Energy scheduling method and platform system of photovoltaic energy storage system

Abstract

The invention relates to an energy scheduling method of a photovoltaic energy storage system and a platform system. The method comprises the following steps: acquiring working state data of the photovoltaic energy storage system at a plurality of historical time points through a sensor; based on the working state data corresponding to each historical time point, training to obtain a state prediction model; on the basis of the state prediction model, according to the working state data of the previous time point, predicting a prediction scheduling demand of the current time point; and in response to the energy scheduling demand signal at the current time point, determining an energy scheduling instruction corresponding to the photovoltaic energy storage system according to the predicted scheduling demand.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of energy storage system control, in particular to an energy scheduling method and platform system of a photovoltaic energy storage system. BACKGROUND

[0002] With the rapid development of renewable energy technology, photovoltaic power generation as a new type of energy has been widely used in the world. However, photovoltaic energy storage systems still face many challenges in actual operation, especially under complex environmental and variable load conditions. How to achieve efficient energy scheduling and optimal management remains a difficult point in technology development.

[0003] The existing solutions lack time series analysis of historical load data and dynamic prediction of equipment working characteristics, making it difficult to accurately identify scheduling requirements. Common rule mechanisms or empirical formulas cannot adapt to diversified energy scheduling scenarios, resulting in insufficient precision of scheduling strategy generation, which easily causes equipment delay response, limiting the intelligent level and operation safety of photovoltaic energy storage systems. Therefore, the existing technology has defects and needs to be solved. SUMMARY

[0004] The primary purpose of the present application is to solve at least one of the above problems and provide an energy scheduling method and platform system of a photovoltaic energy storage system.

[0005] To achieve the various purposes of the present application, the present application adopts the following technical solutions: An energy scheduling method of a photovoltaic energy storage system is provided to adapt to one of the purposes of the present application, comprising the following steps: Obtain working state data of the photovoltaic energy storage system at a plurality of historical time points through a sensor; Train a state prediction model based on the working state data corresponding to each of the historical time points; Based on the state prediction model, predict the predicted scheduling requirement of the current time point according to the working state data of the previous time point; In response to the energy scheduling requirement signal of the current time point, determine the energy scheduling instruction corresponding to the photovoltaic energy storage system according to the predicted scheduling requirement.

[0006] In an optional embodiment, training a state prediction model based on the working state data corresponding to each of the historical time points comprises: Determine the state characteristic parameter corresponding to each of the working state data based on a state identification rule; Sort each of the state characteristic parameters from early to late based on the historical time points to obtain a parameter time series; The parameter time sequence is taken as a training data set to train a preset long short-term memory neural network, to obtain a state prediction model.

[0007] In an optional embodiment, the working state data includes illumination data, photovoltaic power generation power, energy storage battery voltage, energy storage battery current, energy storage battery temperature, energy storage battery state, and electricity price data; the state characteristic parameters include illumination intensity, photovoltaic power generation amount, energy storage amount, load demand, real-time electricity price, load type, and electricity price type; the power grid load type includes power generation amount greater than the sum of energy storage amount demand and user load demand, power generation amount less than user load demand, and power generation amount greater than or equal to user load demand; and the electricity price type includes off-peak electricity price and peak electricity price.

[0008] In an optional embodiment, the parameter time sequence is taken as a training data set to train a preset long short-term memory neural network, to obtain a state prediction model, including: For any two adjacent state characteristic parameters in the parameter time sequence, a difference value between the illumination intensities of the two adjacent state characteristic parameters is calculated, to obtain an illumination difference value; A difference value between the photovoltaic power generation amounts of the two adjacent state characteristic parameters is calculated, to obtain a power generation difference value; A difference value between the energy storage amounts of the two adjacent state characteristic parameters is calculated, to obtain an energy storage difference value; A difference value between the load demands of the two adjacent state characteristic parameters is calculated, to obtain a load difference value; A difference value between the real-time electricity prices of the two adjacent state characteristic parameters is calculated, to obtain an electricity price difference value; A weighted sum of the illumination difference value, the power generation difference value, the energy storage difference value, the load difference value, and the electricity price difference value is calculated, to obtain a change characteristic parameter; A ratio of the change characteristic parameter to a time difference value between the two state characteristic parameters and a historical time point corresponding to the two state characteristic parameters is calculated, to obtain a correlation characteristic between the two state characteristic parameters; The parameter time sequence and each corresponding correlation characteristic are spliced as a fusion characteristic, which is input to a preset long short-term memory neural network for training, to obtain a state prediction model.

[0009] In an optional embodiment, the loss function of the long short-term memory neural network is set as a product of a first cross entropy, a second cross entropy and a third cross entropy, the first cross entropy is a difference between the predicted output and a label corresponding to the parameter time series; the second cross entropy is a difference between the output correlation feature and an actual correlation feature; the third cross entropy is a difference between the predicted output and a real label corresponding to the actual scheduling demand of the output correlation feature; the output correlation feature is a correlation feature between the predicted output and the predicted output at the previous time step; and the actual correlation feature is an actual correlation feature between corresponding parameters of the parameter time series.

[0010] In an optional embodiment, the sensor includes an illumination sensor, a temperature sensor, a battery state sensor and a system state sensor, the illumination sensor is used to detect the ambient light intensity, the temperature sensor is used to detect the working temperature of the photovoltaic energy storage system, the battery state sensor is used to detect the voltage, current and state of charge of the energy storage battery in the energy storage system, and the system state sensor is used to identify the load type and the electricity price type according to the working state data.

[0011] In an optional embodiment, based on the state prediction model, the predicted scheduling demand at the current time point is predicted according to the working state data at the previous time point, including: obtaining the working state data at at least two time points before the current time point to determine the historical state data; calculating the correlation feature between the two historical state data; inputting the historical state data and the corresponding correlation feature into the state prediction model to obtain the output predicted scheduling demand at the current time point.

[0012] In an optional embodiment, according to the predicted scheduling demand, the energy scheduling instruction corresponding to the photovoltaic energy storage system is determined, including: obtaining the current working state data corresponding to the energy scheduling demand signal and the corresponding scheduling demand parameter; determining the reference state feature corresponding to the predicted scheduling demand in the preset database corresponding to the scheduling demand parameter; calculating the feature similarity between the state feature of the current working state data and the reference state feature; judging whether the feature similarity is greater than a preset similarity threshold; if not, determining the energy scheduling instruction corresponding to the photovoltaic energy storage system through the actual user interface operation of the scheduling platform; if yes, determining the energy scheduling instruction of the photovoltaic energy storage system based on the preset corresponding relationship between the working state and the scheduling strategy according to the predicted scheduling demand.

[0013] In an optional embodiment, the scheduling strategy comprises: When the light is sufficient, the photovoltaic system generates electricity to charge the energy storage battery after balancing the load power supply, and then supplies the grid; When the light is insufficient, the photovoltaic system generates electricity to supply the user load, and the energy storage system discharges to supply the user load, and if the electricity generation and discharge cannot meet the user load demand, then the grid electricity is purchased as a second priority; When the grid power is off, the photovoltaic energy storage system switches to an off-grid mode, and supplies the user load through photovoltaic power generation and energy storage discharge; When the valley electricity price is low, the energy storage system is charged by purchasing low-cost electricity from the grid; When the peak electricity price is high, the energy storage system supplies the user load and discharges to supply the grid for high-price electricity sales.

[0014] On the other hand, an energy scheduling platform system of a photovoltaic energy storage system is provided to adapt to one of the purposes of the present application, comprising: An acquisition module is configured to acquire working state data of the photovoltaic energy storage system at a plurality of historical time points through a sensor; A training module is configured to train a state prediction model based on the working state data corresponding to each of the historical time points; A prediction module is configured to predict a predicted scheduling demand at a current time point based on the state prediction model and the working state data at a previous time point; A determination module is configured to determine an energy scheduling instruction corresponding to the photovoltaic energy storage system according to the predicted scheduling demand in response to an energy scheduling demand signal at the current time point.

[0015] In another aspect, an energy scheduling device of a photovoltaic energy storage system is provided to adapt to one of the purposes of the present application, comprising a central processing unit and a memory, wherein the central processing unit is configured to call and run a computer program stored in the memory to execute the steps of the energy scheduling method of the photovoltaic energy storage system as disclosed in the present application.

[0016] In yet another aspect, a computer readable storage medium is provided to adapt to one of the purposes of the present application, wherein the computer readable storage medium stores computer executable instructions for causing a computer to execute the energy scheduling method of the photovoltaic energy storage system as disclosed in any one of the first aspect of the present application.

[0017] The technical solution of the present application has multiple advantages, including but not limited to the following aspects: This application uses sensors to acquire historical time-point data of the photovoltaic energy storage system to train a state prediction model. Based on the data from the previous time point, it predicts the scheduling needs at the current time point and further determines the energy scheduling instructions. This enables the generation of accurate energy scheduling instructions based on the time-series analysis of the working state, thereby improving the accuracy and operational safety of intelligent energy scheduling control of the photovoltaic energy storage system. Attached Figure Description

[0018] The above and / or additional aspects and advantages of this application will become apparent and readily understood from the following description of the embodiments taken in conjunction with the accompanying drawings, wherein: Figure 1 This is a schematic diagram of the architecture between the photovoltaic energy storage system and the energy dispatch platform system of this application; Figure 2 for Figure 1 A schematic diagram of another architecture; Figure 3 A flowchart illustrating one embodiment of the energy dispatching method for the photovoltaic energy storage system of this application; Figure 4 This is a schematic diagram of the energy dispatch platform system of the photovoltaic energy storage system used in this application. Detailed Implementation

[0019] The technical solution of this application is applicable to the field of energy storage system control technology, and is particularly applicable to the energy dispatching scenario of photovoltaic energy storage system. In this context, the technical solution of this application can be applied in a typical energy dispatching platform system architecture.

[0020] like Figure 1 The embodiment of this disclosure provides an energy dispatching platform system 40 for a photovoltaic energy storage system connected to a photovoltaic energy storage system 10. The photovoltaic energy storage system 10 includes a photovoltaic system 20 for power generation and an energy storage system 30 for storing electricity. The photovoltaic system 20 includes several photovoltaic arrays, and the energy storage system 30 includes several energy storage batteries. The energy dispatching platform system 40 receives the working status data uploaded by the photovoltaic energy storage system and issues a control for real-time control of the energy flow between the energy storage system, the photovoltaic system, the power grid, and user loads, thereby ensuring the efficient operation of the photovoltaic energy storage system.

[0021] The energy dispatching platform system 40 of the photovoltaic energy storage system is used to perform the following steps: acquiring the working status data of the photovoltaic energy storage system at multiple historical time points through sensors; training a state prediction model based on the working status data corresponding to each historical time point; predicting the predicted dispatching demand at the current time point based on the state prediction model and the working status data at the previous time point; and determining the energy dispatching instruction corresponding to the photovoltaic energy storage system according to the predicted dispatching demand in response to the energy dispatching demand signal at the current time point.

[0022] It needs to be explained that the energy scheduling platform system 40 in the embodiment of the present disclosure can be a local device, a remote device, or a combination of a local device and a remote device.

[0023] Please refer to Figure 2 In another embodiment of the present disclosure, the architecture between the photovoltaic energy storage system and the energy scheduling platform system, wherein the energy scheduling platform system 40 of the photovoltaic energy storage system comprises: an edge computing platform 50, configured to acquire working state data of the photovoltaic energy storage system, predict the energy scheduling demand corresponding to the working state at the target time point by using the state prediction model, and determine the energy scheduling instruction of the photovoltaic energy storage system according to the predicted scheduling demand; an energy scheduling database 60, configured to store the working state data, the state characteristic parameters, the state identification rules, the associated features of the historical state data, the scheduling demand parameters, the reference state characteristics, the preset threshold, the state prediction model, and the training data corresponding to the state prediction model, and provide the energy scheduling data and the model to the edge computing platform 50; a cloud management platform 70, configured to visually display the data in the edge computing platform and the energy scheduling database 60, issue control instructions to all devices in the photovoltaic energy storage system, and manually control the energy scheduling of the photovoltaic energy storage system according to artificial experience and prior experience.

[0024] The energy scheduling method of the photovoltaic energy storage system of the present application can be programmed as a computer program product, deployed in a server for running and implementation, for example, in the exemplary application scenario of the present application, it can be deployed and implemented in the server of the energy scheduling platform, thereby the interface opened after the computer program product is run can be accessed, the process of the computer program product is interacted with the graphical user interface to execute the method.

[0025] The specific embodiments below can be combined with each other, and the same or similar concepts or processes will not be described in detail in some embodiments. The embodiments of the present application will be described below with reference to the accompanying drawings.

[0026] Please refer to Figure 3 The present application discloses an energy scheduling method of a photovoltaic energy storage system, comprising the following steps: Step 1100, acquiring working state data of the photovoltaic energy storage system at a plurality of historical time points by a sensor; Optionally, the sensor can be a temperature sensor, a current sensor, a voltage sensor, a power sensor, a humidity sensor, or a combination of multiple sensors, which is not limited by the present application.

[0027] Optionally, the working state data acquisition process can be realized based on real-time collection, periodic sampling, triggered collection, or low-power collection, which is not limited by the present application.

[0028] Step 2100, training a state prediction model based on the working state data corresponding to each of the historical time points; Optionally, the state prediction model can be a long short-term memory network model, a Transformer model, a recurrent neural network model, or a combination of at least two models, which is not limited by the present application.

[0029] Optionally, the training process of the state prediction model can be based on self-supervised learning, semi-supervised learning, or reinforcement learning, which is not limited by the present application.

[0030] Optionally, the state prediction model can be optimized in combination with sensor characteristics, device characteristics of the photovoltaic energy storage system, price time sequence variation characteristics, and load time sequence variation characteristics, which is not limited by the present application.

[0031] Step 3100, predicting a predicted scheduling demand at a current time point based on the working state data at a previous time point based on the state prediction model; Optionally, the predicted scheduling demand can include an energy scheduling demand during power grid outage, an energy scheduling demand during insufficient light, an energy scheduling demand during sufficient light, an energy storage charging demand during low load, an energy storage discharging demand during high load, a power purchase demand during low valley price, and a power selling demand during peak price, which is not limited by the present application.

[0032] Optionally, the prediction process of the state prediction model can be based on real-time inference, batch working state data processing, or incremental calculation, which is not limited by the present application.

[0033] Step 4100, determining an energy scheduling instruction corresponding to the photovoltaic energy storage system according to the predicted scheduling demand in response to an energy scheduling demand signal at the current time point.

[0034] Optionally, the energy scheduling demand signal can be a sensor trigger signal, a local management interaction signal, a cloud management interaction signal, or other external input signals, which is not limited by the present application.

[0035] Optionally, the energy scheduling instruction can include energy storage charging, energy storage discharging, power grid power purchase, power grid power selling, preferential supply to user load, preferential energy storage, preferential power selling, etc., which is not limited by the present application. As can be seen, the above embodiments train a state prediction model by acquiring working state data of the photovoltaic energy storage system at different historical times through sensors, predict the current scheduling demand based on the previous time data, and further determine the energy scheduling instruction, thereby realizing precise energy scheduling instruction generation based on working state data time sequence analysis, and improving the accuracy and operation safety of intelligent operation of the photovoltaic energy storage system energy scheduling.

[0036] In the specific implementation, the step of training a state prediction model based on the working state data corresponding to each of the historical time points comprises: determining a state feature parameter corresponding to each of the working state data based on a state recognition rule; sorting each of the state feature parameters from early to late based on the historical time points to obtain a parameter time sequence; training a preset long short-term memory neural network using the parameter time sequence as a training data set to obtain the state prediction model.

[0037] Optionally, the long short-term memory neural network can be a single-layer network, a multi-layer network, a bidirectional network, or an embedded attention mechanism network, and the application does not make any limitation.

[0038] As can be seen, through the above specific embodiments, the state feature parameters are extracted based on the state recognition rule and sorted by time to generate the parameter time sequence, and the long short-term memory network is trained to obtain the state prediction model, so that on the basis of the accurate energy regulation instruction generation, the accuracy of state prediction is improved through time sequence feature extraction and model training, reliable scheduling demand support is provided for the energy scheduling instruction, the risk of misjudgment of the energy scheduling instruction caused by the deviation of the state prediction from the actual scheduling demand is reduced, and the operation safety of the energy scheduling of the photovoltaic energy storage system is improved.

[0039] In the specific implementation, the working state data comprises illumination data, photovoltaic power generation power, energy storage battery voltage, energy storage battery current, energy storage battery temperature, energy storage battery state, and electricity price data; the state feature parameter comprises illumination intensity, photovoltaic power generation capacity, energy storage capacity, load demand, real-time electricity price, load type, and electricity price type; the power grid load type comprises a condition that the power generation capacity is greater than the sum of the energy storage capacity demand and the user load demand, a condition that the power generation capacity is less than the user load demand, and a condition that the power generation capacity is greater than or equal to the user load demand; and the electricity price type comprises a low-valley electricity price and a peak-peak electricity price.

[0040] As can be seen, through the above specific embodiments, the content of the state feature parameter is limited to comprehensively represent the working state feature of the energy flow process between the energy storage system and the photovoltaic system, the power grid, and the user load, thereby assisting in realizing the accurate energy scheduling instruction generation based on the state time sequence analysis and improving the accuracy and operation safety of the intelligent operation of the energy scheduling of the photovoltaic energy storage system.

[0041] In the specific implementation, the step of training a preset long short-term memory neural network using the parameter time sequence as a training data set to obtain the state prediction model comprises: For any two adjacent state characteristic parameters in the parameter time sequence, calculate the difference between the light intensity of the two adjacent state characteristic parameters to obtain a light difference value; Calculate the difference between the photovoltaic power generation of the two adjacent state characteristic parameters to obtain a power generation difference value; Calculate the difference between the energy storage power of the two adjacent state characteristic parameters to obtain a storage power difference value; Calculate the difference between the load demand of the two adjacent state characteristic parameters to obtain a load difference value; Calculate the difference between the real-time electricity price of the two adjacent state characteristic parameters to obtain an electricity price difference value; Calculate the weighted sum of the light difference value, the power generation difference value, the storage power difference value, the load difference value and the electricity price difference value to obtain a change characteristic parameter; Calculate the ratio of the change characteristic parameter to the time difference between the two historical time points corresponding to the two state characteristic parameters to obtain a correlation characteristic between the two state characteristic parameters; Splice the parameter time sequence and the corresponding each correlation characteristic into a fusion feature, and input it into a preset long short-term memory neural network for training to obtain a state prediction model.

[0042] Optionally, the weighted sum can be calculated by using fixed weight, dynamic weight or adaptive weight, which is not limited in the present application.

[0043] Optionally, the fusion feature can be a multi-dimensional feature vector, a sequence feature matrix or a mixed feature set, which is not limited in the present application.

[0044] Optionally, the splicing process of the parameter time sequence and the correlation characteristic can be realized based on feature splicing, data alignment or dimension transformation, which is not limited in the present application.

[0045] Optionally, the training process of the long short-term memory network can be optimized by combining data enhancement, sequence length optimization or model regularization, which is not limited in the present application.

[0046] As can be seen, through the above specific embodiments, the change characteristic parameter is obtained by calculating the weighted sum of the power generation, storage power, load and electricity price difference values of adjacent state characteristic parameters, and the correlation characteristic is generated by combining the time difference, and the long short-term memory network is trained by splicing the fusion feature, so as to improve the model's ability to capture state changes through multi-dimensional feature analysis and correlation feature fusion on the basis of precise state prediction model training, provide more accurate prediction support for the generation of energy scheduling instructions, and reduce the risk of state prediction error.

[0047] In the specific implementation, in the above steps, the loss function of the long short-term memory neural network is set as the product of the first cross entropy, the second cross entropy and the third cross entropy, the first cross entropy is the difference between the predicted output and the label corresponding to the parameter time sequence; the second cross entropy is the difference between the output correlation feature and the actual correlation feature; the third cross entropy is the difference between the predicted output and the real label corresponding to the actual scheduling demand of the output correlation feature; the output correlation feature is the correlation feature between the predicted output and the predicted output at the previous time step; and the actual correlation feature is the actual correlation feature between the corresponding parameters of the parameter time sequence.

[0048] Optionally, the loss function of the long short-term memory neural network can be realized by combining a regularization term, weight adjustment or dynamic loss balancing, which is not limited in the present application.

[0049] Optionally, the first cross entropy, the second cross entropy and the third cross entropy can be classification cross entropy, regression cross entropy or mixed cross entropy, which is not limited in the present application.

[0050] As can be seen, through the above specific embodiments, by designing the loss function of the long short-term memory network as the product of the first cross entropy, the second cross entropy and the third cross entropy, the prediction accuracy of the model for the state sequence and the correlation feature is improved based on the basic entropy of the precise state prediction model training, which provides more reliable model support for the generation of energy scheduling instructions and reduces the prediction deviation risk caused by insufficient model training.

[0051] In the specific implementation, in the above steps, the sensor includes an illumination sensor, a temperature sensor, a battery state sensor and a system state sensor, the illumination sensor is used to detect the ambient light intensity, the temperature sensor is used to detect the working temperature of the photovoltaic energy storage system, the battery state sensor is used to detect the voltage, current and state of charge of the energy storage battery in the energy storage system, and the system state sensor is used to identify the load type and the electricity price type according to the working state data.

[0052] Optionally, the battery state sensor and the system state sensor can be a combination of sensors and control elements, wherein the control element can be an embedded microcontroller, a single-chip microcomputer or other special-purpose chip, and the control element is built-in with a detection algorithm to identify the battery state and the system state at the current time point according to the working state data.

[0053] As can be seen, through the above specific embodiments, the comprehensiveness and accuracy of state feature extraction are improved through multi-sensor cooperation and system state analysis, high-quality data input is provided for the state prediction model, and the recognition error risk caused by insufficient single-sensor data is reduced.

[0054] In the implementation, the step of predicting the predicted scheduling demand at the current time point based on the state prediction model according to the working state data at the previous time point comprises: obtaining working state data at at least two previous time points of the current time point as historical state data; calculating the correlation features between the two historical state data; inputting the historical state data and the corresponding correlation features into the state prediction model to obtain the output predicted scheduling demand at the current time point.

[0055] It can be seen that, by obtaining the historical state data at at least two previous time points of the current time point and calculating the correlation features, the working state features of the equipment in the photovoltaic energy storage system are predicted by inputting the state prediction model, thereby improving the timing accuracy of the predicted scheduling demand based on the accurate energy scheduling instruction generation, providing accurate scheduling demand basis for the energy scheduling instruction, reducing the prediction error risk caused by insufficient timing data, and improving the operation safety of the photovoltaic energy storage system energy scheduling.

[0056] In the implementation, the step of determining the energy scheduling instruction corresponding to the photovoltaic energy storage system according to the predicted scheduling demand comprises: obtaining the current working state data corresponding to the energy scheduling demand signal and the corresponding scheduling demand parameter; determining the reference state features corresponding to the predicted scheduling demand in the preset database corresponding to the scheduling demand parameter; calculating the feature similarity between the state features of the current working state data and the reference state features; judging whether the feature similarity is greater than a preset similarity threshold; if not, determining the energy scheduling instruction corresponding to the photovoltaic energy storage system through actual user interface operation of the scheduling platform; if yes, determining the energy scheduling instruction of the photovoltaic energy storage system based on the predicted scheduling demand and the preset corresponding relationship between the working state and the scheduling strategy.

[0057] Optionally, the feature similarity can be cosine similarity, Euclidean distance, Jaccard system or dynamic time warping distance, which is not limited in the present application.

[0058] Optionally, the calculation process of the feature similarity can be realized based on vector comparison, statistical analysis or feature matching, which is not limited in the present application.

[0059] Optionally, the current working state data can include power generation data, power storage data, load data, or electricity price data, which are not limited in the present application.

[0060] Optionally, the scheduling demand parameter can include a power generation parameter, a power discharge parameter, a power storage parameter, a power purchase parameter, a power sale parameter, or an energy scheduling priority parameter, which are not limited in the present application.

[0061] Optionally, the process of obtaining the current working state data and the corresponding scheduling demand parameter can be based on direct collection by a sensor, processing by a controller, database calling, or data interface transmission, which are not limited in the present application.

[0062] Optionally, the process of operating the actual user interface of the scheduling platform can be local input operation or cloud input operation, which are not limited in the present application.

[0063] Optionally, the corresponding relationship between the working state and the scheduling strategy can be represented by a mapping table, a rule base, a data-driven model, or a conditional logic table, which are not limited in the present application.

[0064] As can be seen, through the above specific embodiments, by comparing the similarity of the state features of the current working state data and the reference state features and determining the energy scheduling instruction based on the corresponding relationship between the predicted scheduling demand and the scheduling strategy, the accuracy and reliability of the energy scheduling instruction generation are improved through feature similarity verification and relationship mapping on the basis of accurately generating the energy scheduling instruction, the risk of energy scheduling instruction errors caused by misjudgment of energy scheduling demand signals or differences in working states is reduced, and the operation safety of the photovoltaic energy storage system is improved.

[0065] In specific implementation, in the above steps, the scheduling strategy includes: When the light is sufficient, the photovoltaic system generates electricity to charge the energy storage battery after balancing the load power supply, and then supplies the power grid; When the light is insufficient, the photovoltaic system generates electricity to supply the user load, the energy storage system discharges to supply the user load, and if the power generation and discharge do not meet the user load demand, the power grid is then selected to purchase electricity; When the power grid is powered off, the photovoltaic energy storage system switches to an off-grid mode, and supplies the user load through photovoltaic power generation and energy storage discharge; When the electricity price is low, the energy storage system is charged by purchasing electricity from the power grid at a low price; When the electricity price is high, the energy storage system supplies the user load and discharges to supply the power grid for high-price power sale.

[0066] It can be seen that by the above embodiments, the conditions of the scheduling strategy are defined, and the energy scheduling requirements under the conditions of illumination, power grid state, load state and electricity price are comprehensively considered, which can meet the energy scheduling under different conditions, thereby realizing the data time sequence analysis based on different working state conditions, accurately generating the energy scheduling instruction, and improving the accuracy and operation safety of intelligent operation of the photovoltaic energy storage system.

[0067] The unique technical advantage of the present application is that the working state data of the photovoltaic energy storage system at the historical time points is acquired by the sensor to train the state prediction model, the scheduling requirement at the current time point is predicted based on the data at the previous time point, and the energy scheduling instruction is further determined, so as to realize the accurate energy scheduling instruction generation based on the working state time sequence analysis, improve the accuracy and operation safety of intelligent regulation and control of the photovoltaic energy storage system, and have high application value in the field of energy storage system control technology.

[0068] Referring to Figure 4 , according to one aspect of the present application, an energy scheduling platform system of a photovoltaic energy storage system is provided, the platform system comprising: an acquisition module for acquiring working state data of a photovoltaic energy storage system at a plurality of historical time points through a sensor; a training module for training a state prediction model based on the working state data corresponding to each of the historical time points; a prediction module for predicting a predicted scheduling requirement at a current time point based on the state prediction model and the working state data at a previous time point; and a determination module for determining an energy scheduling instruction corresponding to the photovoltaic energy storage system according to the predicted scheduling requirement in response to an energy scheduling requirement signal at the current time point.

[0069] On the basis of any embodiment of the system of the present application, the system of the present application further comprises: a training data acquisition module configured to determine a state characteristic parameter corresponding to each of the working state data based on a state recognition rule; sort each of the state characteristic parameters from early to late based on the historical time points to obtain a parameter time sequence; and use the parameter time sequence as a training data set to train a preset long short-term memory neural network to obtain a state prediction model.

[0070] On the basis of any embodiment of the system of the present application, the training data acquisition module in the system of the present application, the working state data comprises illumination data, photovoltaic power generation power, energy storage battery voltage, energy storage battery current, energy storage battery temperature, energy storage battery state and electricity price data; the state characteristic parameter comprises illumination intensity, photovoltaic power generation capacity, energy storage capacity, load demand, real-time electricity price, load type and electricity price type; the power grid load type comprises a condition that the power generation capacity is greater than the sum of the energy storage capacity demand and the user load demand, a condition that the power generation capacity is less than the user load demand and a condition that the power generation capacity is greater than or equal to the user load demand; and the electricity price type comprises a low valley electricity price and a peak electricity price.

[0071] On the basis of any embodiment of the system of the application, the system of the application further comprises: a model training module configured to calculate a difference value between the light intensity of any two adjacent state characteristic parameters in the parameter time sequence, to obtain a light difference value; calculate a difference value between the photovoltaic power generation amount of the two adjacent state characteristic parameters, to obtain a power generation difference value; calculate a difference value between the energy storage power of the two adjacent state characteristic parameters, to obtain a storage power difference value; calculate a difference value between the load demand of the two adjacent state characteristic parameters, to obtain a load difference value; calculate a difference value between the real-time electricity price of the two adjacent state characteristic parameters, to obtain an electricity price difference value; calculate a weighted sum of the light difference value, the power generation difference value, the storage power difference value, the load difference value and the electricity price difference value, to obtain a change characteristic parameter; calculate a ratio of the change characteristic parameter to a time difference value between the historical time points corresponding to the two state characteristic parameters, to obtain a correlation characteristic between the two state characteristic parameters; splice the parameter time sequence and the corresponding each correlation characteristic as a fusion feature, and input to a preset long short-term memory neural network for training, to obtain a state prediction model.

[0072] On the basis of any embodiment of the system of the application, in the model training module of the system of the application, the loss function of the long short-term memory neural network is set as a product of a first cross entropy, a second cross entropy and a third cross entropy, the first cross entropy is a difference between a prediction output and a label corresponding to the parameter time sequence; the second cross entropy is a difference between an output correlation characteristic and an actual correlation characteristic; the third cross entropy is a difference between a scheduling demand representation of the prediction output and the output correlation characteristic and a real label corresponding to an actual scheduling demand; the output correlation characteristic is a correlation characteristic between the prediction output and a prediction output of a previous time step; and the actual correlation characteristic is an actual correlation characteristic between corresponding parameters of the parameter time sequence.

[0073] On the basis of any embodiment of the system of the application, in the acquisition module of the system of the application, the sensor comprises a light sensor, a temperature sensor, a battery state sensor and a system state sensor, the light sensor is configured to detect the ambient light intensity, the temperature sensor is configured to detect the working temperature of the photovoltaic energy storage system, the battery state sensor is configured to detect the voltage, current and state of charge of the energy storage battery in the energy storage system, and the system state sensor is configured to identify the load type and the electricity price type according to the working state data.

[0074] On the basis of any embodiment of the system of the application, the system of the application further comprises a demand prediction module configured to obtain working state data of at least two time points before a current time point, determine the working state data as historical state data; calculate the correlation features between two of the historical state data; input the historical state data and the corresponding correlation features into the state prediction model to obtain the output predicted scheduling demand of the current time point.

[0075] On the basis of any embodiment of the system of the application, the system of the application further comprises an instruction determination module configured to obtain current working state data corresponding to the energy scheduling demand signal and corresponding scheduling demand parameters; determine reference state features corresponding to the predicted scheduling demand in a preset database corresponding to the scheduling demand parameters; calculate the feature similarity between the state features of the current working state data and the reference state features; determine whether the feature similarity is greater than a preset similarity threshold; if not, determine the energy scheduling instruction of the photovoltaic energy storage system through actual user interface operation of the scheduling platform; if yes, determine the energy scheduling instruction of the photovoltaic energy storage system based on a preset correspondence between working states and scheduling strategies according to the predicted scheduling demand.

[0076] On the basis of any embodiment of the system of the application, in the instruction determination module of the system of the application, the scheduling strategy comprises: when the light is sufficient, the photovoltaic system generates electricity to preferentially supply the energy storage battery after balancing the load power supply, and secondarily preferentially supply the power grid; when the light is insufficient, the photovoltaic system generates electricity to preferentially supply the user load, the energy storage system preferentially discharges to supply the user load, and if the power generation and discharge do not meet the user load demand, then secondarily preferentially select the power grid to purchase power; when the power grid is powered off, the photovoltaic energy storage system switches to an off-grid mode, and preferentially supplies the user load through photovoltaic power generation and energy storage discharge; when the electricity price is low, the energy storage system is charged by purchasing power from the power grid at a low price; when the electricity price is high, the energy storage system supplies the user load and discharges to supply the power grid to sell power at a high price.

[0077] Another embodiment of the application also provides a photovoltaic energy storage system energy scheduling device, which comprises a processor, a computer readable storage medium, a memory and a network interface connected through a system bus. Wherein the computer readable non-volatile storage medium of the photovoltaic energy storage system energy scheduling device stores an operating system, a database and computer readable instructions, the database can store information sequences, and the computer readable instructions can make the processor realize a photovoltaic energy storage system energy scheduling method when executed by the processor.

[0078] The processor of the energy scheduling device of the photovoltaic energy storage system is configured to provide computing and control capabilities to support the operation of the entire photovoltaic energy storage system. The memory of the energy scheduling device of the photovoltaic energy storage system can store computer readable instructions which, when executed by the processor, cause the processor to perform the energy scheduling method of the photovoltaic energy storage system. The network interface of the energy scheduling device of the photovoltaic energy storage system is configured to communicate with the terminal.

[0079] The processor in the embodiment is configured to execute the specific functions of each module in Figure 4 The memory stores the program codes and various data required for executing the above-mentioned modules or sub-modules. The network interface is configured to realize data transmission between the user terminal and the server.

[0080] The non-volatile readable storage medium in the embodiment stores the program codes and data required for executing all modules in the energy scheduling platform system of the photovoltaic energy storage system. The server can call the program codes and data of the server to execute the functions of all modules.

[0081] The application further provides a non-volatile readable storage medium storing computer readable instructions which, when executed by one or more processors, cause the one or more processors to perform the steps of the energy scheduling method of the photovoltaic energy storage system according to any one of the embodiments.

[0082] The application further provides a computer program product including computer programs / instructions which, when executed by one or more processors, implement the steps of the method according to any one of the embodiments.

Claims

1. An energy scheduling method for a photovoltaic energy storage system, characterized in that, The method comprises the following steps: obtaining working state data of a photovoltaic energy storage system at a plurality of historical time points; training a state prediction model based on the working state data corresponding to each of the historical time points; predicting a predicted scheduling requirement at a current time point based on the working state data at a previous time point based on the state prediction model; determining an energy scheduling instruction corresponding to the photovoltaic energy storage system according to the predicted scheduling requirement in response to an energy scheduling requirement signal at the current time point.

2. The energy scheduling method of the photovoltaic energy storage system according to claim 1, wherein, The training of the state prediction model based on the working state data corresponding to each of the historical time points comprises: determining a state characteristic parameter corresponding to each of the working state data based on a state recognition rule; sequencing each of the state characteristic parameters from early to late based on the historical time points to obtain a parameter time sequence; training a preset long short-term memory neural network using the parameter time sequence as a training data set to obtain the state prediction model.

3. The energy scheduling method of the photovoltaic energy storage system according to claim 2, characterized in that, The working state data comprises illumination data, photovoltaic power generation power, energy storage battery voltage, energy storage battery current, energy storage battery temperature, energy storage battery state and electricity price data; the state characteristic parameter comprises illumination intensity, photovoltaic power generation capacity, energy storage capacity, load demand, real-time electricity price, load type and electricity price type; the grid load type comprises a condition that the photovoltaic power generation capacity is greater than the sum of the energy storage capacity demand and the user load demand, a condition that the photovoltaic power generation capacity is less than the user load demand and a condition that the photovoltaic power generation capacity is greater than or equal to the user load demand; the electricity price type comprises a low-valley electricity price and a peak-peak electricity price.

4. The energy scheduling method of the photovoltaic energy storage system according to claim 3, characterized in that, The training of the state prediction model using the parameter time sequence as a training data set and a preset long short-term memory neural network comprises: calculating a difference value between the illumination intensities of any two adjacent state characteristic parameters in the parameter time sequence to obtain an illumination difference value; calculating a difference value between the photovoltaic power generation capacities of the two adjacent state characteristic parameters to obtain a power generation difference value; calculating a difference value between the energy storage capacities of the two adjacent state characteristic parameters to obtain an energy storage difference value; calculating a difference value between the load demands of the two adjacent state characteristic parameters to obtain a load difference value; calculating a difference value between the real-time electricity prices of the two adjacent state characteristic parameters to obtain an electricity price difference value; calculating a weighted sum of the illumination difference value, the power generation difference value, the energy storage difference value, the load difference value and the electricity price difference value to obtain a change characteristic parameter; calculating a ratio of the change characteristic parameter to a time difference value between the historical time points corresponding to the two state characteristic parameters to obtain a correlation characteristic between the two state characteristic parameters; concatenating the parameter time sequence and each of the corresponding correlation characteristics as a fusion characteristic and inputting the fusion characteristic into a preset long short-term memory neural network for training to obtain the state prediction model.

5. The energy scheduling method of the photovoltaic energy storage system according to claim 4, wherein, The loss function of the long short-term memory neural network is set as a product of a first cross entropy, a second cross entropy and a third cross entropy, the first cross entropy is a difference between a predicted output and a label corresponding to the parameter time sequence; the second cross entropy is a difference between an output correlation feature and an actual correlation feature; the third cross entropy is a difference between the predicted output and a real label corresponding to a scheduling demand representation of the output correlation feature; the output correlation feature is a correlation feature between the predicted output and a predicted output of a previous time step; and the actual correlation feature is an actual correlation feature between corresponding parameters of the parameter time sequence.

6. The energy scheduling method of the photovoltaic energy storage system according to claim 3, wherein, The sensor includes an illumination sensor, a temperature sensor, a battery state sensor, and a system state sensor, the illumination sensor is used to detect the ambient light intensity, the temperature sensor is used to detect the working temperature of the photovoltaic energy storage system, the battery state sensor is used to detect the voltage, current and state of charge of the energy storage battery in the energy storage system, and the system state sensor is used to identify the load type and the electricity price type according to the working state data.

7. The energy scheduling method of the photovoltaic energy storage system according to claim 4, wherein, Based on the state prediction model, the predicted scheduling demand at the current time point is predicted according to the working state data at the previous time point, including: Obtaining the working state data of at least two previous time points at the current time point, and determining the historical state data; Calculate the correlation feature between the two historical state data; The historical state data and the corresponding correlation feature are input into the state prediction model to obtain the output predicted scheduling demand at the current time point.

8. The energy scheduling method of the photovoltaic energy storage system according to claim 7, wherein, According to the predicted scheduling demand, the energy scheduling instruction corresponding to the photovoltaic energy storage system is determined, including: Obtaining the current working state data corresponding to the energy scheduling demand signal and the corresponding scheduling demand parameter; Determine the reference state feature corresponding to the predicted scheduling demand in the preset database corresponding to the scheduling demand parameter; Calculate the feature similarity between the state feature of the current working state data and the reference state feature; Determine whether the feature similarity is greater than a preset similarity threshold; If not, the actual user interface operation of the scheduling platform is used to determine the energy scheduling instruction corresponding to the photovoltaic energy storage system; If yes, according to the predicted scheduling demand, based on the corresponding relationship between the preset working state and the scheduling strategy, the energy scheduling instruction of the photovoltaic energy storage system is determined.

9. The energy scheduling method of the photovoltaic energy storage system according to claim 8, wherein, The scheduling strategy includes: When the light is sufficient, the photovoltaic system generates electricity to charge the energy storage battery after balancing the load power supply, and then supplies the power grid; When the light is insufficient, the photovoltaic system generates electricity to supply the user load, and the energy storage system discharges to supply the user load, if the power generation and discharge do not meet the user load demand, then the power grid is selected as the second priority; When the power grid is powered off, the photovoltaic energy storage system switches to off-grid mode, and supplies the user load through photovoltaic power generation and energy storage discharge; When the valley electricity price, the energy storage system is charged by purchasing low-cost electricity from the power grid; When the peak electricity price, the energy storage system supplies the user load and discharges to supply the power grid for high-priced electricity sales.

10. An energy scheduling platform system for a photovoltaic energy storage system, characterized by, The system is used for executing the energy scheduling method of the photovoltaic energy storage system in any one of claims 1-9, and the system comprises: an acquisition module, configured to acquire working state data of the photovoltaic energy storage system at a plurality of historical time points through a sensor; a training module, configured to train a state prediction model based on the working state data corresponding to each of the historical time points; a prediction module, configured to predict a predicted scheduling demand at a current time point based on the state prediction model and working state data at a previous time point; a determination module, configured to determine an energy scheduling instruction corresponding to the photovoltaic energy storage system according to the predicted scheduling demand in response to an energy scheduling demand signal at the current time point.

Images