Substation power utilization system based on green power supply

Abstract

The invention relates to the technical field of substations, and discloses a substation power system based on a green power supply, which comprises an energy router, a substation electronic system, an energy storage subsystem, a distributed photovoltaic panel, a photovoltaic generating capacity prediction module, a direct current load module, an energy management platform and an energy storage battery, the energy router is connected with the station electronic system, the energy storage subsystem, the distributed photovoltaic panel, the photovoltaic generating capacity prediction module, the DC load module, the energy management platform and the energy storage battery. The system is deeply fused with distributed photovoltaic power generation and energy storage technologies, and an energy management router is adopted to intelligently control and manage a distributed photovoltaic power generation and energy storage system, a direct current load and other loads. The application efficiency of green clean energy in the transformer substation is effectively improved, greenhouse gas emission is reduced, and the method is one of important means for meeting the construction of a novel electric power system and has good environmental benefits, economic benefits and social benefits.

Description

Technical Field

[0001] The present invention relates to the technical field of substations, and in particular to a substation power system based on green power supply. Background Art

[0002] The power supply system of a substation is an important part of ensuring the safe and reliable operation of the substation. Once a fault occurs in the power supply system, it will directly or indirectly affect the safe and reliable operation of the substation. In serious cases, it will expand the scope of the accident and lead to serious accidents such as grid disconnection. At present, the update and iteration of new technologies are getting faster and faster, and their applications are becoming more and more extensive, resulting in an increasing load and capacity of substation transformers, but the application of green and clean energy in substations is not satisfactory. Summary of the invention

[0003] The present invention provides a substation power system based on green power, which is deeply integrated with distributed photovoltaic power generation and energy storage technology, and uses energy management routers to intelligently control and manage distributed photovoltaic power generation, energy storage systems, DC loads, and other loads. Effectively improving the application efficiency of green and clean energy in substations and reducing greenhouse gas emissions is one of the important means to meet the construction of new power systems, with good environmental, economic and social benefits.

[0004] The present invention provides a substation power system based on green power supply, including an energy router, a station electronic system, an energy storage subsystem, a distributed photovoltaic panel, a photovoltaic power generation prediction module, a DC load module, an energy management platform and an energy storage battery, wherein the energy router is respectively connected to the station electronic system, the energy storage subsystem, the distributed photovoltaic panel, the photovoltaic power generation prediction module, the DC load module, the energy management platform and the energy storage battery;

[0005] The distributed photovoltaic panels serve as green distributed power sources in the substation, and the energy router is used to realize classified intelligent control and management of station loads, which include DC loads, AC loads, important loads, and general loads. The distributed photovoltaic panels are intelligently controlled to generate electricity, manage various loads, and store energy subsystems by setting strategies to maximize the use of green electricity. The energy storage battery uses DC loads to organically combine new energy power generation, energy storage subsystems, and DC loads. At the same time, an energy manager is used to realize intelligent control and management of new energy power generation, energy storage subsystems, DC loads and other loads. The photovoltaic power generation prediction module predicts photovoltaic power generation based on a preset LSTM neural network model, so as to determine the capacity and type of the energy storage battery in combination with the distributed photovoltaic power generation prediction and the distribution of station variable loads.

[0006] Furthermore, the station electronic system includes Class I loads, Class II loads and Class III loads; the green electricity of the green power substation power system is mainly used to supply power to Class III loads. When the entire station loses power and new energy is generated, the Class III loads are cut off and priority is given to supplying power to Class I and Class II loads of the station power system.

[0007] Furthermore, the distributed photovoltaic panels are arranged on the roofs of the substation main control communication room, comprehensive equipment room, relay room, etc., and the type of the distributed photovoltaic panels is selected according to the substation building type.

[0008] Furthermore, in the photovoltaic power generation prediction module, the preset LSTM neural network model includes an input layer, a hidden layer and an output layer;

[0009] The input layer selects the historical solar radiation intensity, photovoltaic power generation, temperature and humidity data of the similar day training set in the set number of days before the prediction day, and adds the temperature and humidity information of the prediction day, a total of six types of input data;

[0010] The number of nodes in the output layer is set to 1; the number of hidden layers and the number of nodes in each hidden layer are determined using a tentative method and a formula based on historical experience, and the formula is:

[0011]

[0012] Among them, m is the number of hidden layer nodes, n is the number of input layer nodes, l is the number of output layer nodes, and α is a constant between 1 and 10.

[0013] Furthermore, in the input layer, the six types of input data include historical data of solar radiation intensity, humidity, and temperature of historical days, and meteorological data of forecast days;

[0014] 1) Predict the photovoltaic power generation power P m days before the day:

[0015]

[0016] Among them, p in = photovoltaic power generation at point n days before the forecast date, P i = photovoltaic power generation before the forecast date i days ago;

[0017] 2) Predict the solar radiation intensity S m days before the day:

[0018]

[0019] Among them, s in = solar radiation intensity at point n i days before the forecast day, Si = solar radiation intensity i days before the forecast day;

[0020] 3) Predict the temperature T m days before the day:

[0021]

[0022] Among them, t in = the temperature at point n i days before the forecast date, t i = Temperature i days before the forecast date;

[0023] 4) Predict humidity H m days before the day:

[0024]

[0025] Among them, h in = humidity at point n days before the forecast date, H i = humidity before forecast day i;

[0026] 5) Forecast day temperature E:

[0027]

[0028] Among them, e n = the temperature at the nth point of the forecast day;

[0029] 6) Forecast daily humidity R:

[0030]

[0031] Among them, r n = Humidity at the nth point of the forecast day.

[0032] Furthermore, in the input layer, the similar day training set is identified using the K-means clustering algorithm, and the process is as follows:

[0033] S101, inputting historical data of solar radiation intensity for a set number of days before the forecast date, and selecting the sum, maximum value, and average value of each group of data objects as characteristic values;

[0034] S102, setting the number of clusters K to 3, including sunny days, cloudy days, and rainy days, using the Kmeans algorithm to perform cluster analysis, and determining whether it converges;

[0035] S103, if not converged, return to continue using the Kmeans algorithm for cluster analysis and determine whether it converges;

[0036] S104: If convergence has been achieved, three clustering results of sunny day, cloudy day and rainy day are input, and the historical daily data sets corresponding to the clustering results are used as sunny day, cloudy day and rainy day training sets respectively.

[0037] Furthermore, the training process of the preset LSTM neural network model is:

[0038] S201, obtaining a historical similar day training set, and simultaneously initializing the LSTM neural network model parameters, inputting sample data, and setting the number of training times n=0;

[0039] S202, perform operations on each neural network layer, obtain output values, and calculate loss function values. When the loss function value is greater than a set value, determine whether n+1 is greater than a set number of times;

[0040] S203, if n+1 is greater than the set number of times, the training ends and the training model is saved;

[0041] S204, if n+1 is less than or equal to the set number of times, then reversely calculate the gradient of each layer of neurons, calculate the weight and bias correction, and correct the weight w and bias b, return to perform operations on each neural network layer to obtain the output value.

[0042] Furthermore, the DC load module serves as a DC load, including DC load 1, DC load 2, DC load 3, and DC load 4, and the DC load 4 includes an inspection subsystem and a drone.

[0043] Furthermore, in the capacity of the energy router, a DC load is selected according to the renewable energy power generation, energy storage capacity, and station variable load demand, and the capacity of the energy router is determined; when the real-time renewable energy power generation is greater than the station variable load power consumption, a green electricity consumption strategy is set according to the peak power stage and the valley power stage, that is, the station variable system gives priority to absorbing distributed photovoltaic power generation, optimizes the energy router's control over the energy storage subsystem, and intelligently controls the charging and discharging of the energy storage subsystem.

[0044] The beneficial effects of the present invention are:

[0045] 1. Provide a substation power system based on green power supply, innovatively apply distributed photovoltaic power generation, energy storage system, DC load, energy management technology, etc. in the substation power system, realize intelligent control and management of distributed photovoltaic power generation, energy storage system, DC load, and other loads in the substation, and control the flexible conversion and transmission between electric energy with different electrical parameters.

[0046] 2. The present invention reduces the capacity of station transformers, has the characteristics of accepting distributed power sources, optimizing the control scheme of the station power system, and can achieve precise control; the substation power system based on green power has many advantages such as improving power supply reliability and power quality, and assisting the safe operation of large power grids; by adopting energy routers to achieve intelligent control and management, when the entire station loses power and new energy is generated, Class III loads are cut off, and priority is given to Class I and Class II loads of the station power system.

[0047] 3. Rationally utilize the roof space of the substation building without increasing the area occupied by the substation. Use energy routers to optimize the control of the station transformer system, effectively improve the application efficiency of green and clean energy in the substation, and help build a new power system.

[0048] 4. According to the characteristics of renewable energy power generation and the power demand of the station power system, the station power system operation mode is set. The system has the characteristics of unified and efficient power control, diversified power interface, active source and load management, uninterrupted power supply and intelligent human-machine interaction. BRIEF DESCRIPTION OF THE DRAWINGS

[0049] Figure 1 It is a schematic diagram of the structure of the substation power supply system based on green power supply of the present invention.

[0050] Figure 2 It is a schematic diagram of the flow of the K-means clustering algorithm in the present invention.

[0051] Figure 3 Schematic diagram of the training process of the LSTM neural network model in the present invention.

[0052] The realization of the purpose, functional features and advantages of the present invention will be further explained in conjunction with embodiments and with reference to the accompanying drawings. DETAILED DESCRIPTION

[0053] It should be understood that the specific embodiments described herein are only used to explain the present invention, and are not used to limit the present invention.

[0054] The present invention adopts a substation power system based on green power supply, which is based on the installation of distributed photovoltaic components in the substation, takes into account the DC load, sets up an energy storage system, and directly connects the distributed photovoltaic power generation to the energy management router, and the energy router realizes the intelligent control and management of new energy power generation, energy storage system, DC load and other loads. The substation power system based on green power supply is adopted, and the distributed photovoltaic power generation components are arranged according to local conditions without increasing the substation site area and affecting the safe and stable operation of the system. Considering the configuration of the energy storage system, part of the load in the substation is considered as a DC load, and the energy router is used to realize the intelligent control and optimization control of the new energy power generation system, energy storage system, DC load and other loads.

[0055] The present invention innovatively applies renewable energy power generation, energy storage systems, DC loads, energy management technologies, etc. to the substation power system. Energy routers are used to intelligently control and manage renewable energy power generation, energy storage systems, DC loads, other loads, etc., to control the flexible conversion, transmission and routing between different electrical parameters of electric energy, and provide AC / DC hybrid distribution networks with different voltage levels. According to the characteristics of renewable energy power generation and the power demand of the station power system, the station power system operation mode is set. The system has the characteristics of unified and efficient control of electric energy, diversified power interfaces, active source and load management, uninterrupted power supply, and intelligent human-machine interaction.

[0056] like Figure 1 As shown, the substation power system based on green power provided by the present invention includes an energy router, a station electronic system, an energy storage subsystem, a distributed photovoltaic panel, a photovoltaic power generation prediction module, a DC load module, an energy management platform and an energy storage battery, and the energy router is respectively connected to the station electronic system, the energy storage subsystem, the distributed photovoltaic panel, the photovoltaic power generation prediction module, the DC load module, the energy management platform and the energy storage battery;

[0057] The distributed photovoltaic panels are used as green distributed power sources in the substation to increase the application ratio of green power sources for substations. The energy router is used to realize the classified intelligent control and management of station loads, and the distributed photovoltaic power generation, various load management, energy storage system, etc. are intelligently controlled by setting strategies to maximize the use of green electricity; the station loads include DC loads, AC loads, important loads, and general loads; the distributed photovoltaic panels are intelligently controlled by setting strategies to generate electricity, various load management, and energy storage subsystems to maximize the use of green electricity. The energy storage battery uses DC loads to organically combine new energy power generation, energy storage subsystems, and DC loads; at the same time, the energy manager is used to realize the intelligent control and management of new energy power generation, energy storage subsystems, DC loads and other loads; the photovoltaic power generation prediction module predicts the photovoltaic power generation based on the preset LSTM neural network model, so as to combine the distributed photovoltaic power generation prediction and the distribution of station loads to determine the capacity and type of the energy storage battery; optimize the station transformer capacity and station power system wiring, reduce system losses, and improve the operational reliability of the station power system.

[0058] In one embodiment, the station electronic system includes Class I loads, Class II loads and Class III loads, so as to refine the station variable loads, consider the relevant loads of the existing new technology application in Class III loads, and divide the loads according to the importance of the loads to adapt to the refined intelligent control of the loads. The green power of the substation power system of the green power source is mainly used to supply power to Class III loads. When the power of the entire station is lost and new energy is generated, the Class III loads are cut off, and the Class I and Class II loads of the station power system are supplied first.

[0059] In one embodiment, according to the substation layout, the distributed photovoltaic panels are arranged on the roofs of the substation master communication room, comprehensive equipment room, relay room, etc., and the type of the distributed photovoltaic panels is selected according to the substation building type. Distributed photovoltaic power generation is directly connected to reduce the intermediate inverter link, and can directly supply power to the DC load, providing green power application efficiency. According to the layout of distributed photovoltaics and the demand for station variable load, the distributed photovoltaic power generation capacity and station variable capacity are determined.

[0060] In one embodiment, the DC load module serves as a DC load, including DC load 1, DC load 2, DC load 3, and DC load 4, and the DC load 4 includes an inspection subsystem and a drone.

[0061] In one embodiment, the capacity of the energy router is selected based on the renewable energy generation, energy storage capacity, and station variable load demand, and the capacity of the energy router is determined; when the renewable energy real-time generation is greater than the station variable load power consumption, a green power consumption strategy is set according to the peak power stage and valley power stage, that is, the station transformer system preferentially consumes distributed photovoltaic power generation, optimizes the energy router's control over the energy storage subsystem, and intelligently controls the charging and discharging of the energy storage subsystem. The station transformer system preferentially consumes distributed photovoltaic power generation, optimizes the energy router's control over the energy storage system, and intelligently controls the charging and discharging of the energy storage system.

[0062] Distributed photovoltaic panels are arranged on the roofs of the main control and communication room, comprehensive equipment room, relay room, etc. according to local conditions. The type of distributed photovoltaic panels is selected according to the substation building type. The station transformer load is refined and can be divided into DC load, AC load, important load, general load, etc. According to the substation load, the specific load of DC load is determined, and the load capacity is determined. Energy routers are used to realize classified intelligent control and management of station loads. By setting strategies, distributed photovoltaic power generation, various load management, energy storage systems, etc. are intelligently controlled to maximize the use of green electricity; compared with directly connecting distributed photovoltaics to the station power system, fault isolation, power quality control, wiring simplification, station system power optimization management, and system loss reduction can be achieved. At the same time, the station transformer capacity is reduced, the station power system wiring is optimized, the system loss is reduced, and the overall operation reliability of the station power system is improved.

[0063] Distributed photovoltaics are arranged in the substation, energy storage batteries are installed, and DC loads are used to organically combine new energy power generation, energy storage systems, and DC loads; at the same time, energy managers are used to achieve intelligent control and management of new energy power generation, energy storage systems, DC loads and other loads. This station-use substation system composed of new energy power generation, energy storage systems, and DC loads has many advantages, such as accepting distributed power sources, improving power supply reliability and power quality, and assisting the safe operation of large power grids.

[0064] In one embodiment, in the photovoltaic power generation prediction module, the capacity of the energy storage battery is reasonably selected in combination with the distributed photovoltaic power generation prediction and the station variable load distribution; the DC load is reasonably selected and the energy router capacity is determined based on the new energy power generation, energy storage capacity, and station variable load requirements.

[0065] The preset LSTM neural network model in the photovoltaic power generation prediction module includes an input layer, a hidden layer, and an output layer;

[0066] The input layer selects the historical solar radiation intensity, photovoltaic power generation, temperature and humidity data of the similar day training set in the set number of days before the prediction day, and adds the temperature and humidity information of the prediction day, a total of six types of input data; based on these input features, the neural network model sets six nodes in the input layer to ensure that the impact of these key factors on photovoltaic power generation can be fully captured.

[0067] The number of nodes in the output layer is set to 1; the number of hidden layers and the number of nodes in each hidden layer are determined using a tentative method and a formula based on historical experience, and the formula is:

[0068]

[0069] Among them, m is the number of hidden layer nodes, n is the number of input layer nodes, l is the number of output layer nodes, and α is a constant between 1 and 10. For example, a prediction model based on 24 neural networks with 2 hidden layers was established using the above method.

[0070] In the input layer, the six types of input data include historical data of solar radiation intensity, humidity, and temperature of historical days, and meteorological data of forecast days;

[0071] 1) Predict the photovoltaic power generation power P m days before the day:

[0072]

[0073] Among them, p in = photovoltaic power generation at point n days before the forecast date, P i = photovoltaic power generation before the forecast date i days ago;

[0074] 2) Predict the solar radiation intensity S m days before the day:

[0075]

[0076] Among them, s in = solar radiation intensity at point n i days before the forecast day, Si = solar radiation intensity i days before the forecast day;

[0077] 3) Predict the temperature T m days before the day:

[0078]

[0079] Among them, t in = the temperature at point n i days before the forecast date, t i = Temperature i days before the forecast date;

[0080] 4) Predict humidity H m days before the day:

[0081]

[0082] Among them, h in = humidity at point n days before the forecast date, H i = humidity before forecast day i;

[0083] 5) Forecast day temperature E:

[0084]

[0085] Among them, e n = the temperature at the nth point of the forecast day;

[0086] 6) Forecast daily humidity R:

[0087]

[0088] Among them, r n = Humidity at the nth point of the forecast day.

[0089] Under similar weather conditions, the trend of photovoltaic power generation shows a high degree of similarity. Therefore, the historical daily data is divided into three different data sets according to the weather type, and the K-Means clustering algorithm is used to build a screening model that can identify similar days. Since solar radiation intensity is a key factor affecting photovoltaic power generation, solar radiation intensity data is used as the basis in the K-Means clustering analysis process to ensure that the screened similar days have a high similarity in solar radiation intensity with the predicted days. Therefore, in the input layer, the similar day training set is identified using the K-means clustering algorithm, and the process is as follows:

[0090] S101, inputting historical data of solar radiation intensity for a set number of days before the forecast date, and selecting the sum, maximum value, and average value of each group of data objects as characteristic values;

[0091] S102, setting the number of clusters K to 3, including sunny days, cloudy days, and rainy days, using the Kmeans algorithm to perform cluster analysis, and determining whether it converges;

[0092] S103, if not converged, return to continue using the Kmeans algorithm for cluster analysis and determine whether it converges;

[0093] S104: If convergence has been achieved, three clustering results of sunny day, cloudy day and rainy day are input, and the historical daily data sets corresponding to the clustering results are used as sunny day, cloudy day and rainy day training sets respectively.

[0094] The present invention sets the distance loss function MSE as the loss function, the ReLU function as the neuron activation function, the expected error is 0.001, and the training is stopped when the error of the results of three consecutive iterations is less than the expected error; the maximum number of iterations is 2000, and when the number of iterations is greater than this value, the training ends even if the result does not meet the expected error condition; the learning rate is 0.006; the initial weight is set to a random number between [0,1]. The training process of the preset LSTM neural network model is:

[0095] S201, obtaining a historical similar day training set, and simultaneously initializing the LSTM neural network model parameters, inputting sample data, and setting the number of training times n=0;

[0096] S202, perform operations on each neural network layer, obtain output values, and calculate loss function values. When the loss function value is greater than a set value, determine whether n+1 is greater than a set number of times;

[0097] S203, if n+1 is greater than the set number of times, the training ends and the training model is saved;

[0098] S204, if n+1 is less than or equal to the set number of times, then reversely calculate the gradient of each layer of neurons, calculate the weight and bias correction, and correct the weight w and bias b, return to perform operations on each neural network layer to obtain the output value.

[0099] In order to achieve accurate prediction of photovoltaic power generation, the present invention introduces the K-Means clustering algorithm and establishes a similar day screening model, which can effectively identify historical dates with similar climate characteristics to the day to be predicted. Based on the similar day data screened above, a power generation prediction model based on the long short-term memory (LSTM) neural network is further constructed and applied to the actual prediction of photovoltaic power generation.

[0100] The above substation power system based on green power can reduce the capacity of station transformers compared with conventional station power system solutions, and has the characteristics of accepting distributed power sources and optimizing the control solution of station power system, and can achieve precise control; this substation power system based on green power is deeply integrated with distributed photovoltaic power generation, energy storage technology, intelligent energy management, etc. The substation power system based on green power has many advantages such as improving power supply reliability and power quality, assisting the safe operation of large power grids, etc.; by adopting energy routers to achieve intelligent control and management, when the whole station power is lost and new energy is generated, Class III loads are cut off, and Class I and Class II loads of the station power system are given priority. The deep integration of substation transformer system and distributed photovoltaic power generation effectively improves the application efficiency of green and clean energy in substations and reduces greenhouse gas emissions. It is one of the important means to meet the construction of new power systems, and has good environmental, economic and social benefits.

[0101] It should be noted that, in this article, the terms "include", "comprises" or any other variations thereof are intended to cover non-exclusive inclusion, so that a process, device, article or method including a series of elements includes not only those elements, but also includes other elements not explicitly listed, or also includes elements inherent to such process, device, article or method. In the absence of further restrictions, an element defined by the sentence "includes a ..." does not exclude the presence of other identical elements in the process, device, article or method including the element.

[0102] The above description is only a preferred embodiment of the present invention, and does not limit the patent scope of the present invention. Any equivalent structure or equivalent process transformation made by using the contents of the present invention specification and drawings, or directly or indirectly applied in other related technical fields, are also included in the patent protection scope of the present invention.

Claims

1. A substation power system based on green power, characterized in that: It includes an energy router, a station electronic system, an energy storage subsystem, a distributed photovoltaic panel, a photovoltaic power generation prediction module, a DC load module, an energy management platform and an energy storage battery, and the energy router is respectively connected to the station electronic system, the energy storage subsystem, the distributed photovoltaic panel, the photovoltaic power generation prediction module, the DC load module, the energy management platform and the energy storage battery; The distributed photovoltaic panels are used as green distributed power sources in the substation, and the energy router is used to realize classified intelligent control and management of station loads, and the station loads include DC loads, AC loads, important loads, and general loads; By setting strategies, the distributed photovoltaic panel power generation, various load management, and energy storage subsystem are intelligently controlled to maximize the use of green electricity. The energy storage battery adopts DC load to organically combine new energy power generation, energy storage subsystem, and DC load; at the same time, an energy manager is used to realize intelligent control and management of new energy power generation, energy storage subsystem, DC load and other loads; the photovoltaic power generation prediction module predicts the photovoltaic power generation based on the preset LSTM neural network model, so as to combine the distributed photovoltaic power generation prediction and the station variable load distribution to determine the capacity and type of the energy storage battery.

2. The substation power system based on green power supply according to claim 1 is characterized in that: The station electronic system includes Class I loads, Class II loads and Class III loads; the green electricity of the green power substation power system is mainly used to supply power to Class III loads. When the entire station loses power and new energy is generated, the Class III loads are cut off and priority is given to supplying power to Class I and Class II loads of the station power system.

3. The substation power system based on green power supply according to claim 1 is characterized in that: The distributed photovoltaic panels are arranged on the roofs of the substation main control communication room, comprehensive equipment room, relay room, etc., and the type of the distributed photovoltaic panels is selected according to the substation building type.

4. The substation power system based on green power supply according to claim 1 is characterized in that: In the photovoltaic power generation prediction module, the preset LSTM neural network model includes an input layer, a hidden layer and an output layer; The input layer selects the historical solar radiation intensity, photovoltaic power generation, temperature and humidity data of the similar day training set in the set number of days before the prediction day, and adds the temperature and humidity information of the prediction day, a total of six types of input data; The number of nodes in the output layer is set to 1; the number of hidden layers and the number of nodes in each hidden layer are determined using a tentative method and a formula based on historical experience, and the formula is: Among them, m is the number of hidden layer nodes, n is the number of input layer nodes, l is the number of output layer nodes, and α is a constant between 1 and 10.

5. The substation power system based on green power supply according to claim 4 is characterized in that: In the input layer, the six types of input data include historical data of solar radiation intensity, humidity, and temperature of historical days, and meteorological data of forecast days; 1) Predict the photovoltaic power generation power P m days before the day: Among them, p in = photovoltaic power generation at point n days before the forecast date, P i = photovoltaic power generation before the forecast date i days ago; 2) Predict the solar radiation intensity S m days before the day: Among them, s in = solar radiation intensity at point n i days before the forecast date, S i = predicted solar radiation intensity i days ago; 3) Predict the temperature T m days before the day: Among them, t in = the temperature at point n i days before the forecast date, t i = Temperature i days before the forecast date; 4) Predict humidity H m days before the day: Among them, h in = humidity at point n days before the forecast date, H i = humidity before forecast day i; 5) Forecast day temperature E: Among them, e n = the temperature at the nth point of the forecast day; 6) Forecast daily humidity R: Among them, r n = Humidity at the nth point of the forecast day.

6. The substation power system based on green power supply according to claim 5 is characterized in that: In the input layer, the similar day training set is identified using the K-means clustering algorithm, and the process is as follows: S101, inputting historical data of solar radiation intensity for a set number of days before the forecast date, and selecting the sum, maximum value, and average value of each group of data objects as characteristic values; S102, setting the number of clusters K to 3, including sunny days, cloudy days, and rainy days, using the Kmeans algorithm to perform cluster analysis, and determining whether it converges; S103, if not converged, return to continue using the Kmeans algorithm for cluster analysis and determine whether it converges; S104: If convergence has been achieved, three clustering results of sunny day, cloudy day and rainy day are input, and the historical daily data sets corresponding to the clustering results are used as sunny day, cloudy day and rainy day training sets respectively.

7. The substation power system based on green power supply according to claim 6 is characterized in that: The training process of the preset LSTM neural network model is: S201, obtaining a historical similar day training set, and simultaneously initializing the LSTM neural network model parameters, inputting sample data, and setting the number of training times n=0; S202, perform operations on each neural network layer, obtain output values, and calculate loss function values. When the loss function value is greater than a set value, determine whether n+1 is greater than a set number of times; S203, if n+1 is greater than the set number of times, the training ends and the training model is saved; S204, if n+1 is less than or equal to the set number of times, then reversely calculate the gradient of each layer of neurons, calculate the weight and bias correction, and correct the weight w and bias b, return to perform operations on each neural network layer to obtain the output value.

8. The substation power system based on green power supply according to claim 1 is characterized in that: The DC load module serves as a DC load and includes DC load 1, DC load 2, DC load 3, and DC load 4. The DC load 4 includes an inspection subsystem and a drone.

9. The substation power system based on green power supply according to claim 1 is characterized in that: In the capacity of the energy router, a DC load is selected according to the renewable energy power generation, energy storage capacity, and station variable load demand, and the capacity of the energy router is determined; when the real-time renewable energy power generation is greater than the station variable load power consumption, a green electricity consumption strategy is set according to the peak power stage and the valley power stage, that is, the station transformer system gives priority to consuming distributed photovoltaic power generation, optimizes the energy router's control over the energy storage subsystem, and intelligently controls the charging and discharging of the energy storage subsystem.

Images