New energy station power intelligent prediction system and method

Abstract

The invention relates to the technical field of solar grid-connected control, in particular to a new energy station power intelligent prediction system and method, and the system comprises a collection and transmission module which is used for collecting and transmitting data of a new energy station; the processing and storage module is used for performing conversion processing and storage on the new energy station data; the model prediction module is used for constructing a power prediction model and is used for predicting the solar power generation power and adjusting and optimizing the power prediction model; the model updating module is used for updating the power prediction model; and the dispatching execution module is used for adjusting the power supply quantity of the power grid dispatching center according to the predicted power and the required power and dispatching the power of the power grid. According to the invention, the efficiency, the stability and the sustainable operation capability of new energy power generation are improved.

Description

Technical Field

[0001] The present invention relates to the technical field of solar power grid-connected control, and particularly to an intelligent power prediction system and method for a new energy power station. Background Art

[0002] New energy power generation has the characteristics of intermittency and volatility. The power grid needs to know in advance the power generation of new energy power stations in order to reasonably arrange scheduling and ensure the balance of power supply and demand and the stable operation of the power grid.

[0003] Chinese Patent Publication No.: CN118868231A discloses a solar power grid-connected control system and method through precise control and monitoring, but does not explain how to implement power grid scheduling and how to update the prediction system. Summary of the Invention

[0004] Therefore, the present invention provides an intelligent power prediction system and method for a new energy power station to overcome the problems of inefficient, unstable and unsustainable operation of new energy power generation caused by the lack of a method for implementing power grid scheduling and updating the prediction system in the prior art.

[0005] To achieve the above object, on the one hand, the present invention provides an intelligent power prediction system for a new energy power station, the system comprising: A data acquisition and transmission module for acquiring new energy power station data and also for transmitting the new energy power station data; A data processing and storage module for removing outliers and filling missing values from the new energy power station data to obtain normal new energy power station data, and also for converting and storing the normal new energy power station data to obtain standard new energy power station data; A model prediction module for constructing a power prediction model, and also for adjusting power generation and energy storage plans, power grid scheduling and equipment maintenance according to the power prediction model and the standard new energy power station data, and also for optimizing the power prediction model; A model update module for evaluating the power prediction model and updating the power prediction model according to the evaluation result; A scheduling execution module for scheduling the energy storage system according to the power prediction model and the required power.

[0006] Further, when the acquisition and transmission module transmits new energy power station data, Z-Wave communication interfaces are equipped on the light intensity sensor, temperature sensor, voltage and current sensor, and energy storage system, and the Z-Wave communication interface is set as the data acquisition end. In the cloud platform database, a Z-Wave gateway is configured and set as the data receiving end. A preset number of data acquisition nodes are set between the data acquisition end and the data receiving end, and each data acquisition node is installed and configured at an interval of J1, where 150m ≤ J1 ≤ 200m. The data transmission method includes: Step S10: The data acquisition end collects new energy power station data every 5 minutes, converts the collected new energy power station data into digital signals, and encapsulates them according to the data format of the Z-Wave wireless communication protocol to obtain encapsulated new energy power station data; Step S20: The data acquisition end checks the channel status of the Z-Wave network. If the channel is idle, the data acquisition end starts data transmission and sends the encapsulated new energy power station data into the Z-Wave network; Step S30: Using the AES-128 bit encryption method, a key is output from the encapsulated new energy power station data to obtain a new energy power station data packet; Step S40: Transmit the new energy power station data packet to the Z-Wave gateway, where: Obtain the distance K between the data acquisition end and the data receiving end, compare the distance K with the preset distance K0, judge the validity of the transmission of the new energy power station data packet according to the comparison result, and output according to the judgment result, where: When K ≤ K0, it is determined that the transmission of the new energy power station data packet is effective. Obtain the signal occlusion situation between the data acquisition end and the data receiving end, adjust the validity of the transmission of the new energy power station data packet according to the signal occlusion situation, and output according to the adjustment result, where: If there is no signal occlusion situation between the data acquisition end and the data receiving end, do not adjust the validity of the transmission of the new energy power station data packet, and directly transmit the new energy power station data packet to the data receiving end; If there is a signal occlusion situation between the data acquisition end and the data receiving end, adjust the validity of the transmission of the new energy power station data packet to invalid transmission of the new energy power station data packet, and use the data acquisition node as a relay node to transmit the new energy power station data packet. The data acquisition node transmits the new energy power station data packet to the next appropriate data acquisition node according to its own routing table and signal strength information until the new energy power station data packet finally reaches the data receiving end; When K > K0, it is determined that the data packet transmission of the new energy power station is invalid, and the data acquisition node is used as a relay node to transmit the data packet of the new energy power station. The data acquisition node transmits the data packet of the new energy power station to the next appropriate data acquisition node according to its own routing table and signal strength information until the data packet of the new energy power station finally reaches the data receiving end; Step S50, the data receiving end receives the data packet of the new energy power station, verifies the data packet of the new energy power station, checks the integrity of the data of the new energy power station and the legality of the source, and verifies whether there is an error in the data of the new energy power station during transmission through the AES - 128 - bit encryption method and the key. If the data of the new energy power station is complete and the source is correct, the Z - Wave gateway will extract the data content in the data packet of the new energy power station; Step S60, check the extracted data of the new energy power station. If it is in the format of the original sensor value, perform format conversion and processing.

[0007] Furthermore, when the processing and storage module removes outliers and fills in missing values for the data of the new energy power station, it compares the obtained light intensity data Lx with the preset light intensity data Lx0, judges the normal situation of the light intensity data Lx according to the comparison result, and outputs according to the judgment result, where: When 0 ≤ Lx ≤ Lx0, it is determined that the light intensity data Lx is normal, and the light intensity data Lx is transmitted to the database; When Lx > Lx0, it is determined that the light intensity data Lx is abnormal, and the light intensity data Lx is removed; When Lx < 0, it is determined that the light intensity data Lx is abnormal, and the light intensity data Lx is removed; Compare the obtained ambient temperature data W with the preset ambient temperature data W0, judge the normal situation of the ambient temperature data W according to the comparison result, and output according to the judgment result, where: When W ≤ W0, it is determined that the ambient temperature data W is normal, and the ambient temperature data W is transmitted to the database; When W > W0, it is determined that the ambient temperature data W is abnormal and the data point is incorrect, and the ambient temperature data W is removed; Compare the obtained solar panel temperature data E with the preset solar panel temperature data E0, judge the normal situation of the solar panel temperature data E according to the comparison result, and output according to the judgment result, where: When E ≤ E0, it is determined that the solar panel temperature data E is normal, and the solar panel temperature data E is transmitted to the database; When E > E0, it is determined that the solar panel temperature data E is abnormal and the data point is incorrect, and the solar panel temperature data E is removed; Introduce the voltage change coefficient α, where α > 0. Compare the obtained device voltage data U with the preset device voltage data U0. Judge the normal situation of the device voltage data U according to the comparison result, and output according to the judgment result, where: When αU > U0, it is determined that the device voltage data U is abnormal, and the device voltage data U is removed; When αU ≤ U0, it is determined that the device voltage data U is normal. Obtain the device current data I, and judge the normal situation of the device operating current direction according to the value of the device current data I, and output according to the judgment result, where: If I > 0, it is determined that the device operating current direction is normal, and the device voltage data U is transmitted to the database; If I ≤ 0, it is determined that the device operating current direction is abnormal, and the device voltage data U is removed; Compare the obtained state of charge data C1 with the preset state of charge data C0 change range, and set 20% ≤ C0 ≤ 80%, where; When C1 = C0, the state of charge data change range is normal, and the state of charge data C1 is transmitted to the database; When C1 < C0, the state of charge data change range is abnormal, and the state of charge data is removed; When C1 > C0, the state of charge data change range is abnormal, and the state of charge data is removed; When data is missing, for the light intensity data, use linear interpolation to fill in the missing values. For the ambient temperature and device voltage and current data, fill in the missing values according to the average of the historical data of the ambient temperature and device voltage and current data, and transmit the filled light intensity data, ambient temperature, and device voltage and current data to the database, and mark the transmitted light intensity data Lx, ambient temperature data W, battery panel temperature data E, device voltage data U, and state of charge data C1 as normal new energy station data.

[0008] Furthermore, when the processing and storage module converts and stores the normal new energy station data, it converts the light intensity data in the normal new energy station data. The sensor outputs the original count Y of the luminous flux, and converts it into the light intensity data X according to the parameters of the sensor, ; Perform unit conversion on the voltage and current data in the normal new energy station data, convert milliamperes to amperes, and millivolts to volts, and calculate the power data P of the device according to the current data I and voltage data U, and set P = U × I; Combine the normal new energy power station data with historical data, perform min-max normalization on the normal new energy power station data, and calculate the standard new energy power station data Xn according to the original data X, the minimum value Xmin and the maximum value Xmax in the normal new energy power station data. The formula is as follows: And store the standard new energy power station data using a relational database.

[0009] Furthermore, when constructing the power prediction model in the model prediction module, according to the non-linear relationship characteristics between the power generation power, light intensity, panel temperature and ambient temperature during solar power generation, select the CART decision tree model for construction. Among them, the construction method of the power prediction model includes: Step A1: Divide the preset new energy power station data set into a training set, a validation set, and a test set. The training set is used to train the model parameters, the validation set is used to adjust the model parameters, and the test set is used to evaluate the final performance of the model. Among them, 70% of the new energy power station data set is used as the training set, 15% is used as the validation set, and 15% is used as the test set; Step A2: When using the CART algorithm to construct the CART decision tree model, set the depth of the CART decision tree model to H1, where 3 ≤ H1 ≤ 5, the minimum number of samples required for node splitting is H2 of the total number of samples, where 5% ≤ H2 ≤ 10%, and the minimum number of samples in the leaf node is H3, where 10 ≤ H3 ≤ 20; Step A3: Use the training set data to train the CART decision tree model. During the training process of the CART algorithm, the CART decision tree model will be constructed according to the solar power generation power. At the node, the CART algorithm will find an optimal feature and threshold to make the data purity in the sub-nodes after division according to the optimal feature and threshold the highest. When processing the light intensity feature, the CART algorithm calculates the variance of the power generation power data in the sub-nodes after division by trying different light intensity thresholds, and selects the threshold that makes the variance the smallest as the basis for node splitting. The CART algorithm will repeat this process for other new energy power station data, continuously splitting the nodes until the stop condition is met; Step A4: Use the validation set to evaluate the performance of the model. The evaluation index is the coefficient of determination R². According to the true power generation power value yi, the power generation power value predicted by the model , the average power generation power and the number of samples n in the validation set, calculate the total sum of squares SST and the sum of squared residuals SSE, and set , , calculate the coefficient of determination R² according to the total sum of squares SST and the sum of squared residuals SSE, and set , 0 < R 2<1, compare according to the difference between the preset determination coefficient R0 and 1 and the determination coefficient R 2 , and judge the fitting effect of the model according to the comparison result, where: When 1 - R² ≤ R0, the model prediction module determines that the model fitting effect is good; When 1 - R² > R0, the model prediction module determines that the model fitting effect is not good, adjusts the parameters of the decision tree, reduces the depth of the CART decision tree model and increases the minimum number of samples required for node splitting, retrains the model until the performance meeting the judgment criteria is obtained on the validation set; Step A5, use the test set data that the model has not contacted during the training process to finally test the model, and evaluate the performance of the model in actual application by calculating the determination coefficient R² on the test set.

[0010] Furthermore, when the model prediction module adjusts the power generation and energy storage plans, grid dispatching and equipment maintenance according to the power prediction model, input the newly collected new energy power station data into the power prediction model, calculate according to the rules of the CART decision tree model, and obtain the predicted power generation value P. Compare the difference between the obtained predicted power generation value P and the actual power generation value P0 with the preset deviation threshold β, judge the compliance of the deviation between the predicted power generation value and the actual power generation value according to the comparison result, and output according to the judgment result, where: When |P - P0| ≤ β, the model prediction module determines that the deviation between the predicted power generation value and the actual power generation value is compliant; When |P - P0| > β, the model prediction module determines that the deviation between the predicted power generation value and the actual power generation value is not compliant, and immediate adjustment measures need to be taken. Compare the predicted power generation value P with the actual power generation value P0, and select adjustment measures according to the comparison result, where: If P ≥ P0, the model prediction module determines that the predicted power generation value is greater than the actual power generation value, checks the working state of the inverter, and adjusts the working point of the inverter to optimize its conversion efficiency on the premise of ensuring equipment safety; If P < P0, the model prediction module determines that the predicted power generation value is less than the actual power generation value, compares the difference between the actual light intensity data Lx and the predicted light intensity data Lx' with the preset deviation value £, and selects adjustment measures according to the comparison result, where: When |Lx - Lx'| ≤ £, the model prediction module determines that the deviation between the predicted light intensity data and the actual light intensity data is compliant; When |Lx - Lx'| > £, the model prediction module determines that the deviation between the predicted light intensity data and the actual light intensity data does not meet the standard, and immediate adjustment measures need to be taken. According to the real-time light direction, adjust the angle of the solar panel to make it perpendicular to the sun's rays, increase the light receiving area, and thus improve the power generation efficiency; Compare the predicted state of charge data SOC with the maximum state of charge data SOCmax and the minimum state of charge data SOCmin, judge the normal situation of the energy storage system according to the comparison result, and output according to the judgment result, where: When SOCmin ≤ SOC ≤ SOCmax, the model prediction module determines that the charge and discharge of the energy storage system are normal, and marks the energy storage system as capable of normal charge and discharge; When SOC < SOCmin, the model prediction module determines that the discharge of the energy storage system is abnormal and limits the discharge power of the energy storage system; When SOC > SOCmax, the model prediction module determines that the charging of the energy storage system is abnormal, stops the charging operation, checks the cause of the charging abnormality, and repairs it.

[0011] Further, when the model prediction module optimizes the power prediction model, the Bayesian optimization method is used. By constructing the prior distribution of the parameters, then updating the posterior distribution according to the coefficient R² of the new observed data, and selecting the next parameter combination to be evaluated, applying the selected parameters to the power prediction model, using the evaluation validation set to evaluate the performance of the power prediction model. The evaluation validation set should be independent of the evaluation training set and include different meteorological conditions and power generation data. According to the true power generation value yi in the evaluation validation set, the power generation value predicted by the model and the number of samples n in the evaluation validation set are used to calculate the RMSE of the power prediction model under the selected parameters, and set ; Compare the RMSE of the power prediction model under the selected parameters with the original RMES0, and adjust the optimization scheme of the power prediction model according to the comparison result, where: When RMES > RMES0, the model prediction module determines that the prediction effect of the power prediction model at this time is worse than that before optimization, and readjusts the parameters for optimization; When RMES = RMES0, the model prediction module determines that the prediction effect of the power prediction model at this time is the same as that before optimization, and readjusts the parameters for optimization; When RMES < RMES0, the model prediction module determines that the prediction effect of the power prediction model at this time is better than that before optimization, inputs the selected parameters into the power prediction model, and replaces the parameter values to optimize the model.

[0012] On the other hand, the present invention also provides a method for intelligent power prediction of new energy power stations, including: Step S1: Collect and transmit new energy power station data; Step S2: Remove abnormal data and fill in missing data from the new energy power station data; Step S3: Convert and store the new energy power station data; Step S4: Build a power prediction model, and adjust the power generation and energy storage plans, grid dispatching, and equipment maintenance according to the power prediction model; Step S5: Optimize the power prediction model, and update the power prediction model according to the evaluation results of the power prediction model; Step S6: Adjust the power supply quantity of the grid dispatching center according to the predicted power and the required power, and dispatch the grid power.

[0013] Compared with the prior art, the beneficial effects of the present invention are as follows. The system obtains new energy power station data in real time through the acquisition and transmission module, ensuring the accuracy and timeliness of the data, thereby improving the power generation efficiency. The system processes and stores abnormal data and missing data through the processing and storage module, and performs conversion and storage to ensure data quality. The system builds a power prediction model through the model prediction module to predict the solar power generation power in real time, making the power generation and energy storage plans scientific and reasonable, improving the utilization rate of new energy, predicting and adjusting the power generation power, and optimizing the power prediction model. It can adjust equipment maintenance according to the prediction results, reduce the frequent switching of equipment and unnecessary losses, thereby reducing the operation and maintenance costs. The system updates the power prediction model through the model update module, which helps to improve the accuracy of power prediction. The system dispatches the power supply quantity of the dispatching center and the grid power through the dispatching execution module, improving the efficiency of grid dispatching and realizing the efficient operation of new energy power generation. Description of the Drawings

[0014] Figure 1 It is a schematic structural diagram of the intelligent power prediction system for new energy power stations in this embodiment; Figure 2 It is a schematic flow diagram of the method for intelligent power prediction of new energy power stations in this embodiment. Detailed Embodiments

[0015] In order to make the objectives and advantages of the present invention clearer and more understandable, the present invention will be further described below in conjunction with embodiments; 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.

[0016] The preferred embodiments of the present invention will be described below with reference to the accompanying drawings. Those skilled in the art should understand that these embodiments are only used to explain the technical principle of the present invention and do not limit the protection scope of the present invention.

[0017] It should be noted that in the description of the present invention, unless otherwise clearly defined and limited, the terms "installation", "connection", and "connection" should be understood in a broad sense. For example, it can be a fixed connection, a detachable connection, or an integral connection; it can be a mechanical connection or an electrical connection; it can be directly connected or indirectly connected through an intermediate medium, and it can be the communication inside two components. For those skilled in the art, the specific meanings of the above terms in the present invention can be understood according to specific situations.

[0018] Please refer to Figure 1 as shown in the figure, which is a schematic structural diagram of the new energy power station power intelligent prediction system of this embodiment. The system includes: An acquisition and transmission module, which is used to acquire new energy power station data and also used to transmit new energy power station data; A processing and storage module, which is used to remove outliers and fill in missing values from new energy power station data to obtain normal new energy power station data, and also used to convert and store the normal new energy power station data to obtain standard new energy power station data. The processing and storage module is connected to the acquisition and transmission module; A model prediction module, which is used to construct a power prediction model, and also used to adjust the power generation and energy storage plans, grid dispatching, and equipment maintenance according to the power prediction model and the standard new energy power station data, and also used to optimize the power prediction model. The model prediction module is connected to the processing and storage module; A model update module, which is used to evaluate the power prediction model and update the power prediction model according to the evaluation result. The model update module is connected to the model prediction module; A scheduling execution module, which is used to schedule the energy storage system according to the power prediction model and the required power. The scheduling execution module is connected to the model update module.

[0019] Specifically, the system is applied to the intelligent management and dispatching terminal of the new energy power system. By integrating modules for data acquisition, processing, prediction, optimization, and dispatching, it realizes the efficient, stable, and sustainable operation of new energy power generation, providing strong support for the intelligent management and dispatching of the new energy power system. In particular, the system obtains new energy power station data in real time through the acquisition and transmission module to ensure the accuracy and timeliness of the data, thereby improving power generation efficiency. The system processes, converts, and stores abnormal and missing data through the processing and storage module to ensure data quality. The system constructs a power prediction model through the model prediction module to predict the solar power generation power in real time, making the power generation and energy storage plans scientific and reasonable, improving the utilization rate of new energy, predicting and adjusting the power generation power, and optimizing the power prediction model. It can adjust equipment maintenance according to the prediction results, reduce the frequent switching of equipment and unnecessary losses, and thus reduce the operation and maintenance costs. The system updates the power prediction model through the model update module, which helps to improve the accuracy of power prediction. The system dispatches the power supply of the dispatching center and the grid power through the dispatching execution module to improve the efficiency of grid dispatching and realize the efficient operation of new energy power generation.

[0020] Specifically, when the acquisition and transmission module collects new energy power station data, it is set that the new energy power station data includes light intensity data, temperature data, equipment voltage and current data, and energy storage system data. For the collection of light intensity data, a silicon photovoltaic cell sensor is used for collection, and its light intensity collection range is D1, where 0 W / m² ≤ D1 ≤ 1000 W / m² is set, and the error is D10, where -5% ≤ D10 ≤ 5% is set. The silicon photovoltaic cell sensor should be installed in the central area and the edge area of the solar panel array, and an adjustable-angle mounting bracket is used to ensure that the sensor is always perpendicular to the sun's rays. For the collection of ambient temperature data, a platinum resistance temperature sensor is used for the collection of ambient temperature data, and its measurement range is D2, where -200°C ≤ D2 ≤ 850°C is set, and the accuracy is D3, where -0.3°C ≤ D3 ≤ 0.3°C is set. For the collection of the temperature data of the battery panel, a thermistor temperature sensor is used and installed on the back of the battery panel to be close to the battery cells to measure the working temperature, and its measurement range is D4, where 0°C ≤ D4 ≤ 100°C is set. For the collection of equipment voltage and current data and energy storage system data, voltage sensors and current sensors are respectively installed at the output end of the solar panel, the input and output ends of the inverter, and the output end of the energy storage system. For the voltage sensor, a voltage-dividing voltage sensor is used, and its accuracy is D5, where -0.3% ≤ D5 ≤ 0.3% is set. The current sensor uses a Hall current sensor, and its accuracy is D6, where 0.2% ≤ D6 ≤ 0.2% is set. For the energy storage system data, the state of charge data is collected.

[0021] Specifically, the light intensity data refers to the data of the intensity of light irradiating on the silicon photovoltaic cell sensor. The silicon photovoltaic cell sensor is a photoelectric conversion device made of silicon material that can convert light energy into electrical signals. The solar panel array refers to a power generation unit composed of multiple solar panels. The ambient temperature data refers to the information and values describing and recording the ambient temperature state. The panel temperature data refers to the information and values describing and recording the temperature state of the solar panel. The platinum resistance temperature sensor is a temperature measurement device based on the characteristic of the resistance of platinum metal changing with temperature. The thermistor temperature sensor is a temperature measurement device based on the thermistor element. The voltage and current data refers to the information and values describing and recording the voltage and current states in the circuit. The energy storage system data refers to the information and values describing and recording the operating state, performance parameters, and fault conditions of the energy storage system. The inverter is a converter that converts direct current electrical energy into alternating current electrical energy. The voltage sensor is a sensor that can sense the measured voltage and convert it into an available output signal. The current sensor is a sensor that can sense the measured current and convert it into an available output signal. The voltage divider voltage sensor is a measuring device that reduces a high-voltage signal to a voltage signal within the input range. The Hall current sensor is a current measurement device based on the Hall effect principle. The state of charge data refers to the relative measure of the energy stored in the battery, and its full name is State of Charge, and the Chinese translation is the proportion of the current electricity to the total capacity.

[0022] Specifically, the acquisition and transmission module collects the new energy station data, reflects the operating conditions of the new energy station, and provides strong support for real-time monitoring, early warning, and decision-making.

[0023] Specifically, when the acquisition and transmission module transmits the new energy station data, a Z-Wave communication interface is equipped on the light intensity sensor, temperature sensor, voltage and current sensor, and energy storage system, and the Z-Wave communication interface is set as the data acquisition end. In the cloud platform database, a Z-Wave gateway is configured and set as the data receiving end. A preset number of data acquisition nodes are set between the data acquisition end and the data receiving end, and each data acquisition node is installed and configured at an interval of J1, with 150m ≤ J1 ≤ 200m. The data transmission method includes: Step S10, the data acquisition end collects the new energy station data every 5 minutes, converts the collected new energy station data into digital signals, and encapsulates them according to the data format of the Z-Wave wireless communication protocol to obtain the encapsulated new energy station data; Step S20, the data acquisition end checks the channel status of the Z-Wave network. If the channel is idle, the data acquisition end starts data transmission and sends the encapsulated new energy power station data into the Z-Wave network; Step S30, using the AES-128 bit encryption method, form a key output from the encapsulated new energy power station data to obtain a new energy power station data packet; Step S40, transmit the new energy power station data packet to the Z-Wave gateway, where: Obtain the distance K between the data acquisition end and the data receiving end, compare the distance K with the preset distance K0, judge the validity of the transmission of the new energy power station data packet according to the comparison result, and output according to the judgment result, where: When K≤K0, it is determined that the transmission of the new energy power station data packet is effective. Obtain the signal blocking situation between the data acquisition end and the data receiving end, adjust the validity of the transmission of the new energy power station data packet according to the signal blocking situation, and output according to the adjustment result, where: If there is no signal blocking situation between the data acquisition end and the data receiving end, do not adjust the validity of the transmission of the new energy power station data packet, and directly transmit the new energy power station data packet to the data receiving end; If there is a signal blocking situation between the data acquisition end and the data receiving end, adjust the validity of the transmission of the new energy power station data packet to invalid transmission of the new energy power station data packet, and use the data acquisition node as a relay node to transmit the new energy power station data packet. The data acquisition node transmits the new energy power station data packet to the next suitable data acquisition node according to its own routing table and signal strength information until the new energy power station data packet finally reaches the data receiving end; When K>K0, it is determined that the transmission of the new energy power station data packet is invalid. Use the data acquisition node as a relay node to transmit the new energy power station data packet. The data acquisition node transmits the new energy power station data packet to the next suitable data acquisition node according to its own routing table and signal strength information until the new energy power station data packet finally reaches the data receiving end; Step S50, the data receiving end receives the new energy power station data packet, verifies the new energy power station data packet, checks the integrity of the new energy power station data and the legality of the source, and verifies whether there is an error in the transmission of the new energy power station data through the AES-128 bit encryption method and the key. If the new energy power station data is complete and the source is correct, the Z-Wave gateway will extract the data content in the new energy power station data packet; Step S60, check the extracted new energy power station data. If it is in the original sensor value format, perform format conversion and processing.

[0024] Specifically, the Z-Wave communication interface refers to the interface connecting Z-Wave devices and data acquisition devices. The data acquisition node refers to the physical location for deploying data acquisition devices. The cloud platform database refers to the database service deployed on the cloud platform. The Z-Wave gateway refers to the bridge that connects the network of the Z-Wave wireless communication protocol and the user interaction interface. The data receiving node refers to the network node responsible for receiving data. The Z-Wave network refers to the network composed of Internet of Things devices and gateways that follow the Z-Wave wireless communication protocol. The channel refers to the communication channel for transmitting data under the Z-Wave wireless communication protocol. The new energy station data packet refers to the data obtained by preliminarily processing the collected new energy station data and encapsulating it according to the data format of the Z-Wave wireless communication protocol. The new energy station data packet includes light intensity data, temperature data, device voltage and current data, and energy storage system data. The AES-128-bit encryption method refers to an encryption method that uses a 128-bit key length in the Advanced Encryption Standard for data encryption. The full name of AES is Advanced Encryption Standard, and its Chinese translation is Advanced Encryption Standard. The key refers to the specific value used to control access to data during the encryption and decryption processes. The relay node refers to the device in the network that forwards the incoming data packet to another network node. The original sensor numerical format refers to the numerical data format directly output by the sensor without further processing and conversion.

[0025] Specifically, the acquisition and transmission module transmits the new energy station data, realizes the functions of mutual communication and relay transmission between devices, improves the coverage and reliability of data transmission, and provides reliable security protection for the new energy station data transmission.

[0026] Specifically, when removing outliers and filling in missing values for the new energy station data, the processing and storage module compares the obtained light intensity data Lx with the preset light intensity data Lx0, judges the normal situation of the light intensity data Lx according to the comparison result, and outputs according to the judgment result, where: When 0 ≤ Lx ≤ Lx0, it is determined that the light intensity data Lx is normal, and the light intensity data Lx is transmitted to the database; When Lx > Lx0, it is determined that the light intensity data Lx is abnormal, and the light intensity data Lx is removed; When Lx < 0, it is determined that the light intensity data Lx is abnormal, and the light intensity data Lx is removed; Compare the acquired ambient temperature data W with the preset ambient temperature data W0, judge the normality of the ambient temperature data W according to the comparison result, and output according to the judgment result, where: When W ≤ W0, it is determined that the ambient temperature data W is normal, and the ambient temperature data W is transmitted to the database; When W > W0, it is determined that the ambient temperature data W is abnormal, the data point is in error, and the ambient temperature data W is removed; Compare the acquired solar panel temperature data E with the preset solar panel temperature data E0, judge the normality of the solar panel temperature data E according to the comparison result, and output according to the judgment result, where: When E ≤ E0, it is determined that the solar panel temperature data E is normal, and the solar panel temperature data E is transmitted to the database; When E > E0, it is determined that the solar panel temperature data E is abnormal, the data point is in error, and the solar panel temperature data E is removed; Introduce a voltage change coefficient α, where α > 0, compare the acquired device voltage data U with the preset device voltage data U0, judge the normality of the device voltage data U according to the comparison result, and output according to the judgment result, where: When αU > U0, it is determined that the device voltage data U is abnormal, and the device voltage data U is removed; When αU ≤ U0, it is determined that the device voltage data U is normal, obtain the device current data I, and judge the normality of the device operating current direction according to the value of the device current data I, and output according to the judgment result, where: If I > 0, it is determined that the device operating current direction is normal, and the device voltage data U is transmitted to the database; If I ≤ 0, it is determined that the device operating current direction is abnormal, and the device voltage data U is removed; Compare the acquired state of charge data C1 with the preset state of charge data C0 change range, set 20% ≤ C0 ≤ 80%, where; When C1 = C0, the state of charge data change range is normal, and the state of charge data C1 is transmitted to the database; When C1 < C0, the state of charge data change range is abnormal, and the state of charge data is removed; When C1 > C0, the state of charge data change range is abnormal, and the state of charge data is removed; When data is missing, for the light intensity data, the linear interpolation method is used to fill in the missing values. For the ambient temperature and device voltage and current data, the missing values are filled according to the average values of the historical data of the ambient temperature and device voltage and current data. Then, the light intensity data, ambient temperature, and device voltage and current data after filling in the missing values are transmitted to the database, and the light intensity data Lx, ambient temperature data W, battery panel temperature data E, device voltage data U, and state of charge data C1 transmitted to the database are marked as normal new energy power station data.

[0027] Specifically, the linear interpolation method refers to a method used to estimate and predict the values of unknown data points between known data points. The average value of the historical data of the ambient temperature and device voltage and current data refers to the average value calculated based on the non-missing and valid data points recorded before the occurrence of the missing value.

[0028] Specifically, by effectively identifying and removing abnormal data in the processing and storage module, it helps to reduce errors in data analysis, improve the quality of data-driven decision-making, and provide support for the operation optimization and strategy formulation of new energy power stations.

[0029] Specifically, when the processing and storage module converts and stores normal new energy power station data, it converts the light intensity data in the normal new energy power station data. The original count Y of the luminous flux is output by the sensor, and it is converted into light intensity data X according to the parameters of the sensor. ; For the voltage and current data in the normal new energy power station data, unit conversion is performed, converting milliamperes to amperes and millivolts to volts, and the power data P of the device is calculated based on the current data I and voltage data U, with P = U × I. Combining the normal new energy power station data with historical data, the minimum-maximum normalization processing is performed on the normal new energy power station data. According to the original data X, the minimum value Xmin and maximum value Xmax in the normal new energy power station data, the standard new energy power station data Xn is calculated, and the formula is: , and the standard new energy power station data is stored using a relational database.

[0030] Specifically, the luminous flux refers to the physical quantity of the total light emission or light radiation energy of a light source. The original count refers to the unprocessed digital value directly output by a sensor. The historical data refers to various data collected and stored by a new energy power station within a preset historical time period. In this embodiment, the preset historical time period is not specifically limited. For example, the preset historical time period can be set to 24 hours. The min-max normalization process refers to a data preprocessing technique that scales the original data to the interval [0, 1]. The relational database refers to a database system established on the basis of a relational model.

[0031] Specifically, the processing and storage module converts, performs unit conversion, normalizes, and stores the new energy power station data, providing a solid foundation for subsequent data analysis, mining, and decision-making.

[0032] Specifically, when constructing the power prediction model in the model prediction module, according to the non-linear relationship characteristics between the power generation power, light intensity, panel temperature, and ambient temperature during the solar power generation process, a CART decision tree model is selected for construction. Among them, the construction method of the power prediction model includes: Step A1: Divide the preset new energy power station dataset into a training set, a validation set, and a test set. The training set is used to train the model parameters, the validation set is used to adjust the model parameters, and the test set is used to evaluate the final performance of the model. Among them, 70% of the new energy power station dataset is used as the training set, 15% is used as the validation set, and 15% is used as the test set. Step A2: When using the CART algorithm to construct the CART decision tree model, set the depth of the CART decision tree model to H1, where 3 ≤ H1 ≤ 5. The minimum number of samples required for node splitting is H2 of the total number of samples, where 5% ≤ H2 ≤ 10%. The minimum number of samples in the leaf node is H3, where 10 ≤ H3 ≤ 20. Step A3: Use the training set data to train the CART decision tree model. During the training process of the CART algorithm, the CART decision tree model will be constructed according to the solar power generation power. At the node, the CART algorithm will find an optimal feature and threshold to make the data purity in the sub-nodes after division the highest. When processing the light intensity feature, the CART algorithm calculates the variance of the power generation power data in the sub-nodes after division by trying different light intensity thresholds, and selects the threshold that minimizes the variance as the basis for node splitting. The CART algorithm will repeat this process for other new energy power station data, continuously splitting nodes until the stop condition is met. Step A4: Use the validation set to evaluate the performance of the model. The evaluation index is the coefficient of determination R². According to the true power generation power value yi, the power generation power value predicted by the model and the average power generation power Calculate the total sum of squares SST and the residual sum of squares SSE based on the training set sample number m and the validation set sample number n, and set , , calculate the coefficient of determination R² based on the total sum of squares SST and the residual sum of squares SSE, and set , 0 < R 2 < 1, compare according to the difference between the preset coefficient of determination R0 and 1 and the coefficient of determination R 2 , and judge the fitting effect of the model according to the comparison result, where: When 1 - R² ≤ R0, the model prediction module determines that the model fitting effect is good; When 1 - R² > R0, the model prediction module determines that the model fitting effect is not good, adjusts the parameters of the decision tree, reduces the depth of the CART decision tree model and increases the minimum number of samples required for node splitting, and retrains the model until the performance meeting the judgment criteria is obtained on the validation set; Step A5, use the test set data that the model has not been exposed to during the training process to finally test the model, and evaluate the performance of the model in actual application by calculating the coefficient of determination R² on the test set.

[0033] Specifically, the CART decision tree model refers to a decision tree learning technique for constructing a model by recursively dividing the feature space into binary nodes, the CART algorithm refers to a decision tree learning technique for classification tasks and regression tasks, the key parameter tree refers to the key parameters that need to be determined when constructing the CART decision tree model, the maximum depth refers to the parameter that controls the complexity of the decision tree, the minimum number of samples required for node splitting refers to the minimum number of samples that the node needs to contain when performing node splitting, the minimum number of samples in the leaf node refers to the minimum number of samples required for the leaf node, the best feature and threshold refer to the optimal feature selected by the CART algorithm and the corresponding threshold when performing node splitting, and the stopping condition refers to reaching the maximum tree depth and the number of samples in the node being less than the minimum number of samples.

[0034] Specifically, the model prediction module improves the accuracy and applicability of the solar power generation prediction model, and provides support for the power prediction and scheduling of solar power generation.

[0035] Specifically, when the model prediction module adjusts the power generation and energy storage plans, grid dispatching, and equipment maintenance according to the power prediction model, the newly collected data of the new energy power station is input into the power prediction model, and calculations are performed according to the rules of the CART decision tree model to obtain the predicted power generation value P. The difference between the obtained predicted power generation value P and the actual power generation value P0 is compared with the preset deviation threshold β, and based on the comparison result, the compliance of the deviation between the predicted power generation value and the actual power generation value is judged, and the output is made according to the judgment result, where: When |P - P0| ≤ β, the model prediction module determines that the deviation between the predicted power generation value and the actual power generation value is compliant; When |P - P0| > β, the model prediction module determines that the deviation between the predicted power generation value and the actual power generation value is non-compliant, and immediate adjustment measures need to be taken. The predicted power generation value P is compared with the actual power generation value P0, and adjustment measures are selected according to the comparison result, where: If P ≥ P0, the model prediction module determines that the predicted power generation value is greater than the actual power generation value, checks the working state of the inverter, and adjusts the working point of the inverter to optimize its conversion efficiency on the premise of ensuring equipment safety; If P < P0, the model prediction module determines that the predicted power generation value is less than the actual power generation value, and the difference between the actual light intensity data Lx and the predicted light intensity data Lx' is compared with the preset deviation value £, and adjustment measures are selected according to the comparison result, where: When |Lx - Lx'| ≤ £, the model prediction module determines that the deviation between the predicted light intensity data and the actual light intensity data is compliant; When |Lx - Lx'| > £, the model prediction module determines that the deviation between the predicted light intensity data and the actual light intensity data is non-compliant, and immediate adjustment measures need to be taken. According to the real-time light direction, the angle of the solar panel is adjusted to make it perpendicular to the sun's rays, increasing the light receiving area to improve the power generation efficiency; The predicted state of charge data SOC is compared with the maximum state of charge data SOCmax and the minimum state of charge data SOCmin, and based on the comparison result, the normal condition of the energy storage system is judged, and the output is made according to the judgment result, where: When SOCmin ≤ SOC ≤ SOCmax, the model prediction module determines that the charge and discharge of the energy storage system are normal, and marks the energy storage system as capable of normal charge and discharge; When SOC < SOCmin, the model prediction module determines that the discharge of the energy storage system is abnormal and limits the discharge power of the energy storage system; When SOC > SOCmax, the model prediction module determines that the charging of the energy storage system is abnormal, stops the charging operation, checks the cause of the charging abnormality, and repairs it.

[0036] Specifically, the inverter refers to a converter that converts direct current electrical energy into alternating current electrical energy, and the operating point refers to the operating state and parameter settings of the inverter during operation.

[0037] Specifically, in the model prediction module, by collecting new energy power station data in real time and inputting it into the power prediction model, and calculating according to the rules of the CART decision tree model, the power generation power can be accurately predicted, thereby optimizing the power generation efficiency system.

[0038] Specifically, when the model prediction module optimizes the power prediction model, the Bayesian optimization method is used. By constructing the prior distribution of the parameters, then determining the coefficient R² according to the new observed data to update the posterior distribution, and selecting the next parameter combination to be evaluated, the selected parameters are applied to the power prediction model, and the evaluation validation set is used to evaluate the performance of the power prediction model. The evaluation validation set should be independent of the evaluation training set and include different meteorological conditions and power generation power data. According to the true power generation power value yi in the evaluation validation set, the power generation power value predicted by the model and the sample number n of the evaluation validation set are used to calculate the RMSE of the power prediction model under the selected parameters, and it is set that ; The RMSE of the power prediction model under the selected parameters is compared with the original RMES0, and the optimization scheme of the power prediction model is adjusted according to the comparison result, where: When RMES > RMES0, the model prediction module determines that the prediction effect of the power prediction model at this time is worse than that before optimization, and adjusts the parameters again for optimization; When RMES = RMES0, the model prediction module determines that the prediction effect of the power prediction model at this time is the same as that before optimization, and adjusts the parameters again for optimization; When RMES < RMES0, the model prediction module determines that the prediction effect of the power prediction model at this time is better than that before optimization, and inputs the selected parameters into the power prediction model to replace the parameter values to optimize the model.

[0039] Specifically, the Bayesian optimization refers to a black-box algorithm for optimizing the objective function applicable to parameter tuning of machine learning models. The prior distribution refers to an assumption about the parameter distribution of the power prediction model before observing new data. The posterior distribution refers to the parameter distribution updated according to Bayesian optimization after observing new data. The selected parameters refer to the next set of parameter combinations intelligently selected and determined by Bayesian optimization for evaluation, such as the depth of the tree, the minimum number of samples required for node splitting, the minimum number of samples in the leaf node, and the feature selection criteria for splitting nodes.

[0040] Specifically, using the Bayesian optimization method in the model prediction module to optimize the power prediction model can improve the prediction accuracy, reliability, and generalization ability of the model, while realizing the automation of the optimization process and the efficient utilization of resources.

[0041] Specifically, in the model update module, the power prediction model is evaluated within a preset update time range. The root mean square error RMSE1 of the current power prediction model is compared with the root mean square error RMSE01 of the preset power prediction model. According to the comparison result, the performance of the power prediction model is judged, where: When RMSE1 ≤ RMSE01, the model update module determines that the performance of the power prediction model is stable and there is no need to update the model; When RMSE1 > RMSE01, the model update module determines that the performance of the power prediction model has declined and the model needs to be updated. Before updating the model, it is necessary to clarify the requirements corresponding to the update reason, where: For update requirements caused by equipment changes, data related to the characteristics of the new equipment should be collected, and a plan to incorporate the new equipment characteristics into the model should be determined; For update requirements caused by changes in meteorological conditions, the change factors of the meteorological conditions should be added to the model; Use the brand-new collected data and adjusted parameters, and retrain and validate the power prediction model to update the power prediction model; In addition, when introducing new variables to improve the prediction accuracy, if it is found that there are influencing factors that have a significant impact on the generated power and are not included in the power prediction model, the influencing factors are added to the power prediction model as new variables and the power prediction model is updated.

[0042] Specifically, the preset update time range refers to the time period preset for regularly evaluating the power prediction model. In this embodiment, the preset update time range is not specifically limited. For example, the preset update time range can be set to one week. The new equipment characteristics refer to the characteristics of the new equipment introduced due to equipment changes. The meteorological condition change factors refer to various meteorological conditions and their changes that affect the generated power. The influencing factors refer to the factors that have a significant impact on the generated power, such as small changes in local meteorological phenomena.

[0043] Specifically, through regular evaluation, the model update module can promptly detect the decline in the performance of the power prediction model and take corresponding update measures, which helps to ensure that the model always maintains an accurate prediction level.

[0044] Specifically, when the scheduling execution module inputs the standard new energy power station data into the power prediction model to obtain the power prediction result W, it compares the power prediction result W with the power demand result W0, judges the future power generation amount according to the comparison result, and schedules the energy storage system according to the judgment result, where: When W > W0, the scheduling execution module determines that the future power generation is large, adjusts the charging strategy of the energy storage system to store the excess power, and observes the state of charge of the energy storage system. When the state of charge data C1 = 100%, the energy storage system stops charging; When W ≤ W0, the scheduling execution module determines that the future power generation is small, controls the energy storage system to discharge to supplement the insufficient power generation and meet the electricity demand of the load, and observes the state of charge of the energy storage system. When the state of charge data C1 = 0%, the energy storage system stops discharging.

[0045] Specifically, the future power generation amount refers to the predicted power generation amount in the future time period obtained according to the prediction result of the power prediction model.

[0046] Specifically, the scheduling execution module can accurately understand the future power generation and electricity demand of the power grid, so as to perform power scheduling and allocation in advance, which helps to ensure the stable operation of the power grid, avoid power shortages and surpluses, and improve the power supply reliability and efficiency of the power grid.

[0047] Please refer to Figure 2 The following is a schematic flow chart of the new energy power station power intelligent prediction method of this embodiment. The method includes: Step S1, collecting and transmitting new energy power station data; Step S2, removing abnormal data and filling missing data from the new energy power station data; Step S3, converting and storing the new energy power station data; Step S4, constructing a power prediction model, and adjusting the power generation and energy storage plans, power grid scheduling, and equipment maintenance according to the power prediction model; Step S5, optimizing the power prediction model, and updating the power prediction model according to the evaluation result of the power prediction model; Step S6, adjusting the power supply amount of the power grid dispatching center according to the predicted power and the required power, and dispatching the power grid power.

[0048] So far, the technical solution of the present invention has been described in conjunction with the preferred embodiments shown in the accompanying drawings. However, it is easily understood by those skilled in the art that the protection scope of the present invention is obviously not limited to these specific embodiments. Without departing from the principle of the present invention, those skilled in the art can make equivalent changes or substitutions to the relevant technical features, and the technical solutions after these changes or substitutions will all fall within the protection scope of the present invention.

Claims

1. A new energy station power intelligent prediction system, characterized in that: The system comprises: The acquisition and transmission module is used to collect and transmit data from new energy stations; A processing and storage module is used to remove abnormal values ​​and fill missing values ​​of the new energy station data to obtain normal new energy station data, and is also used to convert and store normal new energy station data to obtain standard new energy station data; Model prediction module, used to build a power prediction model, and also used to adjust power generation and storage plans, grid dispatching and equipment maintenance according to the power prediction model and standard new energy station data, and also used to optimize the power prediction model; A model updating module is used to evaluate the power prediction model and update the power prediction model according to the evaluation result; The scheduling execution module is used to schedule the energy storage system according to the power prediction model and the required power.

2. The new energy station power intelligent prediction system according to claim 1 is characterized in that: When the acquisition and transmission module transmits data of the new energy station, a Z-Wave communication interface is provided on the light intensity sensor, the temperature sensor, the voltage and current sensor and the energy storage system, and the Z-Wave communication interface is set as the data acquisition end. In the cloud platform database, a Z-Wave gateway is configured and set as the data receiving end. A preset number of data acquisition nodes are set between the data acquisition end and the data receiving end, and each data acquisition node is installed and configured at an interval of J1, and 150m≤J1≤200m is set. The data transmission method includes: Step S10, the data acquisition terminal collects the new energy station data every 5 minutes, converts the collected new energy station data into digital signals, and encapsulates them according to the data format of the Z-Wave wireless communication protocol to obtain encapsulated new energy station data; Step S20, the data acquisition end checks the channel status of the Z-Wave network. If the channel is idle, the data acquisition end starts data transmission and sends the packaged new energy station data to the Z-Wave network; Step S30, using AES-128 bit encryption method, encapsulating the new energy station data to form a key output, and obtaining a new energy station data packet; Step S40, transmitting the new energy station data packet to the Z-Wave gateway, wherein: Obtain the distance K between the data acquisition end and the data receiving end, and compare the distance K with the preset distance K0. According to the comparison result, judge the validity of the data packet transmission of the new energy station, and output it according to the judgment result, where: When K≤K0, the data packet transmission of the new energy station is determined to be valid, the signal shielding situation between the data acquisition end and the data receiving end is obtained, and the validity of the data packet transmission of the new energy station is adjusted according to the signal shielding situation, and the adjustment result is output, where: If there is no signal blocking between the data collection end and the data receiving end, the validity of the new energy station data packet transmission is not adjusted, and the new energy station data packet is directly transmitted to the data receiving end; If there is signal blocking between the data collection end and the data receiving end, the validity of the new energy station data packet transmission is adjusted to invalid, and the data collection node is used as a relay node to transmit the new energy station data packet. The data collection node transmits the new energy station data packet to the next appropriate data collection node according to its own routing table and signal strength information until the new energy station data packet finally reaches the data receiving end; When K>K0, it is determined that the transmission of the new energy station data packet is invalid, and the data collection node is used as a relay node to transmit the new energy station data packet. The data collection node transmits the new energy station data packet to the next appropriate data collection node according to its own routing table and signal strength information until the new energy station data packet finally reaches the data receiving end; Step S50, the data receiving end receives the new energy station data packet, verifies the new energy station data packet, checks the integrity of the new energy station data and the legitimacy of the source, and verifies whether there is any error in the transmission process of the new energy station data through AES-128 bit encryption and key. If the new energy station data is complete and the source is correct, the Z-Wave gateway will extract the data content in the new energy station data packet; Step S60, checking the extracted new energy station data, and if it is in the original sensor value format, performing format conversion and processing.

3. The new energy station power intelligent prediction system according to claim 1 is characterized in that: When removing outliers and filling missing values ​​for the new energy station data, the processing and storage module compares the acquired light intensity data Lx with the preset light intensity data Lx0, judges the normality of the light intensity data Lx according to the comparison result, and outputs it according to the judgment result, wherein: When 0≤Lx≤Lx0, the light intensity data Lx is determined to be normal, and the light intensity data Lx is transmitted to the database; When Lx>Lx0, it is determined that the light intensity data Lx is abnormal, and the light intensity data Lx is removed; When Lx<0, it is determined that the light intensity data Lx is abnormal, and the light intensity data Lx is removed; The acquired ambient temperature data W is compared with the preset ambient temperature data W0, and the normal situation of the ambient temperature data W is judged according to the comparison result, and output according to the judgment result, wherein: When W≤W0, the ambient temperature data W is determined to be normal, and the ambient temperature data W is transmitted to the database; When W>W0, it is determined that the ambient temperature data W is abnormal, and an error occurs in the data point, and the ambient temperature data W is removed; The obtained solar panel temperature data E is compared with the preset solar panel temperature data E0, and the normal situation of the solar panel temperature data E is judged according to the comparison result, and output according to the judgment result, wherein: When E≤E0, the solar panel temperature data E is determined to be normal, and the solar panel temperature data E is transmitted to the database; When E>E0, the panel temperature data E is determined to be abnormal, and an error occurs in the data point, and the panel temperature data E is removed; The voltage variation coefficient α is introduced, where α>0, and the acquired device voltage data U is compared with the preset device voltage data U0. The normal situation of the device voltage data U is judged according to the comparison result, and the judgment result is output, where: When αU>U0, the device voltage data U is determined to be abnormal and the device voltage data U is removed; When αU≤U0, the device voltage data U is determined to be normal, the device current data I is obtained, and the normal situation of the device operating current direction is judged according to the value of the device current data I, and the judgment result is output, where: If I>0, it is determined that the current direction of the equipment is normal, and the voltage data U of the equipment is transmitted to the database; If I≤0, it is determined that the current direction of the equipment is abnormal, and the voltage data U of the equipment is removed; Compare the acquired state of charge data C1 with the preset state of charge data C0 variation range, and set 20%≤C0≤80%, where; When C1=C0, the charge state data variation range is normal, and the charge state data C1 is transmitted to the database; When C1<C0, the change range of the state of charge data is abnormal, and the state of charge data is removed; When C1>C0, the change range of the state of charge data is abnormal, and the state of charge data is removed; When data is missing, for light intensity data, linear interpolation is used to fill the missing values. For ambient temperature and equipment voltage and current data, the missing values ​​are filled according to the average value of the historical data of ambient temperature and equipment voltage and current data. The light intensity data, ambient temperature and equipment voltage and current data after filling the missing values ​​are transmitted to the database, and the light intensity data Lx, ambient temperature data W, battery panel temperature data E, equipment voltage data U and charge status data C1 transmitted to the database are marked as normal new energy station data.

4. The new energy station power intelligent prediction system according to claim 1 is characterized in that: When the processing and storage module converts and stores the normal new energy station data, it converts the light intensity data in the normal new energy station data. The sensor outputs the original count Y of the light flux, which is converted into light intensity data X according to the parameters of the sensor. ; Convert the units of voltage and current data in normal new energy station data, convert milliamperes to amperes, and millivolts to volts, and calculate the power data P of the equipment based on the current data I and voltage data U, setting P=U×I; Combine the normal new energy station data with the historical data, perform minimum-maximum normalization on the normal new energy station data, and calculate the standard new energy station data Xn based on the original data X, the minimum value Xmin and the maximum value Xmax in the normal new energy station data. The formula is: And use relational database to store standard new energy station data.

5. The new energy station power intelligent prediction system according to claim 1 is characterized in that: When the model prediction module constructs the power prediction model, according to the nonlinear relationship between the power generation power and the light intensity, the panel temperature and the ambient temperature in the solar power generation process, the CART decision tree model is selected for construction, wherein the power prediction model construction method includes: Step A1, divide the preset new energy site data set into a training set, a validation set, and a test set, the training set is used to train the parameters of the model, the validation set is used to adjust the parameters of the model, and the test set is used to evaluate the final performance of the model, wherein 70% of the new energy site data set is used as a training set, 15% as a validation set, and 15% as a test set; Step A2, when using the CART algorithm to construct the CART decision tree model, the depth of the CART decision tree model is set to H1, 3≤H1≤5, the minimum number of samples required for node splitting is H2 of the total number of samples, 5%≤H2≤10%, the minimum number of samples for leaf nodes is H3, 10≤H3≤20; Step A3, using the training set data to train the CART decision tree model. During the training process, the CART algorithm will construct the CART decision tree model according to the solar power generation power. At the node, the CART algorithm will find an optimal feature and threshold value so that the data purity in the sub-node divided according to the optimal feature and threshold value is the highest. When processing the light intensity feature, the CART algorithm calculates the variance of the power generation data in the sub-node after the division by trying different light intensity thresholds, and selects the threshold value with the smallest variance as the basis for node splitting. The CART algorithm for processing other new energy station data will repeat this process and continuously split the node until the stop condition is met; Step A4: Use the validation set to evaluate the performance of the model. The evaluation index is the determination coefficient R², which is calculated based on the actual power generation value yi and the power generation value predicted by the model. , average power generation Calculate the total sum of squares SST and the residual sum of squares SSE with the number of samples n in the validation set, and set , , calculate the coefficient of determination R² based on the total sum of squares SST and the residual sum of squares SSE, and set , 0<R 2 <1, according to the preset determination coefficient R0 and 1 and the determination coefficient R 2 The difference is compared, and the fitting effect of the model is judged according to the comparison results, where: When 1-R²≤R0, the model prediction module determines that the model fitting effect is good; When 1-R²>R0, the model prediction module determines that the model fitting effect is not good, adjusts the parameters of the decision tree, reduces the depth of the CART decision tree model and increases the minimum number of samples required for node splitting, and retrains the model until the performance that meets the judgment standard is obtained on the validation set; In step A5, the model is tested on the test set data that the model has not been exposed to during the training process. The performance of the model in actual application is evaluated by calculating the coefficient of determination R² on the test set.

6. The new energy station power intelligent prediction system according to claim 5 is characterized in that: When the model prediction module adjusts the power generation and storage plan, grid dispatching and equipment maintenance according to the power prediction model, the newly collected new energy station data is input into the power prediction model, and the calculation is performed according to the rules of the CART decision tree model to obtain the power generation prediction value P. The difference between the obtained power generation prediction value P and the actual power generation value P0 is compared with the preset deviation threshold β. The compliance of the deviation between the power generation prediction value and the actual power generation value is judged according to the comparison result, and the judgment result is output, wherein: When |P-P0|≤β, the model prediction module determines that the deviation between the predicted value of generated power and the actual value of generated power meets the standard; When |P-P0|>β, the model prediction module determines that the deviation between the predicted power generation value and the actual power generation value does not meet the standard, and adjustment measures need to be taken immediately. The predicted power generation value P is compared with the actual power generation value P0, and the adjustment measures are selected according to the comparison results, where: If P≥P0, the model prediction module determines that the predicted value of generated power is greater than the actual value of generated power, checks the working state of the inverter, adjusts the working point of the inverter on the premise of ensuring the safety of the equipment, and optimizes its conversion efficiency; If P<P0, the model prediction module determines that the predicted value of generated power is less than the actual value of generated power, compares the difference between the actual light intensity data Lx and the predicted light intensity data Lx' with the preset deviation value £, and selects adjustment measures according to the comparison result, wherein: When |Lx-Lx'|≤£, the model prediction module determines that the deviation between the predicted light intensity data and the actual light intensity data meets the standard; When |Lx-Lx'|>£, the model prediction module determines that the deviation between the predicted light intensity data and the actual light intensity data does not meet the standard, and immediate adjustment measures need to be taken. According to the real-time light direction, the angle of the solar panel is adjusted to make it perpendicular to the sunlight, increasing the light receiving area and thus improving the power generation efficiency; The predicted state of charge data SOC is compared with the maximum value SOCmax and the minimum value SOCmin of the state of charge data, and the normal situation of the energy storage system is judged according to the comparison result, and the judgment result is output, where: When SOCmin≤SOC≤SOCmax, the model prediction module determines that the energy storage system is charging and discharging normally, and marks the energy storage system as being able to be charged and discharged normally; When SOC<SOCmin, the model prediction module determines that the energy storage system discharge is abnormal and limits the discharge power of the energy storage system; When SOC>SOCmax, the model prediction module determines that the energy storage system is abnormally charged, stops the charging operation, investigates the cause of the abnormal charging, and performs repairs.

7. The new energy station power intelligent prediction system according to claim 6 is characterized in that: When the model prediction module optimizes the power prediction model, the Bayesian optimization method is used to construct a prior distribution of the parameters, and then update the posterior distribution according to the new observation data determination coefficient R², and select the next parameter combination to be evaluated, and apply the selected parameters to the power prediction model. The evaluation validation set is used to evaluate the performance of the power prediction model. The evaluation validation set should be independent of the evaluation training set, including different meteorological conditions and power generation data. According to the actual power generation value yi in the evaluation validation set, the power generation value predicted by the model The RMSE of the power prediction model under the selected parameters is calculated by evaluating the number of samples in the validation set n. ; The RMSE of the power prediction model under the selected parameters is compared with the original RMES0, and the optimization scheme of the power prediction model is adjusted according to the comparison results, where: When RMES>RMES0, the model prediction module determines that the prediction effect of the power prediction model at this time is worse than that before optimization, and readjusts the parameters for optimization; When RMES=RMES0, the model prediction module determines that the prediction effect of the power prediction model at this time is consistent with that before optimization, and readjusts the parameters for optimization; When RMES<RMES0, the model prediction module determines that the prediction effect of the power prediction model at this time is better than that before optimization, inputs the selected parameters into the power prediction model, and replaces the parameter values ​​to optimize the model.

8. A method applied to the new energy station power intelligent prediction system as described in claims 1-7, characterized in that: include: Step S1, collecting and transmitting new energy station data; Step S2, removing abnormal data and filling missing data from the new energy station data; Step S3, converting and storing the new energy station data; Step S4, constructing a power prediction model, and adjusting power generation and storage plans, grid dispatching, and equipment maintenance according to the power prediction model; Step S5, optimizing the power prediction model, and updating the power prediction model according to the evaluation result of the power prediction model; Step S6, adjusting the power supply of the power grid dispatching center and dispatching the power grid according to the predicted power and the required power.

Images