry out iterative optimization evaluation, obtain optimal distribution scheme to carry out power distribution. Through the application, the technical problems that traditional methods cannot efficiently cope with load fluctuation and photovoltaic power generation deficiency, and it is difficult to ensure that energy storage units can respond in time and provide stable and accurate power support, resulting in that micro-grid has large operation fluctuation and high frequency modulation and peak shaving difficulty during high load period can be solved.

Description

Technical Field

[0001] This invention relates to the field of smart microgrids, and more particularly to a method for frequency regulation, peak shaving, and power distribution in smart microgrids based on photovoltaic power generation. Background Technology

[0002] With the widespread application of renewable energy, photovoltaic (PV) power generation has become the main energy source in microgrids. However, the volatility and intermittency of PV power generation make it impossible to stably meet load demand at all times, especially during periods of insufficient sunshine or peak electricity consumption. To compensate for the shortcomings of PV power generation, energy storage systems have been introduced into microgrids to store excess electricity and provide support when demand increases.

[0003] Currently, energy storage units in smart microgrids mainly rely on traditional power distribution methods to regulate the charging and discharging process of the energy storage system. These methods are usually based on preset rules or simple load forecasts, which cannot fully cope with the complex situation of load fluctuations and insufficient photovoltaic power generation. During periods of low photovoltaic power generation or high demand load, traditional methods often fail to achieve accurate power distribution, resulting in energy storage units being unable to provide sufficient power support in a timely and effective manner, which in turn causes significant fluctuations in the operation of the microgrid.

[0004] In addition, existing power distribution methods do not adequately support the frequency regulation and peak shaving capabilities of energy storage units, making it difficult to meet the requirements of rapid response to load fluctuations and ensuring grid stability. This situation is particularly prominent during high-load periods of microgrids, which may lead to increased grid frequency fluctuations, affect power supply quality, and even cause power outages. Summary of the Invention

[0005] This invention addresses the technical problems of traditional methods being unable to efficiently cope with load fluctuations and insufficient photovoltaic power generation, making it difficult to ensure that energy storage units can respond in a timely manner and provide stable and accurate power support, resulting in large fluctuations in the operation of microgrids during high-load periods and high difficulty in frequency regulation and peak shaving. It provides a smart microgrid frequency regulation, peak shaving and power distribution method based on photovoltaic power generation to solve these problems.

[0006] The technical solution of this invention to solve the above-mentioned technical problems is as follows: This invention provides a method for frequency regulation, peak shaving, and power distribution in a smart microgrid based on photovoltaic power generation. The method is applied to a microgrid, which includes photovoltaic generator sets and energy storage units. The method includes: acquiring historical photovoltaic power generation sequences and historical demand load sequences within the microgrid; analyzing and obtaining historical volatility; predicting photovoltaic power generation and demand load for future periods based on the historical volatility; and obtaining predicted photovoltaic power generation sequences and predicted demand load sequences. Based on the historical photovoltaic power generation sequences, historical demand load sequences, predicted photovoltaic power generation sequences, and predicted demand load sequences, the method analyzes and obtains the predicted volatility and the power supply deviation amplitude. The system calculates the volatility by combining the predicted volatility and historical volatility. Based on the volatility and the power supply deviation, it configures the frequency regulation coefficient and peak shaving allocation coefficient. A power allocation scheme for the energy storage unit is randomly generated. Peak shaving and frequency regulation analyses are performed based on the predicted photovoltaic power generation sequence and the predicted demand load sequence. The allocation fitness is calculated according to the frequency regulation coefficient and peak shaving allocation coefficient. The power allocation scheme includes the allocation power sequence for future periods, and the allocation fitness is calculated based on the deviation between adjacent allocation powers. Based on the allocation fitness, the power allocation scheme is iteratively optimized and evaluated to obtain the optimal power allocation scheme. The system then controls the energy storage unit to allocate power within the microgrid.

[0007] Optionally, the method for frequency regulation, peak shaving, and power distribution in a smart microgrid based on photovoltaic power generation further includes: obtaining the photovoltaic power generation of the photovoltaic generator sets in the microgrid over the past K historical moments, and the demand load power on the electricity consumption side, to obtain historical photovoltaic power generation sequences and historical demand load power sequences, where K is a positive integer; analyzing the volatility of the historical photovoltaic power generation sequences and historical demand load power sequences to obtain historical photovoltaic volatility and historical demand volatility, and calculating the historical volatility; and using ensemble machine learning to construct a photovoltaic power generation prediction branch array and a demand load prediction branch array, wherein the photovoltaic power generation prediction branch array and the demand load prediction branch array each include P photovoltaic power generation prediction branches and... There are P branches for demand load forecasting, where P is a positive integer. Based on the historical volatility and P, Q is calculated, where Q is a positive integer. Q photovoltaic power generation forecasting branches and Q demand load forecasting branches are randomly selected. Photovoltaic power generation and demand load are forecasted for K future times within a future period based on the historical photovoltaic power generation sequence and historical demand load sequence, resulting in Q branch-predicted photovoltaic power generation sequences and Q branch-predicted demand load power sequences. The average values ​​of the Q branch-predicted photovoltaic power generation and Q branch-predicted demand load power at each future time within the Q branch-predicted photovoltaic power generation sequence and Q branch-predicted demand load power sequence are calculated to obtain the predicted photovoltaic power generation sequence and the predicted demand load power sequence.

[0008] Optionally, the method for frequency regulation, peak shaving, and power distribution in a smart microgrid based on photovoltaic power generation further includes: selecting a first historical photovoltaic power generation at a first historical moment within the historical photovoltaic power generation sequence, and randomly selecting multiple random historical photovoltaic power generation; calculating the average of the multiple random historical photovoltaic power generation and the fluctuation amplitude of the first historical photovoltaic power generation to obtain a first photovoltaic volatility; continuing to calculate to obtain K photovoltaic volatility values, and calculating the average to obtain historical photovoltaic volatility; calculating historical demand volatility based on the historical demand load power sequence; and calculating the average of the historical photovoltaic volatility and historical demand volatility to obtain historical volatility.

[0009] Optionally, the method for frequency regulation, peak shaving, and power distribution in a smart microgrid based on photovoltaic power generation further includes: collecting a set of sample historical photovoltaic power generation sequences based on microgrid operation logs over a historical period, and using the photovoltaic power generation sequence following each sample historical photovoltaic power generation sequence as a sample predicted photovoltaic power generation sequence to obtain a set of sample predicted photovoltaic power generation sequences, and integrating them to obtain a photovoltaic prediction sample dataset; randomly selecting P photovoltaic prediction sample data with replacement from the photovoltaic prediction sample dataset, and training P photovoltaic power generation prediction branches using ensemble machine learning; collecting a set of sample historical demand load power sequences based on microgrid operation logs over a historical period, and using the demand load power sequence following each sample historical demand load power sequence as a sample predicted demand load power sequence to obtain a set of sample predicted demand load power sequences, and integrating them to obtain a load prediction sample dataset; randomly selecting P load prediction sample data with replacement from the load prediction sample dataset, and training P load power generation prediction branches using ensemble machine learning; and combining the P photovoltaic power generation prediction branches and the P load power generation prediction branches to obtain a photovoltaic power generation prediction branch array and a demand load prediction branch array.

[0010] Optionally, the method for frequency regulation, peak shaving, and power distribution in a smart microgrid based on photovoltaic power generation further includes: analyzing and calculating the predicted volatility based on the predicted photovoltaic power generation sequence and the predicted demand load power sequence; calculating the average of the predicted volatility and historical volatility to obtain the volatility; calculating the power deviation amplitude of each demand load power and the corresponding photovoltaic power based on the historical photovoltaic power generation sequence, the historical demand load power sequence, the predicted photovoltaic power generation sequence, and the predicted demand load power sequence, and calculating the average to obtain the power supply deviation amplitude; and configuring the volatility and power supply deviation amplitude as a frequency regulation coefficient and a peak shaving distribution coefficient.

[0011] Optionally, the method for frequency regulation, peak shaving, and power distribution in a smart microgrid based on photovoltaic power generation further includes: obtaining the power distribution space for the energy storage unit to distribute power to the power consumption side; randomly generating K first power distribution powers within the power distribution space as first power distribution sequences for K future times in the future period to obtain a first power distribution scheme; calculating a first actual power supply sequence based on the first power distribution sequence and the predicted photovoltaic power generation sequence; subtracting the corresponding predicted demand load power in the predicted demand load power sequence from each first actual power supply in the first actual power supply sequence to obtain a first redundant power supply sequence; and calculating a first distribution fitness based on the first redundant power supply sequence and the first power distribution sequence, according to the frequency regulation coefficient and the peak shaving distribution coefficient, wherein the distribution fitness is calculated based on the deviation between adjacent power distributions within the first power distribution sequence.

[0012] Optionally, the method for frequency regulation, peak shaving, and power distribution in a smart microgrid based on photovoltaic power generation further includes: calculating a first distribution fitness based on the first redundant power supply sequence and the first distribution power sequence, according to the frequency regulation coefficient and the peak shaving distribution coefficient, as shown in the following formula: Among them, F f To assign fitness, w1, w2, and w3 are weights, and their sum is 1. N is a positive integer, TP is the frequency modulation coefficient, TF is the peak shaving allocation coefficient, K is the number of redundant power supplies within the redundant power supply sequence, σ is the variance of the K redundant power supplies within the redundant power supply sequence, and G... i M represents the redundant power supply at the i-th future time. i For the power allocation at the i-th future time, M i+1 The power allocated at the (i+1)th future time.

[0013] Optionally, the method for frequency regulation, peak shaving, and power distribution of a smart microgrid based on photovoltaic power generation further includes: continuing to randomly generate power distribution schemes and calculating the distribution fitness and iterative optimization; until the iterative optimization converges, retaining the power distribution scheme with the largest distribution fitness as the optimal power distribution scheme, and controlling the energy storage unit to distribute power supply in the microgrid at K future moments in the future time period.

[0014] The beneficial effects of this invention are as follows: By acquiring historical photovoltaic power generation sequences and historical demand load sequences within a microgrid, historical volatility is analyzed and obtained. Based on the historical volatility, photovoltaic power generation and demand load are predicted for future periods, resulting in predicted photovoltaic power generation sequences and predicted demand load sequences. Then, based on the historical photovoltaic power generation sequences, historical demand load sequences, predicted photovoltaic power generation sequences, and predicted demand load sequences, predicted volatility and power supply deviation amplitude are analyzed and obtained. Combining the predicted volatility and historical volatility, volatility is calculated. Based on the volatility and power supply deviation amplitude, frequency regulation coefficients and peak shaving allocation coefficients are configured. Finally, a power allocation scheme for the energy storage unit is randomly generated, and based on the predicted photovoltaic power generation sequences and predicted demand load sequences... Peak shaving and frequency regulation analysis is performed. Based on the frequency regulation coefficient and peak shaving allocation coefficient, the allocation fitness is calculated. The power allocation scheme includes a power allocation sequence for future time periods, and the allocation fitness is calculated based on the deviation between adjacent power allocations. Further, based on the allocation fitness, the power allocation scheme is iteratively optimized and evaluated to obtain the optimal power allocation scheme. Finally, the energy storage unit is controlled to allocate power within the microgrid according to the optimal power allocation scheme. In other words, by optimizing the power allocation of the energy storage unit, demand load can be accurately matched during peak electricity consumption periods, improving the scientific nature and accuracy of power allocation, effectively mitigating the negative impact of photovoltaic power generation fluctuations, and ensuring that the energy storage unit can quickly respond to load fluctuations, achieving efficient frequency and peak regulation, thereby significantly improving the overall operational stability and reliability of the microgrid. Attached Figure Description

[0015] Figure 1 A flowchart illustrating a method for frequency regulation, peak shaving, and power distribution in a smart microgrid based on photovoltaic power generation, provided by this invention.

[0016] Figure 2 This invention provides a schematic diagram of the process for obtaining the predicted photovoltaic power generation sequence and the predicted demand load power sequence in a smart microgrid frequency regulation, peak shaving, and power distribution method based on photovoltaic power generation. Detailed Implementation

[0017] 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.

[0018] 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 of the stated features. In the description of this invention, "a plurality of" means two or more, unless otherwise explicitly specified.

[0019] 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.

[0020] Examples, such as Figure 1 As shown in the figure, this invention provides a method for frequency regulation, peak shaving, and power distribution in a smart microgrid based on photovoltaic power generation. The method is applied to a microgrid, which includes photovoltaic generator sets and energy storage units, and specifically includes the following steps:

[0021] S1: Obtain the historical photovoltaic power generation sequence and historical demand load sequence within the microgrid, analyze to obtain the historical volatility, and predict the photovoltaic power generation and demand load in the future period according to the historical volatility to obtain the predicted photovoltaic power generation sequence and the predicted demand load sequence.

[0022] Furthermore, such as Figure 2 As shown, step S1 of the present invention further includes:

[0023] The photovoltaic power generation of the photovoltaic generator set in the microgrid and the demand load power on the electricity consumption side are obtained at the past K historical moments to obtain the historical photovoltaic power generation power sequence and the historical demand load power sequence, where K is a positive integer.

[0024] Specifically, by querying the microgrid operation log, the photovoltaic power generation of the photovoltaic generators in the microgrid at the past K historical moments is obtained. The photovoltaic power generation usually fluctuates with weather changes (such as sunlight intensity) and time (such as the difference between day and night). Therefore, this data is crucial for analyzing the stability of the photovoltaic power generation system and effectively predicting future power generation. K is a positive integer, and the specific value of K can be set according to the actual scenario. For example, 20 means selecting the photovoltaic power generation at the past K (20) historical moments (monitoring time points, such as data monitoring every 2 minutes) and arranging the K photovoltaic power generation in chronological order to obtain the historical photovoltaic power generation sequence.

[0025] On the other hand, the demand load power of the electricity consumption side at the past K historical moments is obtained. The demand load refers to the electricity required by the electricity consumption side of the microgrid at the same time. It is usually affected by weather, time and electricity consumption behavior. The collection time points of the demand load power and photovoltaic power generation are consistent. Then, the K demand load power are arranged in chronological order to obtain the historical demand load power sequence.

[0026] Analyze the volatility of the historical photovoltaic power generation power series and the historical demand load power series to obtain the historical photovoltaic volatility and historical demand volatility, and calculate the historical volatility.

[0027] Furthermore, the present invention also includes the following steps:

[0028] Within the historical photovoltaic power generation sequence, select the first historical photovoltaic power generation at a first historical moment, and randomly select multiple random historical photovoltaic power generation values; calculate the average of the multiple random historical photovoltaic power generation values ​​and the fluctuation amplitude of the first historical photovoltaic power generation value to obtain the first photovoltaic volatility; continue to calculate to obtain K photovoltaic volatility values, calculate the average to obtain the historical photovoltaic volatility; calculate the historical demand volatility based on the historical demand load power sequence; calculate the average of the historical photovoltaic volatility and the historical demand volatility to obtain the historical volatility.

[0029] Specifically, firstly, the first historical photovoltaic power generation at a first historical moment (any one of the K historical moments) is selected from the historical photovoltaic power generation sequence. Then, a predetermined number of random historical photovoltaic power generation values ​​are randomly selected from the historical photovoltaic power generation sequence. The predetermined number can be set according to the amount of data. For example, the predetermined number can be set to 20%, that is, 20% of the historical photovoltaic power generation values ​​are randomly selected from the historical photovoltaic power generation sequence and set as random historical photovoltaic power generation values. For example, if the number of historical photovoltaic power generation values ​​in the historical photovoltaic power generation sequence is 20, then the number of random historical photovoltaic power generation values ​​is 4.

[0030] Next, the average value of the multiple random historical photovoltaic power generation is calculated to obtain the average random historical photovoltaic power generation. Further, the fluctuation amplitude between the average random historical photovoltaic power generation and the first historical photovoltaic power generation is calculated, where the fluctuation amplitude is the ratio of the absolute value of the power difference between the average random historical photovoltaic power generation and the first historical photovoltaic power generation to the average random historical photovoltaic power generation. This fluctuation amplitude is set as the first photovoltaic volatility, which characterizes the fluctuation range of the first historical photovoltaic power generation compared to other historical photovoltaic power generation. Then, using the same method, the fluctuation amplitudes of other historical photovoltaic power generation within the historical photovoltaic power generation sequence are calculated to obtain K photovoltaic volatility values. The average value of these K photovoltaic volatility values ​​is then calculated, and the result is set as the historical photovoltaic volatility. The larger the historical photovoltaic volatility, the stronger the uncertainty of photovoltaic power generation at the K historical moments.

[0031] On the other hand, using the same algorithm for calculating the historical photovoltaic (PV) volatility, K demand volatility values ​​are calculated based on the historical demand load power sequence. The historical demand volatility is then calculated by averaging these K values. This historical demand volatility reflects the fluctuation of the microgrid's demand-side power at K historical points in time. Finally, the historical PV volatility and historical demand volatility are averaged, and the result is set as the historical volatility. The historical volatility integrates the fluctuation characteristics of PV power generation and demand load, reflecting the overall volatility level of the microgrid. By calculating the historical volatility, the overall volatility state of the microgrid can be obtained, thus providing an important basis for optimizing power distribution and frequency and peak regulation.

[0032] An integrated machine learning approach is used to construct a photovoltaic power generation forecasting branch array and a demand load forecasting branch array, wherein the photovoltaic power generation forecasting branch array and the demand load forecasting branch array each include P photovoltaic power generation forecasting branches and P demand load forecasting branches, respectively, where P is a positive integer.

[0033] Furthermore, the present invention also includes the following steps:

[0034] Based on the microgrid operation logs over a historical period, a set of historical photovoltaic (PV) power generation sequences is collected. The PV power generation sequence following each historical PV power generation sequence is used as a sample predicted PV power generation sequence, resulting in a set of sample predicted PV power generation sequences. These sequences are then integrated to obtain a PV prediction sample dataset. P PV prediction sample data are randomly selected with replacement from this dataset, and ensemble machine learning is used to train P PV power generation prediction branches. Based on the microgrid operation logs over a historical period, a set of historical demand load power sequences is collected. The demand load power sequence following each historical demand load power sequence is used as a sample predicted demand load power sequence, resulting in a set of sample predicted demand load power sequences. These sequences are then integrated to obtain a load prediction sample dataset. P load prediction sample data are randomly selected with replacement from this dataset, and ensemble machine learning is used to train P load power generation prediction branches. The P PV power generation prediction branches and the P load power generation prediction branches are then combined to obtain a PV power generation prediction branch array and a demand load prediction branch array.

[0035] Specifically, firstly, based on the microgrid operation logs from historical periods (e.g., the most recent month), sample historical photovoltaic power generation sequences are collected from multiple different time periods to obtain a set of sample historical photovoltaic power generation sequences. Next, the photovoltaic power generation sequence following each sample historical photovoltaic power generation sequence is used as a sample predicted photovoltaic power generation sequence to obtain a set of sample predicted photovoltaic power generation sequences. Then, the set of sample historical photovoltaic power generation sequences and the set of sample predicted photovoltaic power generation sequences are integrated to obtain a photovoltaic prediction sample dataset. This photovoltaic prediction sample dataset is then divided into P equal parts, where P is a positive integer. The specific value of P can be set according to actual prediction needs, such as setting P to 20, resulting in P sample datasets. Further, P samples are selected with replacement from the P sample datasets to obtain the first photovoltaic prediction sample data. The same method is used to iteratively select P times to obtain P photovoltaic prediction sample data.

[0036] Then, P photovoltaic power generation prediction branches are constructed based on machine learning, such as using a BP neural network to construct photovoltaic power generation prediction branches. The photovoltaic power generation prediction branches are used to learn from historical photovoltaic power generation data and then predict future photovoltaic power generation. They include an input layer, multiple hidden layers and an output layer. The input data of the input layer is the historical photovoltaic power generation sequence, and the output data of the output layer is the predicted photovoltaic power generation sequence. Next, using the historical photovoltaic power generation sequence as input and the predicted photovoltaic power generation sequence as output, supervised training is performed on P photovoltaic power generation prediction branches using the P photovoltaic prediction sample data. The training process is as follows: First, the historical photovoltaic power generation sequence is passed to the hidden layer of the neural network through the input layer, and a series of weighted calculations and activation function processing are performed. The output layer outputs a predicted value. Then, the mean squared error loss function is used to calculate the error between the output of the neural network and the actual predicted photovoltaic power generation (target output). The gradient of the error with respect to the weight of each neuron is calculated through the backpropagation algorithm. Backpropagation updates the weights and biases in the neural network through gradient descent, making the output closer to the target output. Then, the above steps (forward propagation, error calculation, backpropagation, and weight update) are repeated. The network weights are adjusted in each iteration to make the prediction results more and more accurate until the error converges, resulting in P photovoltaic power generation prediction branches that have been trained.

[0037] On the other hand, based on the microgrid operation logs within a historical period (e.g., the most recent month), a set of sample historical demand load power sequences is collected. Then, the demand load power sequence following each sample historical demand load power sequence is used as the sample predicted demand load power sequence, resulting in a set of sample predicted demand load power sequences. The sample historical demand load power sequence set and the sample predicted demand load power sequence set are then integrated to obtain a load prediction sample dataset. Further, the load prediction sample dataset is divided into P equal parts, and P random selections are made from the P parts with replacement to obtain the first load prediction sample data. This process is repeated P times to obtain P load prediction sample data.

[0038] Then, P load generation prediction branches are constructed based on a backpropagation (BP) neural network. Each load generation prediction branch includes an input layer, multiple hidden layers, and an output layer. The input data of the input layer is the historical demand load power sequence of the samples, and the output data of the output layer is the predicted demand load power sequence of the samples. Furthermore, the P load generation prediction branches are trained under supervised supervision using P load prediction sample data until convergence, resulting in P trained load generation prediction branches. The training method for the load generation prediction branches is the same as that for the photovoltaic power generation prediction branches described above, and will not be elaborated here.

[0039] Finally, the P photovoltaic power generation forecasting branches are combined to construct a photovoltaic power generation forecasting branch array, and the P load power generation forecasting branches are combined to construct a demand load forecasting branch array. By constructing the photovoltaic power generation forecasting branch array and the demand load forecasting branch array based on machine learning, the intelligence, accuracy, and efficiency of photovoltaic power generation forecasting and demand load forecasting can be improved, thereby improving the accuracy and efficiency of subsequent power allocation optimization.

[0040] Based on the historical volatility and P, Q is calculated, where Q is a positive integer. Q photovoltaic power generation forecast branches and Q demand load forecast branches are randomly selected. Photovoltaic power generation and demand load are predicted for K future moments within a future time period based on the historical photovoltaic power generation sequence and historical demand load sequence, resulting in Q branch predicted photovoltaic power generation sequences and Q branch predicted demand load sequences. The average values ​​of the Q branch predicted photovoltaic power generation and Q branch predicted demand load sequences at each future moment are calculated to obtain the predicted photovoltaic power generation sequence and the predicted demand load sequence.

[0041] Specifically, the historical volatility is multiplied by P and rounded to obtain Q, where Q is a positive integer. For example, if the historical volatility is 18% and P is 20, then Q is 4. Further, Q photovoltaic power generation forecast branches are randomly selected from the P photovoltaic power generation forecast branches, and Q demand load forecast branches are randomly selected from the P load power generation forecast branches. Then, the photovoltaic power generation of the historical photovoltaic power generation sequence is predicted for K future moments within a future time period using the Q photovoltaic power generation forecast branches, resulting in Q branch-predicted photovoltaic power generation sequences. Finally, the demand load of the historical demand load power sequence is predicted for K future moments within a future time period using the Q demand load forecast branches, resulting in Q branch-predicted demand load power sequences.

[0042] Finally, the mean value of the predicted photovoltaic power generation of the Q branches at each future time point within the predicted photovoltaic power generation power sequence of the Q branches is calculated to obtain the predicted photovoltaic power generation power sequence; the mean value of the predicted demand load power of the Q branches at each future time point within the predicted demand load power sequence of the Q branches is calculated to obtain the predicted demand load power sequence.

[0043] By performing volatility analysis on photovoltaic power generation and demand load over historical periods, and selecting an appropriate number of prediction branches based on the overall volatility for both photovoltaic power generation and demand load prediction, the fit between the number of prediction branches and the actual volatility can be improved. This allows for the addition of more prediction branches when volatility is high, enhancing the model's robustness and prediction accuracy; conversely, reducing the number of prediction branches when volatility is low avoids overfitting and maintains model simplicity. This approach saves computational resources, reduces unnecessary computational burden, and improves prediction efficiency while ensuring prediction accuracy.

[0044] S2: Based on the historical photovoltaic power generation sequence, historical demand load power sequence, predicted photovoltaic power generation sequence, and predicted demand load power sequence, analyze and obtain the predicted volatility and the energy supply deviation amplitude. Combine the predicted volatility and historical volatility to calculate the volatility. Based on the volatility and the energy supply deviation amplitude, configure and obtain the frequency regulation coefficient and peak shaving allocation coefficient.

[0045] Furthermore, step S2 of the present invention also includes:

[0046] Based on the predicted photovoltaic power generation sequence and the predicted demand load power sequence, the predicted volatility is analyzed and calculated; the average of the predicted volatility and the historical volatility is calculated to obtain the volatility; based on the historical photovoltaic power generation sequence, the historical demand load power sequence, the predicted photovoltaic power generation sequence, and the predicted demand load power sequence, the power deviation amplitude of each demand load power and the corresponding photovoltaic power generation is calculated, and the average is calculated to obtain the energy supply deviation amplitude; the volatility and the energy supply deviation amplitude are configured as frequency regulation coefficient and peak shaving allocation coefficient.

[0047] Specifically, firstly, K predicted fluctuations are calculated based on the predicted photovoltaic power generation sequence, and the average is used to obtain the photovoltaic predicted fluctuation. Then, the load predicted fluctuation is calculated based on the predicted demand load power sequence. Next, the average of the photovoltaic predicted fluctuation and the load predicted fluctuation is calculated to obtain the predicted fluctuation. Finally, the average of the predicted fluctuation and the historical fluctuation is calculated to obtain the volatility, which reflects the overall volatility measure of the microgrid within the current time period.

[0048] Then, based on the historical photovoltaic power generation sequence, historical demand load power sequence, predicted photovoltaic power generation sequence, and predicted demand load power sequence, the power deviation amplitude of each demand load power and the corresponding photovoltaic power generation at the same time is calculated respectively. The power deviation amplitude is the ratio of the absolute value of the power difference between the demand load power and the corresponding photovoltaic power generation to the demand load power, resulting in K historical power deviation amplitudes and K predicted power deviation amplitudes. Then, the average of the K historical power deviation amplitudes and the K predicted power deviation amplitudes is calculated, that is, the sum of the K historical power deviation amplitudes and the K predicted power deviation amplitudes is divided by 2K to obtain the power supply deviation amplitude.

[0049] Furthermore, the volatility is set as the frequency regulation coefficient. The larger the volatility, the larger the frequency regulation coefficient, indicating that the frequency fluctuation within the microgrid is greater, and thus the frequency regulation demand is greater. The power supply deviation amplitude is set as the peak shaving allocation coefficient. The larger the power supply deviation amplitude, the greater the power supply gap within the microgrid, meaning that photovoltaic power generation cannot meet the demand during high-load periods, and energy storage units need to provide greater support to ensure that the load is met, thus the peak shaving demand to meet peak electricity demand is greater.

[0050] S3: Randomly generate the power allocation scheme of the energy storage unit, perform peak shaving and frequency regulation analysis based on the predicted photovoltaic power generation sequence and the predicted demand load power sequence, and calculate the allocation fitness according to the frequency regulation coefficient and peak shaving allocation coefficient. The power allocation scheme includes the allocation power sequence in the future time period, and the allocation fitness is calculated based on the deviation of adjacent allocation power.

[0051] Furthermore, step S3 of the present invention also includes:

[0052] Obtain the energy allocation space for the energy storage unit to distribute energy to the power consumption side; randomly generate K first energy allocation powers within the energy allocation space as the first allocation power sequence for K future times in the future period to obtain a first power allocation scheme; calculate a first actual energy supply power sequence based on the first allocation power sequence and the predicted photovoltaic power generation power sequence; subtract the corresponding predicted demand load power in the predicted demand load power sequence from each first actual energy supply power in the first actual energy supply power sequence to obtain a first redundant energy supply power sequence.

[0053] Specifically, firstly, the energy allocation space for the energy storage unit to distribute energy to the power consumption side is obtained, that is, the difference between the maximum and minimum power that the energy storage unit can discharge. Next, K first energy allocation powers are randomly generated within the energy allocation space. If K energy allocation powers are randomly selected sequentially within the energy allocation space, they are used as the first allocation power sequence for K future times in the future time period, thus obtaining a first power allocation scheme. Then, based on the first allocation power sequence and the predicted photovoltaic power generation sequence, the first allocation power and the predicted photovoltaic power generation at the same time are added together to obtain the first actual energy supply power. The first actual energy supply power sequence is then calculated sequentially.

[0054] Further, each first actual energy supply power in the first actual energy supply power sequence is subtracted from the corresponding predicted demand load power in the predicted demand load power sequence, that is, the first actual energy supply power at the same time is subtracted from the predicted demand load power, and set as the first redundant energy supply power, that is, the excess power of the energy storage unit, to obtain the first redundant energy supply power sequence.

[0055] Based on the first redundant power supply sequence and the first power allocation sequence, and according to the frequency modulation coefficient and the peak shaving allocation coefficient, the first allocation fitness is calculated, wherein the allocation fitness is calculated based on the deviation between adjacent power allocations within the first power allocation sequence.

[0056] Furthermore, the present invention also includes the following steps:

[0057] Based on the first redundant power supply sequence and the first power allocation sequence, and according to the frequency modulation coefficient and peak shaving allocation coefficient, the first allocation fitness is calculated as follows:

[0058]

[0059] Among them, F f To assign fitness, w1, w2, and w3 are weights, and their sum is 1. N is a positive integer, TP is the frequency modulation coefficient, TF is the peak shaving allocation coefficient, K is the number of redundant power supplies within the redundant power supply sequence, σ is the variance of the K redundant power supplies within the redundant power supply sequence, and G... i M represents the redundant power supply at the i-th future time. i For the power allocation at the i-th future time, M i+1 The power allocated at the (i+1)th future time.

[0060] Specifically, an assignment fitness evaluation function is constructed, in which F fThe allocation fitness is defined as follows: a higher allocation fitness indicates a better overall effect of the power allocation scheme; w1 is the direct frequency regulation weight, w2 is the peak shaving weight, and w3 is the indirect frequency regulation weight, with a sum of 1. These weights can be set according to the degree of influence of the indicators on the allocation fitness; the greater the influence, the greater the corresponding weight. These weights can also be dynamically adjusted according to the actual control needs of the microgrid; N is a positive integer, TP is the frequency regulation coefficient, TF is the peak shaving allocation coefficient, K is the number of redundant power supplies in the redundant power supply sequence, and σ is the variance of the K redundant power supplies in the redundant power supply sequence. A smaller variance indicates a more stable overall power supply and a better frequency regulation effect; G i Let M be the redundant power supply at the i-th future time. The closer the redundant power supply is to 0, the smaller the deviation between the power supply and the power consumption side. This indicates that the power supply just meets the power consumption side's demand, meaning the peak demand is just met, resulting in a better peak-shaving effect. i For the power allocation at the i-th future time, M i+1 Let be the power allocation at the (i+1)th future time. Energy storage units need efficient response during frequency regulation, but adjusting the power allocation takes time. When the power difference between adjacent allocations is large, it's difficult to adjust the power quickly and accurately, affecting the frequency regulation response speed. Therefore, a smaller allocation fitness indicates a smaller deviation between adjacent allocations, making it easier and more accurate for energy storage units to adjust the power allocation, resulting in better indirect frequency regulation and a larger allocation fitness. A larger frequency regulation coefficient (TP) indicates greater fluctuations in photovoltaic power generation and electricity demand, thus increasing the importance of frequency regulation within the allocation fitness during optimization. Similarly, a larger peak-shaving allocation coefficient indicates a larger current demand load gap on the electricity side, further increasing the importance of peak shaving within the allocation fitness during optimization.

[0061] By constructing an allocation fitness evaluation function, the performance of power allocation schemes in frequency regulation and peak shaving can be comprehensively and quantitatively evaluated, improving the accuracy and precision of power allocation scheme evaluation, providing a scientific basis for achieving efficient frequency regulation and peak shaving, and thus ensuring the accuracy and reliability of setting the optimal power allocation scheme.

[0062] S4: Based on the allocation fitness, iteratively optimize and evaluate the power allocation scheme, optimize to obtain the optimal power allocation scheme, and control the energy storage unit to allocate power within the microgrid.

[0063] Furthermore, step S4 of the present invention also includes:

[0064] Continue to randomly generate power allocation schemes and calculate allocation fitness and iterative optimization; until the iterative optimization converges, retain the power allocation scheme with the largest allocation fitness as the optimal power allocation scheme, and control the energy storage unit to allocate power supply in the microgrid at K future times in the future time period.

[0065] Specifically, a second power allocation scheme is randomly generated in the power allocation space, wherein the second power allocation scheme is different from the first power allocation scheme, and the second allocation fitness of the second power allocation scheme is calculated; the iterative selection of power allocation schemes and the calculation of allocation fitness continue until a predetermined number of selections is reached (which can be set according to the optimization accuracy, such as 100 times), and multiple power allocation schemes and multiple allocation fitnesss are output; then, the power allocation scheme with the largest allocation fitness is retained as the optimal power allocation scheme, and the energy storage unit is controlled to perform power supply and distribution in the microgrid at K future times in the future time period according to the optimal power allocation scheme.

[0066] The present invention provides a method for frequency regulation, peak shaving, and power distribution in a smart microgrid based on photovoltaic power generation, which has at least the following technical effects:

[0067] 1. By optimizing the power distribution of energy storage units, demand load can be accurately matched during peak electricity consumption periods, improving the scientific nature and accuracy of power distribution, effectively mitigating the negative impact of photovoltaic power generation fluctuations, and ensuring that energy storage units can quickly respond to load fluctuations, achieving efficient frequency regulation and peak shaving, thereby significantly improving the overall operational stability and reliability of the microgrid.

[0068] 2. By selecting an appropriate number of prediction branches based on historical volatility for photovoltaic power generation and demand load prediction, the fit between the number of prediction branches and the actual volatility can be improved. More prediction branches can be added when volatility is high, improving the model's robustness and prediction accuracy; while fewer prediction branches can be added when volatility is low, thus avoiding overfitting and maintaining model simplicity. This allows for saving computing resources, reducing unnecessary computational burden, and improving prediction efficiency while ensuring prediction accuracy.

[0069] 3. By constructing an allocation fitness evaluation function, the performance of power allocation schemes in frequency regulation and peak shaving can be comprehensively and quantitatively evaluated, improving the accuracy and precision of power allocation scheme evaluation, providing a scientific basis for achieving efficient frequency regulation and peak shaving, and thus ensuring the accuracy and reliability of setting the optimal power allocation scheme.

[0070] 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.

[0071] 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 frequency modulation, peak regulation and power distribution of an intelligent micro-grid based on photovoltaic power generation, characterized in that, The method is applied to a micro-grid, the micro-grid including a photovoltaic generator set and an energy storage set, and the method comprises: obtaining a historical photovoltaic power sequence and a historical demand load power sequence in the micro-grid, and analyzing to obtain a historical fluctuation degree; performing photovoltaic power generation and demand load prediction in a future period according to the historical fluctuation degree, to obtain a predicted photovoltaic power sequence and a predicted demand load power sequence; according to the historical photovoltaic power sequence, the historical demand load power sequence, the predicted photovoltaic power sequence and the predicted demand load power sequence, analyzing to obtain a predicted fluctuation degree and an energy supply deviation amplitude, combining the predicted fluctuation degree and the historical fluctuation degree, calculating to obtain a fluctuation degree, according to the fluctuation degree and the energy supply deviation amplitude, configuring to obtain a frequency modulation coefficient and a peak regulation distribution coefficient; randomly generating an energy distribution scheme of the energy storage set, performing peak regulation and frequency modulation analysis according to the predicted photovoltaic power sequence and the predicted demand load power sequence, and calculating to obtain a distribution fitness according to the frequency modulation coefficient and the peak regulation distribution coefficient, wherein the energy distribution scheme comprises a distribution power sequence in the future period, and the distribution fitness is calculated according to the deviation of adjacent distribution powers, comprising: obtaining an energy distribution space in which the energy storage set distributes energy to the power consumption side; randomly generating K first energy distribution powers in the energy distribution space as a first distribution power sequence of K future time points in the future period, to obtain a first energy distribution scheme; calculating to obtain a first actual energy supply power sequence according to the first distribution power sequence and the predicted photovoltaic power sequence; subtracting each first actual energy supply power in the first actual energy supply power sequence from a corresponding predicted demand load power in the predicted demand load power sequence to obtain a first redundant energy supply power sequence; calculating to obtain a first distribution fitness according to the first redundant energy supply power sequence and the first distribution power sequence according to the frequency modulation coefficient and the peak regulation distribution coefficient, wherein the distribution fitness is calculated according to the deviation of adjacent distribution powers in the first distribution power sequence, comprising: wherein, is the assigned fitness, , and are weights, the sum of which is 1, N is a positive integer, TP is a tuning factor, TF is a tuning factor for the distribution, K is the number of redundant power supplies in the sequence of redundant power supplies, is the variance of the K redundant power supplies in the sequence of redundant power supplies, is the redundant power supply at the i-th future time, is the assigned power at the i-th future time, is the assigned power at the i+1-th future time; performing iterative optimization evaluation of the energy distribution scheme according to the distribution fitness, and optimizing to obtain an optimal energy distribution scheme, and controlling the energy storage set to distribute energy in the micro-grid. 2.The method of claim 1, wherein, obtaining a historical photovoltaic power sequence and a historical demand load power sequence in a micro-grid, and analyzing to obtain a historical fluctuation degree; performing photovoltaic power generation and demand load prediction in a future period according to the historical fluctuation degree, to obtain a predicted photovoltaic power sequence and a predicted demand load power sequence, comprising: obtaining photovoltaic power of the photovoltaic generator set and demand load power of the power consumption side at K past historical time points in the micro-grid, to obtain a historical photovoltaic power sequence and a historical demand load power sequence, K being a positive integer; analyzing fluctuation degrees of the historical photovoltaic power sequence and the historical demand load power sequence, to obtain a historical photovoltaic fluctuation degree and a historical demand fluctuation degree, and calculating to obtain a historical fluctuation degree; The integrated machine learning is used to construct a photovoltaic power generation prediction branch array and a demand load prediction branch array, wherein the photovoltaic power generation prediction branch array and the demand load prediction branch array respectively include P photovoltaic power generation prediction branches and P demand load prediction branches, and P is a positive integer; Q is obtained according to the historical fluctuation degree and P, and Q is a positive integer; Q photovoltaic power generation prediction branches and Q demand load prediction branches are randomly selected, and photovoltaic power generation and demand load prediction in K future time points in a future time period are performed on the historical photovoltaic power sequence and the historical demand load power sequence to obtain Q branch prediction photovoltaic power sequences and Q branch prediction demand load power sequences; The average values of the Q branch prediction photovoltaic power and the Q branch prediction demand load power in each future time point in the Q branch prediction photovoltaic power sequence and the Q branch prediction demand load power sequence are calculated to obtain a prediction photovoltaic power sequence and a prediction demand load power sequence. 3.The method of claim 2, wherein, The fluctuation degrees of the historical photovoltaic power sequence and the historical demand load power sequence are analyzed to obtain a historical photovoltaic fluctuation degree and a historical demand fluctuation degree, and a historical fluctuation degree is calculated, including: A first historical photovoltaic power in a first historical time point in the historical photovoltaic power sequence is selected, and a plurality of random historical photovoltaic power is randomly selected; The average value of the plurality of random historical photovoltaic power and the fluctuation amplitude of the first historical photovoltaic power are calculated to obtain a first photovoltaic fluctuation degree; K photovoltaic fluctuation degrees are continuously calculated, and the average value is calculated to obtain a historical photovoltaic fluctuation degree; The historical demand fluctuation degree is calculated according to the historical demand load power sequence; The average value of the historical photovoltaic fluctuation degree and the historical demand fluctuation degree is calculated to obtain a historical fluctuation degree. 4.The method of claim 2, wherein, The integrated machine learning is used to construct a photovoltaic power generation prediction branch array and a demand load prediction branch array, including: According to the micro-grid operation record log in the historical time, a sample historical photovoltaic power sequence set is collected, and the photovoltaic power sequence after each sample historical photovoltaic power sequence is taken as a sample prediction photovoltaic power sequence to obtain a sample prediction photovoltaic power sequence set, and a photovoltaic prediction sample data set is integrated; P photovoltaic prediction sample data is randomly selected from the photovoltaic prediction sample data set with replacement, and the integrated machine learning is used to train P photovoltaic power generation prediction branches; According to the micro-grid operation record log in the historical time, a sample historical demand load power sequence set is collected, and the demand load power sequence after each sample historical demand load power sequence is taken as a sample prediction demand load power sequence to obtain a sample prediction demand load power sequence set, and a load prediction sample data set is integrated; P load prediction sample data is randomly selected from the load prediction sample data set with replacement, and the integrated machine learning is used to train P load power generation prediction branches; The P photovoltaic power generation prediction branches and the P load power generation prediction branches are combined respectively to obtain a photovoltaic power generation prediction branch array and a demand load prediction branch array. 5.The method of claim 1, wherein, According to the historical photovoltaic power generation sequence, the historical demand load power sequence, the predicted photovoltaic power generation sequence and the predicted demand load power sequence, a predicted fluctuation degree and a power supply deviation amplitude are obtained by analysis, a fluctuation degree is calculated by combining the predicted fluctuation degree and a historical fluctuation degree, a frequency modulation coefficient and a peak shaving distribution coefficient are configured, and the configuration includes: According to the predicted photovoltaic power generation sequence and the predicted demand load power sequence, a predicted fluctuation degree is obtained by analysis and calculation; The mean value of the predicted fluctuation degree and a historical fluctuation degree is calculated to obtain a fluctuation degree; According to the historical photovoltaic power generation sequence, the historical demand load power sequence, the predicted photovoltaic power generation sequence and the predicted demand load power sequence, the power deviation amplitude of each demand load power and the corresponding photovoltaic power generation power is calculated, and the mean value is calculated to obtain a power supply deviation amplitude; The fluctuation degree and the power supply deviation amplitude are configured as a frequency modulation coefficient and a peak shaving distribution coefficient. 6.The method of claim 1, wherein, According to the distribution fitness, an iterative optimization evaluation of the electric energy distribution scheme is performed, and an optimal electric energy distribution scheme is obtained by optimization, and the configuration includes: The electric energy distribution scheme is continuously randomly generated, and the distribution fitness and iterative optimization are calculated; Until the iterative optimization converges, the electric energy distribution scheme with the maximum distribution fitness is retained as the optimal electric energy distribution scheme, and the energy storage unit is controlled to supply and distribute electric energy in the micro-grid at K future time points in the future period.

Images