User side energy storage regulation capability evaluation method, system and device and medium

Abstract

The invention discloses a user side energy storage regulation capability evaluation method, system and device and a medium, and the method comprises the steps: obtaining historical load data, air temperature data and time characteristics, building a load prediction model based on an LSTM neural network, and obtaining a day-ahead load prediction value through cyclic coding; historical photovoltaic output and weather features are obtained, a photovoltaic output prediction model based on an LSTM neural network is established, and a day-ahead photovoltaic output prediction value is obtained through a category feature coding technology and cyclic coding; analyzing trend and seasonal components in the historical load data by adopting a time sequence prediction technology to obtain a maximum demand prediction value; and based on the day-ahead load prediction value, the day-ahead photovoltaic output prediction value and the maximum demand prediction value, combining operation constraint conditions of the energy storage system to obtain a user side energy storage adjustment capability evaluation result. According to the invention, effective evaluation of the user-side energy storage adjustment capability is realized, and a reference basis is provided for the response capability of power grid flexible resources.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of energy management and power system, and particularly relates to a user-side energy storage regulation capacity evaluation method, system, device and medium. BACKGROUND

[0002] With the transformation of energy structure and the rapid development of distributed energy, the role of user-side energy storage system in the power system is increasingly prominent. Energy storage system not only can balance user load and grid demand, but also can reduce electricity cost and improve power supply reliability through participating in demand response, peak clipping and valley filling strategies, and provide flexible resources for the grid. However, there are still many challenges in scientifically guiding the accurate and efficient response of flexible resources: the volatility of distributed photovoltaic power generation and user load increases the complexity of energy storage scheduling; the existing methods are difficult to comprehensively evaluate the regulation potential of user-side energy storage, and cannot provide reliable basis for grid scheduling. In view of the above problems, the existing technology mainly uses a single prediction model or a static evaluation method, which is difficult to consider the dynamic characteristics of photovoltaic output, load demand and energy storage regulation. SUMMARY

[0003] In view of the above existing problems, the present application is proposed.

[0004] Therefore, the present application provides a user-side energy storage regulation capacity evaluation method, system, device and medium to solve the problem that the existing technology cannot accurately and comprehensively evaluate the regulation potential of user-side energy storage system under dynamic distributed photovoltaic output and load demand.

[0005] To solve the above technical problems, the present application provides the following technical solutions:

[0006] In a first aspect, the present application provides a user-side energy storage regulation capacity evaluation method, comprising:

[0007] Obtain historical load data, temperature data and time characteristics, establish a load prediction model based on LSTM neural network, and obtain day-ahead load prediction value through cyclic encoding;

[0008] Obtain historical photovoltaic output and weather characteristics, establish a photovoltaic output prediction model based on LSTM neural network, and obtain day-ahead photovoltaic output prediction value through category feature encoding technology and cyclic encoding;

[0009] Using time series prediction technology, analyze the trend and seasonal components in historical load data to obtain maximum demand prediction value;

[0010] Based on the day-ahead load prediction value, the day-ahead photovoltaic output prediction value and the maximum demand prediction value, and combined with the operation constraints of the energy storage system, the user-side energy storage regulation capacity evaluation result is obtained.

[0011] As a preferred scheme of the user-side energy storage adjustment capability evaluation method, wherein: the day-ahead load prediction value comprises:

[0012] Obtain historical load data, air temperature data and time characteristics;

[0013] Construct input features of the LSTM neural network;

[0014] Normalize the historical load and air temperature data to obtain normalized data;

[0015] Use cyclic encoding to convert the periodic characteristics in the time characteristics into two-dimensional continuous feature values;

[0016] Design the LSTM neural network structure;

[0017] Divide the data set and set the training parameters;

[0018] Use the training set to train and evaluate the load prediction model to obtain the trained load prediction model;

[0019] Input the corresponding features of the current date into the trained load prediction model to obtain the day-ahead load prediction value.

[0020] As a preferred scheme of the user-side energy storage adjustment capability evaluation method, wherein: the day-ahead load prediction value comprises:

[0021] Obtain historical load data, air temperature data and time characteristics;

[0022] Construct input features of the LSTM neural network;

[0023] Normalize the historical load and air temperature data to obtain normalized data;

[0024] Use cyclic encoding to convert the periodic characteristics in the time characteristics into two-dimensional continuous feature values;

[0025] Use category feature encoding technology to process weather types;

[0026] Design the LSTM neural network structure;

[0027] Divide the data set and set the training parameters;

[0028] Use the training set to train and evaluate the load prediction model to obtain the trained load prediction model;

[0029] Input the corresponding features of the current date into the trained load prediction model to obtain the day-ahead load prediction value.

[0030] As a preferred scheme of the user-side energy storage adjustment capacity evaluation method, the maximum demand prediction value is obtained by:

[0031] Collecting historical maximum load data and determining the length of the seasonal cycle;

[0032] Setting an initial level component, a trend component and a seasonal component, and selecting a smoothing parameter;

[0033] Using time series prediction technology, the level component, the trend component and the seasonal component are updated step by step to obtain updated level component, trend component and seasonal component;

[0034] Based on the updated level component, trend component and seasonal component, the maximum demand prediction value is obtained.

[0035] The beneficial effects of the preferred technical scheme are that through the systematic time series prediction process, the trend and seasonal characteristics in the load data are accurately captured, thereby realizing high-precision prediction of the maximum demand of the user, and providing reliable basis for energy storage operation strategy optimization and electricity cost control.

[0036] As a preferred scheme of the user-side energy storage adjustment capacity evaluation method, the maximum demand prediction value is obtained by:

[0037] Based on the day-ahead load prediction result and the day-ahead photovoltaic output prediction result, the net load value at each time is calculated;

[0038] Based on the net load value, the upper limit of the energy storage charging power and the upper limit of the energy storage discharging power are calculated respectively;

[0039] By setting a charging and discharging time period set, the energy state of the energy storage is simulated to obtain the energy storage capacity at the current time;

[0040] According to the energy storage capacity at the current time, the user energy storage adjustment capacity evaluation result is obtained.

[0041] The beneficial effects of the preferred technical scheme are that by combining the load and photovoltaic output prediction results and the energy storage operation constraints, the energy storage charging and discharging process is dynamically simulated, and the adjustable capacity of the user-side energy storage at different time points is accurately evaluated, thereby providing scientific decision basis for optimizing energy storage scheduling and reducing demand electricity cost.

[0042] As a preferred scheme of the user-side energy storage adjustment capacity evaluation method, the maximum demand prediction value is obtained by:

[0043]

[0044] wherein, is the user gateway meter value at time t. is the upper limit of the energy storage charging power at time t; is the maximum demand of the user in the month; is the rated charging power of the energy storage.

[0045] As a preferred scheme of the user-side energy storage regulation capacity evaluation method provided by the application, the upper limit of the energy storage discharging power is expressed as:

[0046]

[0047] wherein, is the upper limit of the energy storage discharging power at time t; is the rated discharging power of the energy storage.

[0048] In a second aspect, the application provides a user-side energy storage regulation capacity evaluation system, comprising:

[0049] a load prediction module, configured to obtain historical load data, air temperature data and time characteristics, establish a load prediction model based on an LSTM neural network, and obtain a day-ahead load prediction value through loop encoding;

[0050] a photovoltaic output prediction module, configured to obtain historical photovoltaic output, month and weather characteristics, establish a photovoltaic output prediction model based on an LSTM neural network, and obtain a day-ahead photovoltaic output prediction value through category feature encoding technology and loop encoding;

[0051] a maximum demand prediction module, configured to analyze trend and seasonal components in historical load data by using time series prediction technology, and obtain a maximum demand prediction value;

[0052] a user-side energy storage regulation capacity evaluation module, configured to obtain a user-side energy storage regulation capacity evaluation result based on the day-ahead load prediction value, the day-ahead photovoltaic output prediction value and the maximum demand prediction value, and in combination with operation constraints of the energy storage system.

[0053] In a third aspect, the application provides an electronic device, comprising:

[0054] a memory, configured to store a program;

[0055] a processor, configured to execute the computer executable instruction, which, when executed by the processor, implements the steps of the user-side energy storage regulation capacity evaluation method.

[0056] In a fourth aspect, the application provides a computer readable storage medium, comprising: the program, when executed by the processor, implements the steps of the user-side energy storage regulation capacity evaluation method.

[0057] The application has the beneficial effects that the application utilizes the LSTM neural network to combine load and photovoltaic output prediction, improves the accuracy of load and photovoltaic output prediction, and provides accurate basic data for subsequent energy storage strategy optimization; the trend and seasonal components in historical load data are analyzed by using the cubic exponential smoothing method to predict the maximum demand of the user in the month. This method effectively captures the change rule of the load, ensures that the energy storage system will not cause the load to exceed the maximum demand due to charging when running; the energy state change of the energy storage is simulated by setting the charging and discharging time period set, and the actual adjustment capacity of the energy storage at different time points is finally evaluated. This method can dynamically adjust the operation strategy of the energy storage, maximize the adjustment potential of the energy storage system, and improve the electricity economy of the user and the flexibility of the power grid. BRIEF DESCRIPTION OF DRAWINGS

[0058] In order to more clearly illustrate the technical solutions of the embodiments of the application, the drawings needed in the embodiment description will be briefly introduced. Obviously, the drawings in the following description are only some embodiments of the application, and other drawings can be obtained by those skilled in the art without creative labor. Among them:

[0059] Figure 1 The basic flowchart of a user-side energy storage adjustment capacity evaluation method provided for an embodiment of the application is shown in the figure.

[0060] Figure 2 The user energy storage adjustment capacity evaluation result graph of a user-side energy storage adjustment capacity evaluation method provided for an embodiment of the application is shown in the figure. DETAILED DESCRIPTION

[0061] In order to make the above-mentioned purposes, features and advantages of the application more obvious and easy to understand, the specific embodiments of the application will be described in detail below with reference to the drawings of the specification. Obviously, the described embodiments are part of the embodiments of the application, rather than all the embodiments. Based on the embodiments in the application, all other embodiments obtained by those skilled in the art without creative labor should be within the protection scope of the application.

[0062] Embodiment 1, refer to Figure 1 For an embodiment of the application, a user-side energy storage adjustment capacity evaluation method is provided, which comprises:

[0063] S100: Obtain historical load data, temperature data and time characteristics, establish a load prediction model based on the LSTM neural network, and obtain the day-ahead load prediction value through loop coding;

[0064] S200: Obtain historical photovoltaic output and weather characteristics, establish a photovoltaic output prediction model based on an LSTM neural network, obtain a day-ahead photovoltaic output prediction value through category feature coding technology and cycle coding;

[0065] S300: Adopt time series prediction technology to analyze the trend and seasonal components in historical load data to obtain a maximum demand prediction value;

[0066] S400: Based on the day-ahead load prediction value, the day-ahead photovoltaic output prediction value, and the maximum demand prediction value, and combined with the operation constraints of the energy storage system, obtain the user-side energy storage adjustment capacity evaluation result.

[0067] It should be noted that the user-side energy storage system faces multiple challenges during operation, including the intermittency and volatility of photovoltaic output, the uncertainty of load demand, and the limited capacity and charge / discharge rate limit of the energy storage system. These factors make the optimization scheduling of the energy storage system complex, especially in ensuring that the maximum demand is not exceeded to avoid additional electricity charges while maximizing the economic benefits of the energy storage system. Therefore, it is particularly important to explore a user-side energy storage adjustment capacity evaluation method that takes into account the dynamic characteristics of photovoltaic output, load demand, and energy storage adjustment.

[0068] Therefore, in view of the problem that the existing technology cannot accurately and comprehensively evaluate the adjustment potential of the user-side energy storage system under dynamically changing distributed photovoltaic output and load demand, through the steps of S100-S400, through accurate load and photovoltaic output prediction, maximum demand prediction, and dynamic simulation under energy storage operation constraints, efficient evaluation of the adjustment capacity of the user-side energy storage system is achieved. Specifically, the LSTM neural network is used to process historical data and combined with the advanced cubic exponential smoothing method to ensure high accuracy of load and photovoltaic output prediction; by setting a reasonable energy storage charging and discharging strategy, the increase in electricity charges due to energy storage charging and exceeding the maximum demand of the load is avoided. This method not only optimizes the operation efficiency of the energy storage system, improves the electricity economy of the user, but also enhances the flexibility and stability of the power grid, providing strong technical support for realizing intelligent energy management and sustainable development.

[0069] Embodiment 2, refer to Figure 2 For an embodiment of the present application, a user-side energy storage adjustment capacity evaluation method is provided based on the previous embodiment, comprising:

[0070] In the embodiments of the present application, the day-ahead load prediction value obtained in step S100 comprises:

[0071] Obtain historical load data, temperature data, and time characteristics;

[0072] Construct input features of the LSTM neural network;

[0073] The historical load and air temperature data are normalized to obtain normalized data;

[0074] The periodic characteristics in the time characteristics are converted into two-dimensional continuous characteristic values by using cyclic encoding;

[0075] An LSTM neural network structure is designed;

[0076] The data set is divided and the training parameters are set;

[0077] The training set is used to train and evaluate the load prediction model, and a trained load prediction model is obtained;

[0078] The corresponding features of the current date are input into the trained load prediction model to obtain the day-ahead load prediction value.

[0079] In the embodiments of the present application, the air temperature data in step S100 includes daily average air temperature, and the time characteristics include month, week type, holiday flag.

[0080] In the embodiments of the present application, the input features of the LSTM neural network constructed in step S100 include the load data of the previous day at every 15-minute interval (96 time points in total), the daily average air temperature of the previous day, the daily average air temperature of the prediction day, and the month, week type and holiday features of the prediction day.

[0081] In the embodiments of the present application, a 0 / 1 variable is used to represent the holiday feature, and if the prediction day is a holiday, the encoding is 1, otherwise it is 0.

[0082] In the embodiments of the present application, the cyclic encoding is used to process the periodic category features (such as month and week), and is converted into two characteristic values, and the expression is as follows:

[0083]

[0084] Wherein, x and y are two characteristic values of cyclic encoding. When the week is represented by cyclic encoding, i takes 1-7, and max(i)=7; when the month is represented by cyclic encoding, i takes 1-12, and max(i)=12. For example, the cyclic encoding of Monday is [0.78, 0.62], and the cyclic encoding of January is [0.5, 0.87].

[0085] In the embodiment of the present application, constructing the LSTM neural network model structure comprises setting the time step of the LSTM neural network to 1 and the model input feature dimension to 1x103 (including 96 load data + 1 previous day temperature + 1 predicted day temperature + cyclically encoded month x 2 + week x 2 + holiday flag x 1) based on the above feature values; the output layer is 96 neurons, corresponding to the load prediction value of 24 hours (every 15 minutes) of the predicted day; the network model structure comprises 1 input layer (103 neurons), 2 LSTM hidden layers (each 103 neurons), and 1 output layer (96 neurons).

[0086] In the embodiment of the present application, dividing the data set and setting the training parameters comprises dividing the data set into a training set (70%), a validation set (20%), and a test set (10%); using the mean square error to calculate the model loss error; setting the learning rate to 0.001; introducing the early stopping strategy, and stopping training when the validation set loss does not decrease for 5 consecutive rounds.

[0087] In the embodiment of the present application, training and evaluation comprises training the LSTM model using the training set; monitoring the model generalization ability on the validation set to prevent overfitting; and finally evaluating the model performance on the test set to ensure the prediction accuracy.

[0088] In the embodiment of the present application, obtaining the day-ahead load prediction value comprises inputting the related features of the current date into the trained LSTM model; and outputting the load prediction value of the next day (24 hours, 15 minutes interval); obtaining a high-precision and dynamically updated day-ahead load prediction result for subsequent energy storage adjustment capacity evaluation.

[0089] It should be noted that S100 uses the LSTM neural network to establish a day-ahead 24-hour load prediction model by considering historical load levels, temperature levels, and time characteristics and other factors. This step provides a key input for the subsequent steps - the day-ahead load prediction value, i.e., the accurate load prediction value every 15 minutes in the next day. The result of this step directly affects how the energy storage system plans charging and discharging based on the expected load.

[0090] In the embodiment of the present application, obtaining the day-ahead photovoltaic output prediction value in step S200 comprises:

[0091] Obtaining historical photovoltaic output and weather characteristics;

[0092] Constructing LSTM neural network input features;

[0093] Normalizing photovoltaic output data and temperature data to obtain normalized data;

[0094] Using cyclic encoding to convert the periodic characteristics of the month into two-dimensional continuous feature values;

[0095] The weather type is processed by using a category feature encoding technology;

[0096] An LSTM neural network structure is designed;

[0097] The data set is divided and the training parameters are set;

[0098] The training set is used to train and evaluate the load prediction model, and a trained photovoltaic output prediction model is obtained;

[0099] The corresponding features of the current date are input into the trained photovoltaic output prediction model, and a day-ahead photovoltaic output prediction value is obtained.

[0100] In an optional embodiment, the category feature encoding technology in step S300 can be One-Hot encoding, and can also be label encoding; and can also be binary encoding.

[0101] In an optional embodiment, the One-Hot encoding includes converting the weather type (sunny, cloudy, overcast, rainy) into a 4-bit binary vector, each type corresponding to a unique vector representation (for example, sunny is [1, 0, 0, 0], cloudy is [0, 1, 0, 0], etc.), which is used as an input feature of the LSTM neural network for photovoltaic output prediction.

[0102] In an optional embodiment, the label encoding includes converting the weather type (sunny, cloudy, overcast, rainy) into a unique integer label (for example, sunny is 0, cloudy is 1, overcast is 2, and rainy is 3), and directly using these integer labels as input features of the LSTM neural network for photovoltaic output prediction.

[0103] In an optional embodiment, the binary encoding includes converting the weather type (sunny, cloudy, overcast, rainy) into corresponding binary values (for example, sunny is 00, cloudy is 01, overcast is 10, and rainy is 11), and using these binary values as input features of the LSTM neural network for photovoltaic output prediction.

[0104] It should be noted that One-Hot encoding maps each weather type to an independent vector, avoiding the implicit relationships caused by numerical order in label encoding (such as the model mistakenly believing that labels with larger values ​​are more important than those with smaller values), thus preventing the model from learning incorrect associations. Compared with binary encoding, One-Hot encoding eliminates the concept of numerical distance between different categories, ensuring that each category is equidistant and independent in the input space. This helps improve the LSTM neural network's understanding and processing ability of weather type features, thereby improving the accuracy of photovoltaic power output prediction. One-Hot encoding not only enhances the robustness and stability of the model but also avoids potential numerical misleading, enabling the model to more accurately capture the complex relationship between photovoltaic power output and weather type. Therefore, this invention uses One-Hot encoding to process weather type features, which has significant advantages over label encoding and binary encoding.

[0105] In this application embodiment, the weather characteristics include the month of the forecast day, the weather type (sunny, cloudy, overcast, rainy) and the temperature and precipitation every 15 minutes;

[0106] In this embodiment of the application, the input features for constructing the LSTM neural network include photovoltaic power output data every 15 minutes of the previous day (a total of 96 points), month information for the forecast day, weather type information, and temperature and precipitation forecast values ​​every 15 minutes of the forecast day.

[0107] In this embodiment, the photovoltaic output data and temperature data are normalized; the months are cyclically encoded; the weather type is encoded using a 4-bit One-Hot code, where 1000 represents sunny, 0100 represents cloudy, 0010 represents overcast, and 0001 represents rainy; precipitation is represented by different values, where 0 represents no precipitation, 0.25 represents light rain, 0.5 represents light rain, 0.75 represents moderate rain, and 1 represents heavy rain.

[0108] In this embodiment, the LSTM neural network structure is designed as follows: based on the aforementioned feature values, the time step of the LSTM neural network is set to 1, and the model input parameters are 1×294 feature vectors. The model output parameters are the 24-hour photovoltaic power output prediction values ​​for the prediction day, with a time interval of 15 minutes. The model consists of one input layer, two hidden layers, and one output layer. The input layer has 294 neurons, corresponding to 294 input feature values; the number of neurons in the hidden layer is empirically set to 294; and the output layer has 96 neurons, corresponding to the 24-hour (15-minute interval) photovoltaic power output prediction values. The mean squared error is used to calculate the model loss error.

[0109] In the embodiment of the present application, the data set is divided and the training parameters are set, including selecting historical photovoltaic output data and historical weather data as the data set, wherein 70% is divided as the training set, 20% as the validation set, and 10% as the test set. The learning rate is set to 0.001, and the early stopping strategy is adopted, and the training is stopped when the validation set loss does not decrease for 5 consecutive rounds.

[0110] It should be noted that S200 uses the LSTM neural network again to establish a day-ahead 24-hour photovoltaic output prediction model by considering the correlation factors such as historical photovoltaic output, month, and weather characteristics. The model is used to predict the photovoltaic output in the next day. Due to the uncertainty of photovoltaic output, this prediction is crucial for understanding the renewable energy supply situation and directly affects when the energy storage system is charged (when the photovoltaic output is high) and when it is discharged (when the photovoltaic output is low but the load demand is high).

[0111] In the embodiment of the present application, the maximum demand prediction value obtained in step S300 includes:

[0112] Collecting historical maximum load data and determining the length of the seasonal cycle;

[0113] Setting the initial level component, trend component, and seasonal component, and selecting the smoothing parameter;

[0114] Using time series prediction technology to gradually update the level component, trend component, and seasonal component to obtain updated level component, trend component, and seasonal component;

[0115] Based on the updated level component, trend component, and seasonal component, the maximum demand prediction value is obtained.

[0116] In an optional implementation, the time series prediction technology in step S300 can be a cubic exponential smoothing method, an ARIMA model, or a Prophet model.

[0117] In an optional implementation, the cubic exponential smoothing method includes initializing the level component, trend component, and seasonal component, updating these three components through iteration using historical load data, and finally predicting the maximum demand value of a future month based on the updated level, trend, and seasonal components.

[0118] In an optional implementation, the ARIMA model includes performing stationarity test and necessary difference processing on historical load data, determining the parameters (p, d, q) of the ARIMA model, training the ARIMA model using these parameters, verifying the accuracy of the model through historical data, and predicting the maximum demand of a future month using the trained model.

[0119] In an alternative embodiment, the Prophet model includes collecting and organizing historical load data, including dates and load values, configuring basic parameters of the Prophet model such as growth trend type and seasonal components, training the Prophet model using historical load data, and evaluating model performance through cross-validation or test set, and using the trained model to predict the maximum demand of a certain month in the future.

[0120] It should be noted that the advantages of the present application using the triple exponential smoothing method over the ARIMA model and the Prophet model are its simplicity and efficiency, and it is particularly suitable for time series data with obvious trend and seasonality. The triple exponential smoothing method can quickly and intuitively capture the periodicity and trend changes in the load data by directly updating the level, trend and seasonal components, reducing the complex parameter adjustment process. In contrast, the ARIMA model requires data stationarity test and difference processing, and the parameter selection is more complex, increasing the difficulty of model tuning; while the Prophet model performs well in handling multiple seasonality and holiday effects, but its configuration is relatively complex, relying on more prior knowledge and feature engineering. The triple exponential smoothing method provides accurate demand prediction results without excessive parameter adjustment, and is particularly suitable for real-time monitoring and prediction of maximum demand in user-side energy storage systems. In addition, the triple exponential smoothing method has higher transparency, which is easy to understand and explain the prediction results, thereby providing a reliable basis for energy storage scheduling strategies.

[0121] In the embodiments of the present application, in order to avoid the user load exceeding the maximum demand of the month (i.e. the maximum load) due to the charging of energy storage when the user-side energy storage participates in demand response, thereby increasing the demand charge, the triple exponential smoothing method is used to predict the maximum demand of the month. The prediction formula of the triple exponential smoothing method is as follows:

[0122] l m =α(d m -s m-k )+(1-α)(l m-1 +t m-1 )

[0123] t m =β(l m -l m-1 )+(1-β)t m-1

[0124] s m =γ(d m -l m )+(1-γ)s m-k

[0125]

[0126] wherein, l m is the horizontal component of the maximum load in the mth month; d m is the actual observation value of the maximum load in the mth month; s m is the seasonal component of the maximum load in the mth month; t m is the trend component of the maximum load in the mth month; a,

[0127] b and g are respectively the horizontal smoothing parameter, the trend smoothing parameter and the seasonal smoothing parameter, and all take values in the range [0, 1]; k is the length of the seasonal period, and for monthly data, k = 12; is the predicted value of the maximum load in the mth + h month, h is the prediction step, such as predicting the next month, h = 1; h mod k represents the remainder after h is divided by k.

[0128] It should be noted that the predicted value of the maximum load is the predicted value of the maximum demand, which is encapsulated in the parameter used in this step.

[0129] It should be noted that S300 uses the cubic exponential smoothing method to predict the maximum load of the user in the current month, i.e., the maximum demand. Understanding the maximum demand helps to avoid additional cost increases due to energy storage charging, and also provides an important boundary condition for the operation strategy of the energy storage system. This step ensures that the energy storage scheduling strategy does not lead to an increase in the total electricity cost of the user.

[0130] In an optional embodiment, the user-side energy storage adjustment capability evaluation result obtained in step S400 includes:

[0131] Based on the day-ahead load prediction result and the day-ahead photovoltaic output prediction result, the net load value at each time is calculated;

[0132] Based on the net load value, the upper limit of the energy storage charging power and the upper limit of the energy storage discharging power are calculated respectively;

[0133] By setting a set of charging and discharging time periods, the energy state of the energy storage is simulated to obtain the energy storage capacity at the current time;

[0134] According to the energy storage capacity at the current time, the user energy storage adjustment capability evaluation result is obtained.

[0135] In the embodiments of the present application, the maximum demand of the user in the current month is taken as a constraint boundary to calculate the upper limit of the charging power of the energy storage at time t:

[0136]

[0137] wherein, is the user gateway meter value at time t, i.e., the net load of the user; and respectively represent the user predicted load and the predicted output value of photovoltaic at time t; is the upper limit of the energy storage charging power at time t; is the maximum demand of the user in the month; is the rated charging power of the energy storage.

[0138] In the embodiments of the present application, the energy storage only supports user-side charging and discharging, and cannot send electricity back to the power grid. Therefore, when the user net load is less than 0 (i.e., the photovoltaic output is greater than the user load), the energy storage system is prohibited from discharging. Therefore, the upper limit of the discharging power of the energy storage at time t can be represented as:

[0139]

[0140] wherein, is the upper limit of the discharging power of the energy storage at time t; is the rated discharging power of the energy storage.

[0141] In the embodiments of the present application, the regulation capacity of the energy storage is embodied as the total amount of stored electric energy accumulated in the charging period. Therefore, based on the user-set charging and discharging time of the energy storage, the energy state change of the energy storage can be simulated to obtain the energy storage capacity at the current time, which is represented as:

[0142]

[0143] wherein, E t is the energy storage capacity at time t; Δt is the unit time; T cha and T dis are the user-set charging and discharging time sets of the energy storage. Therefore, the regulation capacity of the energy storage at time t can be represented as:

[0144]

[0145] It should be noted that S400 provides a load prediction value, a photovoltaic output prediction value, and a maximum demand prediction value based on the previous steps, and combines the operating constraints (such as charging and discharging power limits) of the energy storage system to propose an energy storage regulation capacity evaluation method. This step is essentially to integrate all the information above, simulate the actual operation of the energy storage system, evaluate its regulation capacity under different conditions, and thus provide a reference basis for the response capacity of the grid flexibility resource.

[0146] In the embodiment of the present application, a certain user has a storage device with a capacity of 1000kW·h and a charging and discharging power of 500kW. The fixed charging time of the storage is set to 0:00-8:00 and 11:00-13:00, and the discharging time is set to 8:00-11:00 and 15:00-17:00. The initial capacity of the storage is 500kW·h. The storage adjustment capacity evaluation result is as shown in Table 1. Figure 2

[0147] Embodiment 3, this is an embodiment of the present application, which is different from the first embodiment, a user side storage adjustment capacity evaluation system is provided.

[0148] It should be noted that the technical scheme of the user side storage adjustment capacity evaluation system belongs to the same concept as the technical scheme of the user side storage adjustment capacity evaluation method described above. The details of the technical scheme of the user side storage adjustment capacity evaluation system in the present embodiment are not described in detail, and can be referred to the description of the technical scheme of the user side storage adjustment capacity evaluation method described above.

[0149] The user side storage adjustment capacity evaluation system in the present embodiment comprises:

[0150] A load prediction module is configured to obtain historical load data, temperature data and time characteristics, establish a load prediction model based on an LSTM neural network, and obtain a day-ahead load prediction value through recurrent encoding.

[0151] A photovoltaic output prediction module is configured to obtain historical photovoltaic output, month and weather characteristics, establish a photovoltaic output prediction model based on an LSTM neural network, and obtain a day-ahead photovoltaic output prediction value through category feature encoding technology and recurrent encoding.

[0152] A maximum demand prediction module is configured to analyze the trend and seasonal components in the historical load data by using time series prediction technology, and obtain a maximum demand prediction value.

[0153] A user side storage adjustment capacity evaluation module is configured to obtain a user side storage adjustment capacity evaluation result based on the day-ahead load prediction value, the day-ahead photovoltaic output prediction value and the maximum demand prediction value, and in combination with the operation constraints of the storage system.

[0154] The present embodiment also provides an electronic device suitable for the case of a user side storage adjustment capacity evaluation method, comprising:

[0155] A memory and a processor; the memory is configured to store computer executable instructions, and the processor is configured to execute the computer executable instructions to implement the user side storage adjustment capacity evaluation method as proposed in the above embodiment.

[0156] ​The embodiment also provides a storage medium, which stores a computer program, and the computer program is executed by a processor to implement the method for evaluating the energy storage adjustment capability of a user side.

[0157] The storage medium proposed by the embodiment belongs to the same inventive concept as the method for evaluating the energy storage adjustment capability of a user side proposed by the above embodiment, and the technical details not described in the embodiment can be referred to the above embodiment, and the embodiment has the same beneficial effects as the above embodiment.

[0158] Through the above description of the embodiments, those skilled in the art can clearly understand that the present application can be realized by means of software and necessary universal hardware, and of course can also be realized by hardware, but in many cases the former is a better embodiment. Based on such understanding, the technical solutions of the present application or the part that contributes to the prior art can be embodied in the form of a software product, which can be stored in a computer readable storage medium, such as a floppy disk, a read-only memory (ROM), a random access memory (RAM), a FLASH memory, a hard disk, or an optical disc, etc., and includes a number of instructions to make a computer device (which can be a personal computer, a server, or a network device, etc.) execute the methods of various embodiments of the present application.

[0159] It should be noted that the above embodiments are only used to illustrate the technical solutions of the present application but not limit the present application, and although the present application has been described in detail with reference to the preferred embodiments, those skilled in the art should understand that the technical solutions of the present application can be modified or replaced equivalently without departing from the spirit and scope of the present application, and all of them should be covered in the scope of the claims of the present application.

Claims

1. A method for evaluating user-side energy storage regulation capability, characterized in that, The method comprises the following steps: Obtain historical load data, temperature data and time characteristics, and establish a load prediction model based on LSTM neural network, and obtain day-ahead load prediction value through cycle coding; Obtain historical photovoltaic output and weather characteristics, and establish a photovoltaic output prediction model based on LSTM neural network, and obtain day-ahead photovoltaic output prediction value through category feature coding technology and cycle coding; Use time series prediction technology to analyze the trend and seasonal components in the historical load data to obtain the maximum demand prediction value; Based on the day-ahead load prediction value, the day-ahead photovoltaic output prediction value and the maximum demand prediction value, and combined with the operation constraints of the energy storage system, the user-side energy storage adjustment capability evaluation result is obtained. 2.The user-side energy storage regulation capability evaluation method of claim 1, wherein: The day-ahead load prediction value comprises the following steps: Obtain historical load data, temperature data and time characteristics; Construct the input features of the LSTM neural network; Normalize the historical load and temperature data to obtain normalized data; Use cycle coding to convert the periodic characteristics in the time characteristics into two-dimensional continuous feature values; Design the structure of the LSTM neural network; Divide the data set and set the training parameters; Use the training set to train and evaluate the load prediction model to obtain the trained load prediction model; Input the corresponding features of the current date into the trained load prediction model to obtain the day-ahead load prediction value. 3.The user-side energy storage regulation capability evaluation method according to claim 1 or 2, characterized in that: The day-ahead photovoltaic output prediction value comprises the following steps: Obtain historical photovoltaic output and weather characteristics; Construct the input features of the LSTM neural network; Normalize the photovoltaic output data and temperature data to obtain normalized data; Use cycle coding to convert the periodic characteristics of the month into two-dimensional continuous feature values; Use category feature coding technology to process the weather type; Design the structure of the LSTM neural network; Divide the data set and set the training parameters; Use the training set to train and evaluate the load prediction model to obtain the trained photovoltaic output prediction model; Input the corresponding features of the current date into the trained photovoltaic output prediction model to obtain the day-ahead photovoltaic output prediction value.

4. The user-side energy storage regulation capability assessment method of claim 3, wherein: The maximum demand prediction value comprises the following steps: Collect historical maximum load data and determine the seasonal cycle length; Set the initial level component, trend component and seasonal component, and select the smoothing parameter; Use time series prediction technology to update the level component, trend component and seasonal component step by step to obtain the updated level component, trend component and seasonal component; Based on the updated level component, trend component and seasonal component, the maximum demand prediction value is obtained.

5. The user-side energy storage regulation capability assessment method of claim 4, wherein: The user-side energy storage adjustment capability evaluation result comprises the following steps: Based on the day-ahead load prediction result and the day-ahead photovoltaic output prediction result, the net load value at each time is calculated; Based on the net load value, the upper limit of the energy storage charging power and the upper limit of the energy storage discharging power are calculated respectively; By setting the charging and discharging time period set, the energy state of the energy storage is simulated to obtain the energy storage capacity at the current time; According to the energy storage capacity at the current time, the user energy storage adjustment capability evaluation result is obtained.

6. The user-side energy storage regulation capability assessment method of claim 5, wherein: The upper limit of the energy storage charging power is represented as: wherein, is the user gateway meter value at time t; is the upper limit of the energy storage charging power at time t; is the user's maximum demand for the month; is the rated charging power of the energy storage.

7. The user-side energy storage regulation capability assessment method of claim 6, wherein: The upper limit of the energy storage discharging power is represented as: wherein, is the upper limit of the discharge power of the energy storage at time t; is the rated discharge power of the energy storage.

8. A user-side energy storage regulation capability assessment system applying the method of any one of claims 1-7, characterized in that, The method comprises the following steps: The load prediction module is configured to acquire historical load data, air temperature data and time characteristics, establish a load prediction model based on an LSTM neural network, and obtain a day-ahead load prediction value through recurrent encoding. The photovoltaic output prediction module is configured to acquire historical photovoltaic output, month and weather characteristics, establish a photovoltaic output prediction model based on an LSTM neural network, and obtain a day-ahead photovoltaic output prediction value through category feature encoding technology and recurrent encoding. The maximum demand prediction module is configured to analyze trend and seasonal components in historical load data by using time series prediction technology, and obtain a maximum demand prediction value. The user-side energy storage adjustment capability evaluation module is configured to obtain a user-side energy storage adjustment capability evaluation result based on the day-ahead load prediction value, the day-ahead photovoltaic output prediction value and the maximum demand prediction value, and in combination with operation constraint conditions of the energy storage system.

9. An electronic device, comprising: The program is executed by the processor to implement the steps of the method according to any one of claims 1-7. The program is executed by the processor to implement the steps of the method according to any one of claims 1-7. ​ 10. A computer-readable storage medium storing a program, characterized in that, ​

Images