Electricity price tracking and electricity selling method based on AI model and related equipment

Abstract

The invention relates to the technical field of energy storage equipment, provides an electricity price tracking and selling method based on an AI model and related equipment, is applied to photovoltaic energy storage equipment, and comprises the following steps: acquiring an electricity utilization habit of a user, predicting a solar radiation value of the next day and an electricity price fluctuation curve of the current day, and calculating photovoltaic power generation capacity of the next day based on the solar radiation value of the next day. And if the capacity is higher than the capacity of the photovoltaic energy storage equipment, predicting the highest electricity price time period of the electricity price based on an electricity price fluctuation curve, and analyzing an actual electricity consumption demand range and an actual electricity consumption time period according to the electricity consumption habit of the user so as to analyze the optimal partition electricity selling time of the photovoltaic energy storage equipment. Through intelligent electricity price prediction and user electricity demand analysis, optimal electric quantity distribution and flexible electricity selling of the photovoltaic energy storage equipment are realized, the electricity demand of the user is guaranteed preferentially, and the problem that the electricity price fluctuation is highly frequent, the power grid response mechanism is complex or the user load behavior is uncertain is solved. The problems that response is not timely and energy storage discharge opportunities are missed exist.

Description

TECHNICAL FIELD

[0001] The present application relates to the field of energy storage devices, in particular to a power price tracking and power selling method based on an AI model and related equipment. BACKGROUND

[0002] With the rapid development of distributed energy, the proportion of photovoltaic energy storage systems in power systems is increasing year by year. Especially in industrial and residential electricity scenarios, users install photovoltaic components and supporting energy storage devices to realize self-generation and self-use of energy, and sell surplus electricity to the grid, which has become one of the important paths to promote the popularization of green energy. At the same time, with the increasing volatility and uncertainty of new energy output, how to improve the economic utilization efficiency of energy storage power under the premise of ensuring power reliability has become a core problem in current energy management.

[0003] In related technical means, the power price tracking and power selling method based on static time strategy or rule engine allocates photovoltaic power generation power by presetting peak and valley periods or fixed thresholds. For example, some existing systems use pre-set power price thresholds or power price levels to prioritize the use of energy storage device power to meet user load during early and late peak periods, and the remaining part is sold or stored under low-price valley periods or capacity restrictions. Improving user-side income and reducing electricity costs also achieves preliminary scheduling optimization of photovoltaic energy storage resources.

[0004] For the above technical solutions, although rule-based time matching and power price threshold judgment can achieve basic power price tracking and power selling control, in the case of highly frequent power price fluctuations, complex grid response mechanisms, or uncertain user load behavior, there are problems of delayed response and missed energy storage discharge opportunities. SUMMARY

[0005] In order to improve the problem of delayed response and missed energy storage discharge opportunities in the case of highly frequent power price fluctuations, complex grid response mechanisms, or uncertain user load behavior, the present application provides a power price tracking and power selling method based on an AI model and related equipment.

[0006] The application provides an electricity price tracking and selling method based on an AI model, applied to a photovoltaic energy storage device, and includes the following steps: obtaining user electricity usage habits, and predicting solar radiation values for the next day and electricity price fluctuation curves for the current day; calculating photovoltaic power generation for the next day based on the solar radiation values for the next day; if the photovoltaic power generation for the next day is higher than the capacity of the photovoltaic energy storage device, predicting the highest electricity price time period of the electricity price based on the electricity price fluctuation curves for the current day, and analyzing the actual electricity demand range and actual electricity time period of the user according to the user electricity usage habits; comparing the highest electricity price time period with the electricity time period, if the highest electricity price time period and the electricity time period are within a preset time interval range, reserving the electricity quantity meeting the electricity demand range of the photovoltaic energy storage device, and selling the remaining electricity quantity of the photovoltaic energy storage device; if the highest electricity price time period and the electricity time period are not within the preset time interval range, selling all the electricity quantity of the photovoltaic energy storage device.

[0007] As a preferred solution, the step of obtaining user electricity usage habits and predicting solar radiation values for the next day and electricity price fluctuation curves for the current day includes the following steps: obtaining user historical electricity data, performing time series analysis on the user historical electricity data, extracting electricity frequency, electricity quantity change amplitude and periodic characteristics of the user in different time periods to construct user electricity usage habits; inputting the user electricity usage habits and weather historical data into a preset time sequence prediction network to predict user electricity trends for the next 24 hours and key meteorological parameters for the next day, inputting the key meteorological parameters into a preset solar radiation prediction sub-model to output solar radiation values for the next day; using a feature extraction network to extract features of grid historical electricity price records, holiday information and demand response policies to construct electricity price fluctuation related features, inputting the electricity price fluctuation related features into a preset short-term price prediction model to output electricity price fluctuation curves for the current day.

[0008] As a preferred solution, the step of inputting the electricity price fluctuation related features into a preset short-term price prediction model to output electricity price fluctuation curves for the current day includes the following steps: inputting the electricity price fluctuation related features into a preset short-term price prediction model to obtain predicted hourly electricity price sequences, electricity price fluctuation rate sequences and abnormal factor sets of high-frequency price fluctuations; performing fluctuation window sliding analysis on the hourly electricity price sequences, and extracting local maximum points and slope information based on the analysis results, using the local maximum points and the slope information to construct price fluctuation intensity and price trend change rate of each time period; weighting and fusing the price fluctuation intensity and the electricity price fluctuation rate sequences to generate a price fluctuation weight curve; identifying cross-time points of the price trend change rate and the abnormal factor sets to generate a sensitive time period set of price jumps; performing joint clustering analysis on the price fluctuation weight curve and the sensitive time period set to obtain electricity price fluctuation curves for the current day.

[0009] As a preferred solution, the step of calculating the photovoltaic power generation of the next day based on the solar radiation value of the next day comprises: combining and calculating the solar radiation value of the next day with the area, inclination parameter and conversion efficiency parameter of the photovoltaic panel device by using a preset combination calculation formula to obtain the photovoltaic power generation of each period of the next day; wherein the combination calculation formula is as follows: P t =I t ×A×η×cos(θ); wherein, the P t is the photovoltaic power generation at time t; the I t is the solar radiation value at time t; the A is the effective area of the photovoltaic panel; the η is the conversion efficiency of the photovoltaic panel; and the θ is the inclination of the photovoltaic panel.

[0010] As a preferred solution, the step of predicting the highest electricity price time period based on the electricity price fluctuation curve of the day comprises: performing standardization processing on the electricity price fluctuation curve of the day to construct a time gradient curve reflecting the price fluctuation range, performing peak value detection on the time gradient curve to identify candidate time periods of rapid price rise; extracting the price growth rate and duration corresponding to the candidate time periods, and performing normalized aggregation analysis on the price growth rate and the duration to obtain a fluctuation intensity score matrix at the time period level, inputting the fluctuation intensity score matrix into a preset price risk identification model to output a confidence-weighted price abnormal risk factor and a price rise rate change map; performing hierarchical threshold analysis on the price abnormal risk factor to identify price high-rise candidate time periods, and performing gradient reconstruction on the price high-rise candidate time periods and the price rise rate change map to extract a main peak period; and comparing the main peak period with a preset electricity price peak threshold to screen out the highest electricity price time period with a price higher than the average peak value.

[0011] As a preferred solution, the step of analyzing the actual electricity demand range and actual electricity time period of the user according to the user electricity habit comprises: performing clustering modeling on the user electricity habit to obtain a user typical daily electricity template, inputting the user typical daily electricity template into a preset user behavior prediction model to predict the electricity consumption at different time periods in the next 24 hours of the user; and performing segmented statistics and sliding window regression analysis on the electricity consumption at different time periods to obtain the actual electricity demand range and actual electricity time period of the user.

[0012] As a preferred solution, the step of performing segmented statistics and sliding window regression analysis on the electricity consumption in the different time periods to obtain the user's actual electricity demand range and actual electricity consumption time period includes: performing segmented statistics on the electricity consumption in the different time periods to obtain statistical results, extracting the mean and deviation based on the statistical results, and identifying the basic electricity demand range and the high-intensity electricity demand range based on the mean and the deviation; fusing the basic electricity demand range with the high-intensity electricity demand range to generate a continuous non-overlapping electricity intensity curve, performing time series fitting on the electricity intensity curve, and extracting the peak period, duration and sudden increase point based on the fitting result, and constructing the user's actual electricity demand range according to the peak period, duration and sudden increase point; performing sliding window regression analysis on the electricity consumption in the different time periods to extract stable load areas and sudden load areas, and constructing a load change trend curve according to the stable load area and the sudden load area; performing regression fitting and cross-analysis on the load change trend curve through the actual electricity demand range to obtain the user's actual electricity consumption time period.

[0013] The present application also provides an AI-model-based electricity price tracking and electricity sales system, which is applied to photovoltaic energy storage equipment, including: an acquisition module, used to obtain the user's electricity usage habits, and predict the solar radiation value of the next day and the electricity price fluctuation curve of the current day, and calculate the photovoltaic power generation of the next day based on the solar radiation value of the next day; a prediction module, used to predict the highest electricity price time period based on the electricity price fluctuation curve of the current day if the photovoltaic power generation of the next day is higher than the capacity of the photovoltaic energy storage equipment, and analyze the user's actual electricity demand range and actual electricity usage time period according to the user's electricity usage habits; a comparison module, used to compare the highest electricity price time period with the electricity usage time period, if the highest electricity price time period and the electricity usage time period are within a preset time interval range, the photovoltaic energy storage equipment reserves the electricity that meets the electricity demand range, and sells the remaining electricity of the photovoltaic energy storage equipment through inter-wall electricity sales; a power sales module, used to sell all the electricity of the photovoltaic energy storage equipment through inter-wall electricity sales if the highest electricity price time period and the electricity usage time period are not within the preset time interval range.

[0014] The present application also provides an electronic device including a memory and a processor, wherein the memory stores a computer program that can be run on the processor, and when the processor executes the computer program, it implements any of the above-mentioned AI model-based electricity price tracking and electricity selling methods.

[0015] The present application also provides a computer-readable storage medium having a computer program stored thereon. When the computer program is executed by a processor, the processor executes the electricity price tracking and electricity selling method based on the AI ​​model as described in any one of the above.

[0016] Compared with the prior art, the application has the following beneficial effects: high flexibility and strong economic benefit. Through intelligent electricity price prediction and user electricity demand analysis, optimal electricity distribution and flexible electricity selling of the photovoltaic energy storage device are realized. When the electricity price peak period is predicted, the user electricity demand is preferentially guaranteed, and the remaining electricity is sold during the electricity price valley period, thereby maximizing the economic benefit of the energy storage device. Through the introduction of the electricity price fluctuation prediction model and the user electricity habit analysis, the function of real-time adjustment of electricity distribution according to the electricity price fluctuation and the user demand change is realized. Not only the economy of user electricity is improved, but also the efficient use of the photovoltaic energy storage device is promoted. The problems of untimely response and missed energy storage discharge opportunities in the case of high frequency of electricity price fluctuation, complex power grid response mechanism or uncertain user load behavior are improved. BRIEF DESCRIPTION OF DRAWINGS

[0017] In order to more clearly illustrate the technical solutions in the embodiments of the present application or the prior art, the drawings needed to be used in the embodiments or the prior art description will be briefly introduced below. Obviously, the drawings in the following description are only some embodiments of the present application, and other drawings can be obtained by those skilled in the art without creative labor.

[0018] The structures, proportions, sizes, etc. shown in the drawings of the present specification are only used to cooperate with the content disclosed in the specification, to enable those skilled in the art to understand and read, and are not used to limit the limiting conditions of the embodiments of the present application. Therefore, any modification of structure, change of proportion relationship or adjustment of size, which does not affect the effects and purposes that can be achieved by the present application, should still fall within the scope of the technical content disclosed by the present application.

[0019] Figure 1 is a flowchart of the electricity price tracking and electricity selling method based on the AI model provided by the embodiments of the present application; Figure 2 is a structural schematic block diagram of the electricity price tracking and electricity selling system based on the AI model provided by the embodiments of the present application; Figure 3 is a structural schematic block diagram of the electronic device provided by the embodiments of the present application.

[0020] Explanation of reference signs: 10, electricity price tracking and electricity selling system based on AI model; 11, acquisition module; 12, prediction module; 13, comparison module; 14, electricity selling module; 20, electronic device; 21, memory; 22, processor. DETAILED DESCRIPTION

[0021] The following will clearly and completely describe the technical solutions in the embodiments of the present invention in conjunction with the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of them. All other embodiments obtained by ordinary technicians in this field based on the embodiments of the present invention without making any creative efforts shall fall within the scope of protection of the present invention.

[0022] The flowcharts shown in the accompanying drawings are for illustrative purposes only and do not necessarily include all contents and operations / steps, nor must they be executed in the order described. For example, some operations / steps may be decomposed, combined, or partially merged, so the actual execution order may vary depending on the actual situation.

[0023] It should also be understood that the terms used in this specification are for the purpose of describing specific embodiments only and are not intended to limit the present application. As used in this specification and the appended claims, the singular forms "a," "an," and "the" are intended to include the plural forms unless the context clearly indicates otherwise.

[0024] It should be further understood that the term "and / or" used in this specification and the appended claims refers to and includes any and all possible combinations of one or more of the associated listed items.

[0025] The technical solution of the present invention will be further described below with reference to the accompanying drawings and through specific implementation methods.

[0026] Example 1: like Figure 1 As shown, this application provides an electricity price tracking and electricity selling method based on an AI model, which is applied to photovoltaic energy storage equipment and includes the following steps: Step S100: Obtain the user's electricity usage habits, predict the solar radiation value of the next day and the electricity price fluctuation curve of the current day, and calculate the photovoltaic power generation of the next day based on the solar radiation value of the next day.

[0027] In this step, users' electricity usage habits are first analyzed through historical electricity data. Specifically, a time series prediction network (such as a long short-term memory (LSTM) network) is used to analyze past electricity usage data, extracting the frequency, fluctuations, and periodicity of electricity usage over different time periods. These characteristics help build a model of users' electricity usage habits and predict future electricity usage trends. Simultaneously, based on historical weather data and combined with a meteorological forecast model (for example, using a support vector machine (SVR) for weather data regression), key meteorological parameters for the next day, including temperature, cloud cover, and humidity, are predicted. This data is then fed into the solar radiation prediction sub-model.

[0028] For example, when obtaining the user's electricity usage habits, by analyzing the data of the past 30 days, it is identified that the user's electricity demand is higher during the morning peak (7:00-9:00) and the evening peak (17:00-19:00), which will affect the prediction of future electricity demand. In terms of solar radiation prediction, combined with the historical radiation data of the local meteorological bureau and the current weather conditions, the model predicts the solar radiation intensity of the next day to ensure the accuracy of the photovoltaic power generation calculation.

[0029] Step S200, if the photovoltaic power generation of the next day is higher than the capacity of the photovoltaic energy storage device, the highest electricity price period of the electricity price is predicted based on the electricity price fluctuation curve of the current day, and the actual electricity demand range and actual electricity time period of the user are analyzed according to the user's electricity usage habits.

[0030] In this step, the photovoltaic power generation of the next day is calculated by the photovoltaic power generation calculation model according to the predicted solar radiation value of the next day and the physical parameters of the photovoltaic panel (such as the area of the photovoltaic panel, conversion efficiency, etc.). If the power generation exceeds the capacity of the energy storage device, the highest time period of the electricity price is predicted according to the electricity price fluctuation curve (analyzed by combining historical electricity price data of the power grid and user electricity load). In order to further meet the user's demand, the actual electricity demand range is analyzed according to the user's electricity usage habits, and the actual electricity time period is identified.

[0031] For example, in calculating the photovoltaic power generation, the solar radiation value is combined with the inclination angle of the photovoltaic panel, conversion efficiency and other parameters to obtain the photovoltaic power generation curve of the next day. Through analysis of the electricity price fluctuation curve of the current day, the peak period of the electricity price can be predicted, such as from 10:00 to 12:00 in the morning and from 18:00 to 20:00 in the afternoon. At the same time, according to the user's electricity usage habits (for example, the user uses air conditioning from 8:00 to 10:00 in the morning and kitchen appliances during the noon period), the actual electricity demand time period is calculated, and the electricity consumption during this period is predicted.

[0032] Step S300, the highest electricity price time period is compared with the electricity time period, if the highest electricity price time period and the electricity time period are within the preset time interval range, the photovoltaic energy storage device reserves the electricity quantity that meets the electricity demand range, and the remaining electricity quantity of the photovoltaic energy storage device is used for wall separation electricity sales.

[0033] In this step, the predicted highest electricity price time period is compared with the actual electricity time period of the user. If there is overlap between the two, and the overlapping part meets the preset time interval range (for example, the set time interval is ±30 minutes), the photovoltaic energy storage device reserves the electricity quantity that meets the user's demand, and the remaining electricity quantity is used for wall separation electricity sales.

[0034] For example, if the predicted highest electricity price time period is from 10:00 to 12:00, and the user's electricity demand time period is from 8:00 to 10:00, the energy storage capacity will be reserved for the user from 10:00 to 10:30, and the remaining energy will be sold after 10:30. In this way, sufficient energy can be provided for the user during the highest electricity price period, and the remaining energy can be efficiently sold.

[0035] In step S400, if the highest electricity price time period and the electricity demand time period are not within the preset time interval range, all the energy of the photovoltaic energy storage device is sold to the grid.

[0036] In this step, when the highest electricity price time period and the user's electricity demand time period are not within the preset time interval range, all the energy of the photovoltaic energy storage device is sold to the grid, ensuring that the energy in the energy storage device is timely and economically rewarded.

[0037] For example, if the predicted highest electricity price time period is from 18:00 to 20:00, and the user's electricity demand time period is from 8:00 to 10:00, there is no overlap between the two. In this case, all the energy in the photovoltaic energy storage device is sold to the grid to obtain the highest electricity price reward in the market.

[0038] In this embodiment, by obtaining the user's electricity usage habit, combining weather history data and a time series prediction network, the user's electricity usage trend in the next 24 hours and the key meteorological parameters of the next day are predicted, and the solar radiation value of the next day is also predicted. Next, based on the solar radiation value of the next day, the photovoltaic power generation calculation model is used to calculate the photovoltaic power generation capacity of the next day, and compared with the capacity of the photovoltaic energy storage device. If the photovoltaic power generation capacity of the next day exceeds the capacity of the photovoltaic energy storage device, the highest electricity price time period is predicted according to the electricity price fluctuation curve of the day. Then, according to the user's electricity usage habit, the actual electricity demand range and the actual electricity time period of the user are analyzed, and the highest electricity price time period and the electricity time period are compared. If the highest electricity price time period and the electricity time period are within the preset time interval range, the photovoltaic energy storage device reserves the energy to meet the electricity demand range, and the remaining energy of the photovoltaic energy storage device is sold to the grid; if the highest electricity price time period and the electricity time period are not within the preset time interval range, all the energy of the photovoltaic energy storage device is sold to the grid, realizing the intelligent energy distribution and optimized electricity selling of the photovoltaic energy storage device.

[0039] By predicting electricity price fluctuations based on an AI model and analyzing user electricity demand, the user's demand can be prioritized during periods of high electricity prices, avoiding excessive discharge during periods of low electricity prices, and maximizing the economic benefits of photovoltaic energy storage devices. At the same time, the electricity of the photovoltaic energy storage device can be flexibly scheduled according to the electricity price fluctuations, ensuring that the electricity of the energy storage device is maximized during the highest electricity price period. Through the allocation of power resources, not only can unnecessary power waste be reduced, but also the cost of power purchase can be effectively reduced, the economic efficiency of electricity use can be improved, the efficient use of green energy can be promoted, and the problem of delayed response and missed energy storage discharge opportunities in the case of highly frequent electricity price fluctuations, complex power grid response mechanisms, or uncertain user load behavior can be improved.

[0040] Embodiment 2 In step S100, the user's electricity usage habits are obtained, and the solar radiation value of the next day and the electricity price fluctuation curve of the day are predicted. Specifically, it includes: Obtain the user's historical electricity data, perform time series analysis on the user's historical electricity data, extract the user's electricity frequency, electricity quantity change amplitude and periodicity characteristics in different time periods, and construct the user's electricity usage habits.

[0041] The historical electricity data is analyzed by the time series prediction network. Specifically, first, the historical electricity data is obtained from the user's smart meter or power management. These data include the user's electricity consumption in different time periods (such as every hour, every minute of every day). Then, a time series analysis method (such as a long short-term memory network LSTM or an ARIMA model) is used to process these data. This method can effectively capture the electricity usage rules, such as peak hours, low valleys, and periodic changes such as holidays. Through these analysis results, the user's electricity frequency, electricity quantity fluctuation amplitude, seasonal electricity change, etc. can be extracted, and the user's electricity usage habit model can be constructed.

[0042] For example, when obtaining the user's electricity usage habits, suppose the user's historical data shows that the user's electricity peak occurs from 7am to 9am and from 5pm to 7pm every day, and the electricity consumption is generally higher on weekends. Through LSTM model analysis of the data, the high peak period can be identified, and the user's future electricity demand can be predicted, and a targeted scheduling strategy can be provided based on the analysis.

[0043] The user's electricity usage habits and weather historical data are input into a preset time series prediction network to predict the user's electricity trend in the next 24 hours and the key meteorological parameters of the next day, including temperature, cloud cover and humidity. The key meteorological parameters are input into a preset solar radiation prediction sub-model, combined with historical similar sunshine data and current geographic location parameters, to output the solar radiation value of the next day.

[0044] By combining meteorological data and user electricity data for joint prediction; specifically, first input the user's electricity habits and historical meteorological data (including temperature, humidity, cloud cover, etc.) into a time series prediction network, such as an LSTM model or a GRU model (gated recurrent unit model). The time series network can predict the user's electricity trend in the next 24 hours, and at the same time, predict the key meteorological parameters (such as temperature, cloud cover, humidity, etc.) in the next 24 hours by combining weather changes. These data will be further transmitted to the solar radiation prediction sub-model, which predicts the solar radiation value of the next day by historical similar sunshine data and the user's current geographic location parameters. In this way, the user's electricity demand and photovoltaic power generation potential can be comprehensively predicted based on multiple factors.

[0045] For example, in actual application, assuming that the weather in the area where the user is located is relatively cold in winter, and the meteorological data predicts that the cloud cover will be large the next day. This information is input into the solar radiation prediction model. The model prediction result shows that the radiation value of the next day is low, which will affect the expected amount of photovoltaic power generation. At the same time, according to the user's electricity mode, the user's electricity demand in low temperature weather is predicted, and the electricity scheduling is further optimized.

[0046] The feature extraction network is used to extract features from the historical electricity price records, holiday information and demand response policies of the power grid to construct electricity price fluctuation related features. The electricity price fluctuation related features are input into a preset short-term price prediction model to output the electricity price fluctuation curve of the day.

[0047] The electricity price fluctuation feature model is constructed for prediction; specifically, historical electricity price data is obtained from the power grid management platform or third-party electricity market data provider. In order to accurately capture the trend of electricity price fluctuation, electricity price records, holiday information (such as electricity price fluctuation before and after holidays) and demand response policies are extracted. The feature extraction method can use convolutional neural network (CNN) or feature selection method based on regression analysis to ensure that the key factors of electricity price fluctuation are captured. The extracted electricity price fluctuation features are input into a preset short-term price prediction model (such as LSTM or random forest model). The model outputs the electricity price fluctuation curve of the day according to historical electricity price and demand response policy, helping to predict future electricity price trend.

[0048] For example, assuming that the extracted holiday information shows that the demand for electricity rises during the holiday period, and the electricity price also fluctuates. Therefore, combined with historical data and holiday information, it is predicted that the electricity price will rise significantly the day before the holiday. Based on this prediction, the discharge plan of the energy storage device can be adjusted in advance to sell electricity when the electricity price fluctuates most dramatically.

[0049] The step of inputting the electricity price fluctuation related features into the preset short-term price prediction model to output the electricity price fluctuation curve of the day specifically includes: The electricity price fluctuation related features are input into the preset short-term price prediction model to obtain the predicted hourly electricity price sequence, the electricity price fluctuation rate sequence, and the abnormal factor set of high-frequency price fluctuation.

[0050] The features are selected and calculated by the short-term price prediction model. Specifically, first, the key data in the historical electricity price records of the power grid are extracted by using a feature selection algorithm (such as the information gain method in the decision tree or the principal component analysis PCA based on correlation). These key data include the hourly electricity price variation trend, the peak-valley difference, seasonal fluctuation, etc. After the feature selection is completed, these data and the holiday information, demand response policy related features are input into the preset short-term price prediction model (such as the LSTM model or the random forest model) together. The model outputs three results through time series regression analysis: The hourly electricity price sequence reflects the fluctuation prediction of the electricity price of each hour of the day; The electricity price fluctuation rate sequence quantifies the rate of electricity price fluctuation; The abnormal factor set of high-frequency price fluctuation identifies the key influence indicators of the region of sharp price change.

[0051] For example, in electricity price prediction, the input historical data shows that the electricity price in a certain area is low before 8 am, and rapidly breaks through the peak from 3 pm to 5 pm. The short-term price prediction model outputs the hourly electricity price sequence of the day by analyzing these historical data and demand response policies, indicating that the electricity price will jump to a peak from 4 pm to 4:30 pm. At the same time, the electricity price fluctuation rate sequence identifies that the fluctuation rate significantly increases from 3 pm to 5 pm, and the abnormal factor set identifies that “near holiday” and “seasonal change” are the influencing factors of the electricity price jump.

[0052] The hourly electricity price sequence is analyzed by fluctuation window sliding, and based on the analysis results, the local maximum points and slope information are extracted, and the price fluctuation intensity and price trend change rate of each time period are constructed using the local maximum points and slope information.

[0053] The price fluctuation feature information is constructed by time window sliding analysis. Specifically, a fixed time window (such as 3 hours or 5 hours) is selected, and the hourly electricity price sequence is processed by sliding window. During the sliding window analysis, the local maximum points and time slope in the sequence are extracted by local function fitting (such as calculating the fitting slope by using the least squares method). The local maximum points represent the peak value of price fluctuation, and the slope represents the rate of price change. Combined with these two features, the price fluctuation intensity and price trend change rate of each time period are constructed: The price fluctuation intensity quantifies the peak amplitude during fluctuation; Price trend change rate: identify the change direction and rate of electricity price trend.

[0054] For example, in the analysis of the predicted electricity price from 3 pm to 6 pm, it is found that the local maximum point appears at 4 pm, and the slope increases significantly during 14:30-15:30. By fitting this window, the price fluctuation intensity is generated, indicating that the electricity price is expected to peak at 4 pm, while the trend change rate shows that the price gradually decreases after 4:30.

[0055] The price fluctuation intensity is weighted and fused with the electricity price volatility rate sequence to generate a price fluctuation weight curve.

[0056] The price fluctuation weight curve is generated by weighted fusion; specifically, the price fluctuation intensity is weighted and calculated with the electricity price volatility rate sequence. The weighting method can use the Poisson distribution method or a dynamic weight adjustment mechanism based on the standard deviation of electricity price changes to ensure that the price changes in different time periods are accurately quantified. The final price fluctuation weight curve can reflect the overall trend of electricity price fluctuations and the weight values of each time period, which is used for subsequent sensitive period analysis.

[0057] For example, the electricity price weight from 4 pm to 5 pm in the evening is given a higher weight value (such as 0.8-0.9) due to the higher slope, while the weight in the morning is only 0.2-0.3 due to the lower price fluctuation. The final generated price fluctuation weight curve can accurately indicate the importance of each fluctuation period.

[0058] Cross-time point identification is performed on the price trend change rate and the abnormal factor set to generate a sensitive period set of price jumps.

[0059] The sensitive period set is generated by cross-identification; specifically, by combining the cross-analysis of the price trend change rate and the abnormal factor set, the time points are filtered and marked through the dynamic time warping algorithm (DTW) or a clustering method based on graph structure, to identify the sensitive period of price jumps. The sensitive period set identifies important time points with significant price changes, which is used for operation scheduling strategies.

[0060] For example, in the daily electricity price prediction, the period from 16:00 to 16:30 is identified as a sensitive period due to the price fluctuation intensity and the effect of abnormal factors. This period, combined with holiday factors and seasonal changes, has a significant impact on discharging operations, and it is recommended that energy storage devices release electricity during this period to achieve the highest revenue.

[0061] The price fluctuation weight curve and the sensitive period set are jointly clustered to obtain the daily electricity price fluctuation curve.

[0062] The electricity price fluctuation curve is generated through joint clustering analysis. Specifically, the price fluctuation weight curve and the set of sensitive time periods are input into a joint clustering algorithm (such as the density-based DBSCAN algorithm or K-Means clustering). This classifies and analyzes the characteristics of electricity price fluctuations, mining their relationships and generating a complete electricity price fluctuation curve for the day. This curve not only reflects the overall trend of electricity prices but also identifies important sensitive periods.

[0063] For example, analyzing the daily electricity price weight curve and sensitive time periods ultimately generates a fluctuation curve, showing that prices are relatively stable from morning to noon, but fluctuate significantly from 3:00 to 5:00 p.m., with 4:00 p.m. to 4:30 p.m. as the critical discharge window. This fluctuation curve can be applied to scheduling decisions for PV and energy storage to ensure maximum returns.

[0064] In step S100, the step of calculating the photovoltaic power generation of the next day based on the solar radiation value of the next day specifically includes: The solar radiation value of the next day is combined with the area, inclination parameters and conversion efficiency parameters of the photovoltaic panel equipment using a preset combination calculation formula to obtain the photovoltaic power generation in each time period of the next day.

[0065] The power generation of each period is calculated using the photovoltaic power generation formula. Specifically, based on the predicted solar radiation value of the next day and combined with the physical model of photovoltaic power generation, the power generation of the photovoltaic panel equipment in each period of the next day, P, is calculated using the following combined formula: t : The combination calculation formula is as follows: P t =I t ×A×η×cos(θ); Among them, P t is the power generation of the photovoltaic equipment at time t (unit: watt W); I t is the solar radiation value at time t (unit: watts per square meter, W / m²), output by the solar radiation prediction sub-model; A is the effective area of ​​the photovoltaic panel (unit: m²), directly obtained from the equipment installation parameters. For example, a typical standard photovoltaic panel has an area of ​​1.6 m²; η is the conversion efficiency of the photovoltaic panel (dimensionless), directly read from the photovoltaic panel equipment parameter table. A typical value is 0.18, or 18%; θ is the angle of incidence of the solar radiation (unit: rad). This is obtained by geometrically calculating the angle between the solar radiation and the surface normal of the photovoltaic panel and analytically combining the fixed tilt angle of the photovoltaic panel.

[0066] In each time period, the above formula is used to iteratively calculate all time points t=1,2,…,n to generate the time-of-use power output curve for the next day.

[0067] For example, if a photovoltaic panel device effective area A is 1.6 m², the conversion efficiency η is 0.18, and the solar radiation value It at time t is 500 W / m², the inclination angle of the photovoltaic panel is 30 degrees, and the angle between the solar radiation direction and the normal line (i.e. θ) is 20 degrees, then the photovoltaic power generation at time t is calculated as follows: Substitute the relevant values into the formula: Pt=500•1.6•0.18•cos(20°); According to the trigonometric function: cos(20°)≈0.9397; The calculation result is: Pt=500•1.6•0.18•0.9397=135.11W; Therefore, at time t, the output power of the photovoltaic panel is 135.11 watts. Iterative application of the above calculation to the entire day cycle (such as 24 hours) can obtain the power generation per hour or finer time granularity.

[0068] In step S200, the step of predicting the highest electricity price period of the electricity price based on the electricity price fluctuation curve of the day includes: The electricity price fluctuation curve of the day is standardized to construct a time gradient curve reflecting the electricity price variation amplitude, and the time gradient curve is peak detected to identify the candidate time period of rapid electricity price rise.

[0069] The time gradient curve is constructed and the candidate period is identified by normalization transformation and local peak detection method. Specifically, first, the electricity value of each time node in the obtained daily predicted electricity price fluctuation curve is subjected to Min-Max standardization processing, so that the electricity value is compressed to the interval of 0, 1, thereby unifying the electricity price amplitude scale of different time periods. Then, based on the standardized electricity price sequence, the electricity price change rate (i.e. first order difference) between adjacent time points is calculated to form a time gradient curve. The sliding window detection method is used to scan the gradient curve, and by comparing each time point with its adjacent values, the local peak points are extracted as the candidate time points of rapid electricity price rise.

[0070] For example, if the electricity price sequence of a certain time period after standardization is: [0.25, 0.32, 0.50, 0.78, 0.91, 0.95, 0.96, 0.85], the change rate sequence calculated by the gradient curve is: [+0.07, +0.18, +0.28, +0.13, +0.04, +0.01, -0.11]. When the sliding window is 3, it is identified that the third time point (i.e. 0.50) to the sixth time point (0.95) is a section with significant electricity price growth rate, so it is determined as the candidate time period of rapid electricity price rise.

[0071] The price growth rate and duration corresponding to the extracted candidate time period are extracted, and the price growth rate and duration are normalized and aggregated for analysis to obtain a volatility intensity score matrix at the time period level. The volatility intensity score matrix is input into a preset price risk identification model, combined with holiday factors, power grid load factors, and weather conditions, to output a confidence-weighted price anomaly risk factor and a price rise rate change map.

[0072] Risk indicators and price rise structures are obtained through multi-factor modeling and score matrix analysis. Specifically, for each candidate time period, the price growth rate (i.e., the difference between the end and start of the period) and the duration (i.e., the number of time nodes in the period) are calculated. After normalization, these two indicators form a two-dimensional volatility intensity score matrix. Then the score matrix is input into a price risk identification model, which can use a multi-layer perceptron (MLP) or XGBoost classification model, and introduce holiday factors (such as whether it is a pre-holiday weekday), power grid load factors (such as load forecast curve), and weather conditions (such as high temperature, low temperature, thunderstorm, etc.) as additional input features. The model outputs two key results: one is the price anomaly risk factor (risk level label or score); the other is the price rise rate change map, which is used for subsequent trend analysis.

[0073] For example, the price of a certain candidate time period rises from 0.40 to 0.90, lasting 3 hours, with a growth rate of +0.50. After normalization, the growth rate score is 0.83 and the duration score is 0.60, resulting in a volatility intensity score of 0.72. Combined with the pre-holiday factor (assuming it is the day before the National Day) and high load weather (such as high temperature 35°C), the model predicts that the price anomaly risk factor for this period is "high", and marks the period of continuously rising price slope in the rate change map.

[0074] The price anomaly risk factor is analyzed by threshold classification to identify price high-rise candidate time periods, and the price high-rise candidate time periods are reconstructed with the price rise rate change map to extract the main peak period of rapid price rise and high discharge value.

[0075] The high-value main peak period is extracted by threshold classification and gradient trend backtracking; specifically, the price anomaly risk factor is classified (such as into low, medium, and high levels), and a threshold (for example, the high-risk factor threshold is 0.7) is set to retain only time periods greater than the threshold as price high-rise candidate time periods. Then, through first and second order gradient trend analysis of the price rise rate change map, its price growth curve is reconstructed to identify the main peak area, i.e., the time interval in which the price has a significant upward trend and reaches the maximum discharge benefit in a short period of time.

[0076] For example, a certain time period is assigned a price anomaly risk factor of 0.85, exceeding the high-risk threshold of 0.7; it exhibits in the price rising rate atlas as a sustained price rise from 15:30 to 16:30, and reaches a peak at 16:15. After gradient reconstruction, 16:00-16:30 is identified as the main peak period with discharging value.

[0077] The main peak period is compared with the preset peak price threshold to screen out the highest price period when the price is higher than the average peak level.

[0078] The final highest price period is screened out by comparison with the peak price threshold; specifically, the peak price threshold is set as a reference benchmark (e.g., 1.1 times the average daily highest price of the past month), and the predicted prices of all main peak periods are compared one by one with the threshold. Only the time periods with predicted prices higher than the threshold are retained as the highest price periods. The finally screened time periods are used to guide the power scheduling and preferential power selling strategy of the photovoltaic energy storage device.

[0079] For example, the set peak price threshold is 0.85, and the predicted price reaches 0.92 from 16:00 to 16:30, exceeding the threshold. Therefore, the time period is finally confirmed as the highest price period and is used as the preferential period for discharging of the photovoltaic energy storage device.

[0080] In step S200, the actual power demand range and actual power consumption time period of the user are analyzed according to the user's power consumption habits, which specifically includes: The user's power consumption habits are clustered and modeled to identify different types of daily power consumption behavior patterns, and a typical daily power consumption template of the user is obtained. The typical daily power consumption template of the user is input into a preset user behavior prediction model, and the power consumption of the user in different time periods in the next 24 hours is predicted in combination with weather changes and holiday characteristics.

[0081] The user's power consumption pattern is identified and predicted by a clustering algorithm; specifically, the historical power consumption data of the user is first analyzed using a clustering algorithm (such as K-means or DBSCAN algorithm). The power consumption behavior will be classified into different categories according to the characteristics of power consumption (such as morning peak, evening peak, weekday and weekend differences, etc.). Each category corresponds to a typical "daily power consumption template". For example, the templates for weekdays and weekends differ in power consumption peak time and duration.

[0082] After clustering is completed, the output typical daily power consumption template is input into a preset user behavior prediction model. The model can be a neural network-based regression model (such as LSTM network), which can predict the power consumption trend in the next 24 hours according to the user's power consumption pattern, weather changes, holidays, and other external factors, and output the power consumption in different time periods.

[0083] For example, suppose the user's historical power consumption data is analyzed by clustering, and it is identified that the power consumption peak on weekdays occurs from 7:00 to 9:00 in the morning, and the evening peak is from 17:00 to 19:00, and the power consumption pattern on weekends shows a relatively stable load distribution throughout the day. According to these power consumption patterns, the user's typical daily power consumption template will reflect the power consumption at different time periods (for example, 300W at 7am and 200W at 4pm). Then, combined with meteorological data (such as temperature, humidity, etc.) and holiday information, the prediction model calculates the power consumption of each hour in the next 24 hours and adjusts the expected power consumption demand.

[0084] Segmented statistics and sliding window regression analysis of power consumption at different time periods are performed to obtain the actual power consumption demand range and actual power consumption time period of the user.

[0085] The power consumption demand range and time period are determined by sliding window regression analysis. Specifically, the predicted power consumption in the next 24 hours is divided into multiple time periods by hour, and the power consumption in each time period is counted. Then, using sliding window regression analysis, the power consumption data in each time period is smoothed to eliminate sudden fluctuations and noise, thereby extracting the stable power consumption demand range of the user (such as basic power consumption demand and high-intensity power consumption demand). In addition, sliding window regression can also be used to identify the actual power consumption time period of the user, for example, the user's power consumption demand fluctuates significantly in some time periods, which needs to be specially marked and included in the scheduling decision.

[0086] For example, if the user's power consumption from 8am to 9am is 300W, but in some special cases (such as air conditioner starting or more people using electrical appliances), the power consumption in this period suddenly increases to 500W. Through sliding window regression analysis, this fluctuation will be smoothed, and it is determined that the power consumption in this period is 300W under normal circumstances, and it is identified that 8am to 9am is the "basic power consumption demand" period, and 9am to 11am is the "high-intensity power consumption demand" period. In this case, power scheduling will be performed according to the high-intensity power consumption demand period to ensure that the photovoltaic energy storage can provide sufficient power during these time periods.

[0087] The steps of segmented statistics and sliding window regression analysis of power consumption at different time periods to obtain the actual power consumption demand range and actual power consumption time period of the user include: Segmented statistics of power consumption at different time periods are performed to obtain statistical results, and the mean and deviation are extracted according to the statistical results, and the basic power consumption demand range and high-intensity power consumption demand range are identified according to the mean and deviation.

[0088] The basic electricity demand range and high-intensity electricity demand range are extracted by using time interval statistics. Specifically, the electricity consumption data of different time intervals in the next 24 hours is processed by segmentation, for example, divided by hour or finer minute granularity. The mean and deviation of electricity consumption are calculated for each data set. The mean represents the average load demand of the time interval, and the deviation represents the fluctuation amplitude of electricity consumption. According to the preset threshold, the load in the low deviation range is classified as "basic electricity demand range", and the load in the high deviation range is classified as "high-intensity electricity demand range".

[0089] For example, when statistics of electricity consumption from 7:00 to 9:00 in the morning is collected, it is recorded that the average hourly electricity consumption of the user is 400W, and the deviation data is 30W. This time interval is marked as "basic electricity demand range". In the electricity consumption data from 6:00 to 8:00 in the afternoon, the average electricity consumption is 600W, and the deviation is 100W. It is identified that there is significant load fluctuation in this time interval, which is marked as "high-intensity electricity demand range".

[0090] The basic electricity demand range and high-intensity electricity demand range are fused to eliminate time overlapping intervals, generate a continuous non-overlapping electricity intensity curve, perform time series fitting on the electricity intensity curve, and extract peak period, duration and sudden increase point based on the fitting result. The actual electricity demand range of the user is constructed according to the peak period, duration and sudden increase point.

[0091] The actual electricity demand range is constructed by non-overlapping processing and time series fitting. Specifically, in the first step, the basic electricity demand range and high-intensity electricity demand range are cross-fused. If there is an overlap between basic demand and high-intensity demand in a time interval, it is preferentially marked as high-intensity demand, and all overlapping intervals are eliminated, thereby generating a continuous non-overlapping electricity intensity curve. In the second step, the generated electricity intensity curve is time series fitted, and the data trend is extracted by using polynomial fitting or nonlinear regression method to obtain the peak period of electricity change (maximum load period), duration (load fluctuation duration) and sudden increase point (rapid change point of load). Finally, the actual electricity demand range of the user is constructed according to the fitting result, which provides a basis for the power generation scheduling of photovoltaic energy storage equipment combined with demand key parameters.

[0092] For example, through the fusion process, it is found that the electricity demand from 7:00 to 8:30 in the morning is the basic demand, the demand from 8:30 to 9:00 is the high-intensity demand due to the start of high-load equipment (such as kitchen appliances), and the demand from 9:00 to 9:30 is classified as the basic load. Through the time series fitting analysis, it is extracted that the peak period of the user is from 8:00 to 8:50, with a duration of 50 minutes, and 8:30 is identified as the point of sudden increase. Based on these results, the actual electricity demand range is constructed as follows: the peak period is from 8:00 to 8:50 in the morning, and the reserved load is 550W.

[0093] The electricity consumption of different time periods is analyzed by sliding window regression analysis to extract stable load areas and burst load areas, and a load change trend curve is constructed based on the stable load areas and burst load areas.

[0094] The load pattern is extracted by sliding window regression and a change trend is constructed; specifically, the future electricity demand of the user is processed by sliding window, and multi-window regression analysis (such as linear sliding window regression and regression method based on smooth kernel function) is applied to extract two key areas: stable load area and burst load area. The stable load area refers to a section with low load change amplitude of the user, and the burst load area refers to a section with rapid increase or decrease of electricity demand detected by sliding window regression. According to the two types of areas, a load change trend curve is generated for further identification of the electricity consumption pattern characteristics of the user.

[0095] For example, when processing the data from 7:00 to 9:00 in the morning, it is identified that the load fluctuation range from 7:00 to 8:30 does not exceed 20W (stable load area), while the load from 8:30 to 9:00 increases rapidly from 450W to 650W (burst load area). According to the two areas, a trend curve is generated to show the overall change trend of electricity demand in the morning and mark the fluctuation area.

[0096] The actual electricity consumption time period of the user is obtained by regression fitting and cross analysis of the load change trend curve with the actual electricity demand range.

[0097] The actual electricity consumption time period of the user is determined by regression fitting and cross analysis; specifically, the load change trend curve is cross-analyzed with the actual electricity demand range of the user. This analysis is based on the Bayesian decision model or the support vector regression (SVR) method, and according to the intersection of the peak period, high-intensity demand and stable load area of the user, the actual electricity consumption time period of the user is further confirmed. For example, if a time period belongs to high-intensity demand and also appears in the rapid increase area of the load change trend, it is automatically marked as a priority electricity consumption period.

[0098] For example, through cross-analysis confirmation, the interval from 8:30 to 9:00 is both in the actual electricity demand range of the user and belongs to the rapid fluctuation interval of the load change trend. Finally, the interval is calibrated as the actual electricity consumption time period. Combined with other analysis results, the electricity consumption plan of the user is generated to ensure that the photovoltaic energy storage device can preferentially guarantee power supply in this stage.

[0099] In the embodiment, through user electricity consumption habit analysis based on time series analysis and data mining method, solar radiation prediction, and electricity price fluctuation analysis, comprehensive prediction of the next day photovoltaic power generation and the highest electricity price time period of the day is realized. First, the user's historical electricity consumption data is classified through clustering modeling to identify different types of user electricity consumption patterns, and a preset user behavior prediction model is used in combination with weather changes and holiday characteristics to accurately predict the user's electricity consumption in different time periods in the next 24 hours. Then, combined with the physical parameters of the photovoltaic panel device and the solar radiation value of the next day, the photovoltaic power generation in each time period of the next day is calculated through a preset formula, and a more accurate power generation prediction curve is obtained after correction of the temperature coefficient and the equipment attenuation factor.

[0100] In terms of electricity price prediction, first, the historical electricity price records and real-time feature information are input into the short-term price prediction model to output the hourly electricity price sequence, electricity price fluctuation rate sequence, and abnormal factor set, and the candidate time period of rapid price rise is identified through time gradient analysis and peak detection. Subsequently, the price growth rate, duration, and other parameters are normalized and aggregated, and combined with holiday factors, power grid load factors, and weather conditions, the price abnormal risk factor and price rise rate change map are generated. Through hierarchical threshold analysis and gradient reconstruction, the main peak period with rapid price rise and discharge value is extracted, and finally the highest electricity price time period that meets the electricity price peak threshold is selected.

[0101] At the same time, in order to analyze the actual electricity demand range and time period of the user, the future electricity consumption in different time periods is segmented and counted, and the basic electricity demand range and high-intensity electricity demand range are identified through sliding window regression analysis. At the same time, the electricity intensity curve is generated by eliminating overlapping intervals, and the peak period, duration, and sudden increase point are extracted by time series fitting of the electricity intensity curve, which are used to build the actual electricity demand range of the user. Finally, through cross-analysis of the load change trend curve and the actual electricity demand, the actual electricity consumption time period of the user is determined, realizing the effective combination of photovoltaic energy storage power dispatching and electricity price tracking. The embodiment can significantly improve the economic efficiency and intelligent level of photovoltaic energy storage, providing a strong guarantee for efficient operation of energy and maximization of user benefits.

[0102] Embodiment 3: As Figure 2As shown, the application also provides an AI model-based electricity price tracking and electricity selling system 10 applied to a photovoltaic energy storage device, which comprises an acquisition module 11, a prediction module 12, a comparison module 13, and an electricity selling module 14.

[0103] The acquisition module 11 is mainly used for acquiring user electricity usage habits and predicting solar radiation value of the next day and electricity price fluctuation curve of the current day, and calculating photovoltaic power generation of the next day based on the solar radiation value of the next day.

[0104] The prediction module 12 is mainly used for predicting the highest electricity price time period of the electricity price based on the electricity price fluctuation curve of the current day if the photovoltaic power generation of the next day is higher than the capacity of the photovoltaic energy storage device, and analyzing the actual electricity demand range and actual electricity time period of the user according to the user electricity usage habits.

[0105] The comparison module 13 is mainly used for comparing the highest electricity price time period with the electricity time period in terms of time period, and reserving the electricity amount satisfying the electricity demand range by the photovoltaic energy storage device and selling the remaining electricity amount of the photovoltaic energy storage device to the wall if the highest electricity price time period and the electricity time period are within the preset time interval range.

[0106] The electricity selling module 14 is mainly used for selling all the electricity amount of the photovoltaic energy storage device to the wall if the highest electricity price time period and the electricity time period are not within the preset time interval range.

[0107] In the present embodiment, by designing the electricity price tracking and power selling system 10 based on the AI model, the intelligent integration of electricity price tracking, user demand prediction, power selling strategy formulation and execution of the photovoltaic energy storage device is comprehensively realized. The system is collaboratively working by the acquisition module 11, the prediction module 12, the comparison module 13 and the power selling module 14, which are respectively responsible for different functional tasks. The acquisition module 11 acquires the user's electricity habit through time series analysis and machine learning technology, predicts the key meteorological parameters of the next day in combination with the weather prediction model, and calculates the photovoltaic power generation of the next day by using the physical parameters of the photovoltaic panel and the solar radiation value, to provide basic data for the subsequent modules. The prediction module 12 accurately predicts the peak price period in the electricity price fluctuation curve through the AI model, and simultaneously outputs the actual electricity demand range and actual electricity time period of the user according to the clustering model, to dynamically update the user behavior trend and optimize the matching efficiency. The comparison module 13 compares the predicted highest electricity price period with the actual electricity time period of the user, identifies the overlap interval of the electricity price peak and the user demand by using the cross analysis method, and reasonably allocates the electricity of the photovoltaic energy storage device. The power selling module 14 realizes the efficient power wall selling of the photovoltaic power generation under the condition that the user demand and the electricity price period do not match, decides the power selling period by using the target optimization algorithm, and ensures the maximization of the income. The present embodiment integrates the prediction, energy storage scheduling and dynamic management of power selling into one, realizes the maximization of the economic income of the photovoltaic energy storage device in a systematic way, simultaneously reduces the user's power cost, and provides an efficient and intelligent solution for the new energy system.

[0108] It should be noted that, for the convenience and brevity of description, the specific working processes of the system and each module described above can be referred to the corresponding processes in the foregoing embodiment 1, which will not be described herein.

[0109] Embodiment 4: As shown in Figure 3 The present application also provides an electronic device 20, comprising a memory 21 and a processor 22, the memory 21 stores a computer program capable of running on the processor 22, and the processor 22 implements the electricity price tracking and power selling method based on the AI model of embodiment 1 when executing the computer program.

[0110] In this embodiment, the memory 21 and the processor 22 of the electronic device 20 are used to run a computer program to implement the AI model-based electricity price tracking and power selling method, providing a convenient computing platform and an efficient execution mechanism for photovoltaic energy storage device scheduling and power selling solutions. The memory 21 stores a computer program designed for embodiment 1, which includes core modules such as time series analysis, price prediction model, user behavior prediction model, and device optimization algorithm; when the processor 22 executes the computer program, it can accurately complete the feature extraction and prediction of user electricity usage habits, the calculation of the next day's solar radiation value, the estimation of photovoltaic power generation capacity, and the analysis of electricity price fluctuation trends, and identify the highest electricity price period through time series fitting and hierarchical threshold. The processor 22 can dynamically allocate the power of the photovoltaic energy storage device based on the cross comparison result to meet the user's priority electricity demand, and execute the wall selling strategy during the non-demand period. The electronic device 20 can efficiently coordinate solar energy resources, user electricity behavior, and grid electricity price fluctuations during operation, making the allocation of energy storage devices more accurate, and at the same time, it can respond to weather changes and policy requirements in real time, improve user benefits, and promote the widespread use of green energy.

[0111] Embodiment 5: The present application also provides a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, causes the processor to perform the AI model-based electricity price tracking and power selling method of any one of embodiments 1.

[0112] In this embodiment, a photovoltaic energy storage device-oriented AI model-based electricity price tracking and power selling method is implemented through a computer-readable storage medium. The storage medium stores a computer program executable by a processor, which integrates complex time series prediction algorithms, photovoltaic power generation capacity calculation formulas, electricity price fluctuation prediction models, user demand analysis methods, and other functions into one. When the computer program is executed on the processor, it can efficiently complete the comprehensive analysis of user electricity habits and environmental data, accurately predict user electricity trends and the next day's solar radiation value through time series prediction networks, and dynamically plan photovoltaic energy storage device power generation by time period. The program in the storage medium further extracts price anomaly risk factors and the highest electricity price period using a short-term electricity price prediction model, and generates a wall selling strategy by cross-analyzing the electricity demand range and high-intensity demand areas. This computer-readable storage medium can accurately execute energy storage device allocation and intelligent power selling operations, providing reliable algorithm support for photovoltaic energy storage scheduling systems, significantly improving energy storage power utilization efficiency, optimizing electricity price income, and promoting the efficient use of new energy.

[0113] The structure, proportion, size, etc. shown in the drawings of the specification are only used to cooperate with the disclosed content, to be understood and read by those skilled in the art, and do not have technical significance to limit the implementation conditions of the present application. Any modification of the structure, change of the proportion relationship or adjustment of the size, without affecting the effects and purposes that can be achieved by the present application, should still fall within the scope of the disclosed technology.

[0114] The above-described embodiments are only used to illustrate the technical solutions of the present application, but not to limit them; although the present application has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that the technical solutions recorded in the foregoing embodiments can still be modified, or some technical features can be replaced by equivalents; and these modifications or replacements do not make the essence of the corresponding technical solutions deviate from the spirit and scope of the technical solutions of the embodiments of the present application.

Claims

1. An electricity price tracking and electricity selling method based on an AI model, applied to photovoltaic energy storage equipment, characterized in that: include: Obtaining the user's electricity usage habits, and predicting the next day's solar radiation value and the current day's electricity price fluctuation curve, and calculating the next day's photovoltaic power generation based on the next day's solar radiation value; If the photovoltaic power generation on the next day is higher than the capacity of the photovoltaic energy storage device, the highest electricity price time period is predicted based on the electricity price fluctuation curve of the day, and the actual electricity demand range and actual electricity consumption time period of the user are analyzed according to the user's electricity consumption habits; Comparing the maximum electricity price time period with the electricity consumption time period, if the maximum electricity price time period and the electricity consumption time period are within a preset time interval, the photovoltaic energy storage device reserves the amount of electricity that meets the electricity demand range, and sells the remaining electricity of the photovoltaic energy storage device to the wall; If the maximum electricity price time period and the electricity consumption time period are not within the preset time interval range, all the electricity of the photovoltaic energy storage device will be sold through the wall.

2. The method for tracking and selling electricity prices based on an AI model according to claim 1, characterized in that: The steps of obtaining the user's electricity usage habits and predicting the solar radiation value of the next day and the electricity price fluctuation curve of the current day include: Obtaining historical electricity usage data of users, performing time series analysis on the historical electricity usage data of users, extracting the frequency of electricity usage, the amplitude of change in electricity usage, and periodic characteristics of users in different time periods, so as to construct the electricity usage habits of users; Input the user's electricity usage habits and weather history data into a preset time series prediction network to predict the user's electricity usage trend for the next 24 hours and key meteorological parameters for the next day. Input the key meteorological parameters into a preset solar radiation prediction sub-model to output the solar radiation value for the next day. A feature extraction network is used to extract features from the power grid's historical electricity price records, holiday information, and demand response policies to construct features related to electricity price fluctuations. These features are input into a preset short-term price prediction model to output the electricity price fluctuation curve for the day.

3. The method for tracking and selling electricity prices based on an AI model according to claim 2, characterized in that: The step of inputting the electricity price fluctuation related features into a preset short-term price prediction model and outputting the electricity price fluctuation curve for the day includes: Inputting the electricity price fluctuation-related features into a preset short-term price forecasting model to obtain a predicted hourly electricity price series, an electricity price volatility series, and a set of abnormal factors of high-frequency price fluctuations; Performing a fluctuation window sliding analysis on the hourly electricity price series, extracting local maximum points and slope information based on the analysis results, and constructing the price fluctuation intensity and price trend change rate of each time period using the local maximum points and the slope information; Performing weighted fusion of the price fluctuation intensity and the electricity price volatility sequence to generate a price fluctuation weight curve; Identify the intersection time points of the price trend change rate and the abnormal factor set to generate a set of sensitive time periods for price jumps; The price fluctuation weight curve and the sensitive time period set are jointly clustered and analyzed to obtain the electricity price fluctuation curve for the day.

4. The method for tracking and selling electricity prices based on an AI model according to claim 1, characterized in that: The step of calculating the photovoltaic power generation of the next day based on the solar radiation value of the next day includes: The solar radiation value of the next day is combined with the area, inclination parameters and conversion efficiency parameters of the photovoltaic panel equipment using a preset combined calculation formula to obtain the photovoltaic power generation for each time period of the next day; The combination calculation formula is as follows: P t =I t ×A×η×cos(θ); Among them, the P t is the photovoltaic power generation at time t; I t is the solar radiation value at time t; A is the effective area of ​​the photovoltaic panel; η is the conversion efficiency of the photovoltaic panel; θ is the inclination angle of the photovoltaic panel.

5. The method for tracking and selling electricity prices based on an AI model according to claim 1, characterized in that: The step of predicting the maximum electricity price time period based on the electricity price fluctuation curve of the day includes: Normalizing the electricity price fluctuation curve of the day to construct a time gradient curve reflecting the magnitude of electricity price fluctuations, performing peak detection on the time gradient curve to identify candidate time periods with rapid electricity price increases; Extract the price growth rate and duration corresponding to the candidate time period, perform normalized aggregation analysis on the price growth rate and duration, and obtain a time period-level volatility intensity scoring matrix. Input the volatility intensity scoring matrix into a preset price risk identification model, and output a confidence-weighted price anomaly risk factor and a price increase rate change map. Performing a hierarchical threshold analysis on the price anomaly risk factor to identify candidate time periods of price increases, performing gradient reconstruction on the candidate time periods of price increases and the price increase rate change map to extract the main peak period; The main peak period is compared with a preset electricity price peak threshold to filter out the highest electricity price period when the price is higher than the average peak level.

6. The method for tracking and selling electricity prices based on an AI model according to claim 1, characterized in that: The step of analyzing the user's actual electricity demand range and actual electricity usage time period based on the user's electricity usage habits includes: Perform cluster modeling on the user's electricity usage habits to obtain a typical daily electricity usage template for the user, input the typical daily electricity usage template into a preset user behavior prediction model, and predict the user's electricity usage at different time periods in the next 24 hours; The electricity consumption in different time periods is subjected to segmented statistics and sliding window regression analysis to obtain the user's actual electricity demand range and actual electricity consumption time period.

7. The method for tracking and selling electricity prices based on an AI model according to claim 6, characterized in that: The step of performing segmented statistics and sliding window regression analysis on the electricity consumption in different time periods to obtain the user's actual electricity demand range and actual electricity consumption time period includes: Performing segmented statistics on the electricity consumption in the different time periods to obtain statistical results, extracting a mean and a deviation based on the statistical results, and identifying a basic electricity demand range and a high-intensity electricity demand range based on the mean and the deviation; The basic electricity demand range is merged with the high-intensity electricity demand range to generate a continuous non-overlapping electricity intensity curve, the electricity intensity curve is time-series fitted, and the peak period, duration, and sudden increase point are extracted based on the fitting result. The actual electricity demand range of the user is constructed based on the peak period, duration, and sudden increase point; Performing a sliding window regression analysis on the power consumption in different time periods, extracting a stable load area and a sudden load area, and constructing a load change trend curve based on the stable load area and the sudden load area; The load change trend curve is subjected to regression fitting and cross analysis based on the actual power demand range to obtain the user's actual power consumption time period.

8. An electricity price tracking and electricity sales system based on an AI model, applied to photovoltaic energy storage equipment, characterized in that: include: An acquisition module is used to obtain the user's electricity usage habits, predict the solar radiation value of the next day and the electricity price fluctuation curve of the current day, and calculate the photovoltaic power generation of the next day based on the solar radiation value of the next day; a prediction module configured to, if the photovoltaic power generation of the next day is higher than the capacity of the photovoltaic energy storage device, predict the peak electricity price time period based on the electricity price fluctuation curve of the current day, and analyze the user's actual electricity demand range and actual electricity consumption time period based on the user's electricity consumption habits; a comparison module, configured to compare the maximum electricity price time period with the electricity consumption time period; if the maximum electricity price time period and the electricity consumption time period are within a preset time interval, the photovoltaic energy storage device reserves the amount of electricity that meets the electricity demand range, and sells the remaining electricity of the photovoltaic energy storage device to the power grid; The electricity selling module is used to sell the entire amount of electricity of the photovoltaic energy storage device through a wall-to-wall electricity sale if the maximum electricity price time period and the electricity consumption time period are not within a preset time interval range.

9. An electronic device, characterized in that: It includes a memory and a processor, the memory stores a computer program that can be run on the processor, and when the processor executes the computer program, it implements the electricity price tracking and electricity selling method based on the AI ​​model as described in any one of claims 1 to 7.

10. A computer-readable storage medium, characterized in that A computer program is stored thereon, and when the computer program is executed by a processor, the processor is caused to execute the electricity price tracking and electricity selling method based on the AI ​​model as described in any one of claims 1 to 7.

Images