Distributed photovoltaic station reverse power forecasting method and system based on low-frequency system

Abstract

The invention provides a distributed photovoltaic station reverse power forecasting method and system based on a low-frequency system. The method comprises the following steps: acquiring operation data of a distributed photovoltaic station and the maximum transmission capacity of the low-frequency system; performing power prediction on the distributed photovoltaic station by using a neural network combination model according to the operation data to obtain power prediction data; based on the power prediction data and the maximum transmission capacity, judging whether the distributed photovoltaic field station has a reverse transmission power risk or not; when the distributed photovoltaic field station has the reverse transmission power risk, an optimization control algorithm is adopted to solve the target function, an optimal prevention and control strategy is obtained, and reverse transmission power forecasting is carried out on the distributed photovoltaic field station according to the optimal prevention and control strategy; according to the method, collaborative power prediction is carried out on the distributed photovoltaic station by using the neural network combination model, and risk judgment is carried out on the distributed photovoltaic station in combination with the transmission capacity of the low-frequency system, so that prospective and precise identification of potential reverse transmission power risks can be realized.

Description

Technical Field

[0001] This invention relates to the field of power system technology, and specifically to a method and system for predicting the reverse power transmission of distributed photovoltaic power plants based on low-frequency systems. Background Technology

[0002] Currently, distributed photovoltaic (PV) power output is significantly affected by natural factors such as weather, temperature, and solar radiation, exhibiting strong random fluctuations. The supply-demand balance between distributed PV and regional loads relies on accurate power forecasting. Furthermore, the accuracy of power forecasting directly determines the targetedness and effectiveness of backfeeding prevention measures, and is a crucial prerequisite for ensuring the safe and stable operation of the power grid. If the forecast results deviate too much, it can easily lead to misjudgment or underjudgment of backfeeding risks, thereby causing problems such as grid voltage fluctuations and frequency anomalies.

[0003] In existing technologies, power prediction for distributed photovoltaic (PV) power plants often employs either a single model or completely separate prediction models. Both types of models have significant limitations: the former attempts to simultaneously predict PV power and regional load power using the same model, but struggles to adapt to the differentiated characteristics of the two types of power—PV power requires precise correlation with long-term temporal changes in meteorological factors, while load power needs to capture multi-scale periodic patterns and fluctuations caused by social behaviors such as dates and holidays. A single model cannot simultaneously address the core needs of both, inevitably leading to significant prediction bias. The latter, while separating the prediction processes for PV and load power, suffers from a lack of synergy in model design, resulting in misalignments in prediction time granularity and data acquisition dimensions, thus reducing the reliability of the prediction results. Furthermore, traditional reverse transmission power risk assessment mechanisms do not take into account the maximum transmission capacity of the low-frequency system connected to the distributed PV power plant. They simply determine reverse transmission risk based on the difference between PV power and load power, causing the risk assessment results to deviate from the actual transmission capacity of the low-frequency grid-connected scenario and fail to adapt to the carrying capacity characteristics of the low-frequency system.

[0004] In summary, existing technologies suffer from poor adaptability and insufficient accuracy in power prediction models, as well as shortcomings in assessing back-feeding power risk by ignoring low-frequency system characteristics and being detached from actual operating scenarios. Ultimately, these technologies fail to meet the grid's requirements for back-feeding power risk control in low-frequency grid-connected scenarios, thus affecting the safety and economy of distributed photovoltaic power plant grid-connected operation. Summary of the Invention

[0005] To address the technical problems of existing technologies that still rely on single or separate models for backfeed power forecasting, resulting in insufficient power prediction accuracy and poor adaptability, and failing to consider the characteristics of low-frequency systems connected to distributed photovoltaic (PV) power plants, leading to biased backfeed power risk assessment and inability to meet the backfeed power forecasting needs in low-frequency grid-connected scenarios, this invention proposes a backfeed power forecasting method for distributed PV power plants based on low-frequency systems, comprising: Obtain the operating data of the distributed photovoltaic power station and the maximum transmission capacity of the low-frequency system connected to the grid with the distributed photovoltaic power station; Based on the operational data, a pre-built neural network combination model is used to predict the power of the distributed photovoltaic power station, thereby obtaining the power prediction data of the distributed photovoltaic power station. Based on the maximum transmission capacity and the power prediction data, it is determined whether the distributed photovoltaic power station has the risk of reverse power transmission; When the distributed photovoltaic power station faces the risk of reverse power transmission, an optimization control algorithm is used to solve the pre-set objective function to obtain the optimal prevention and control strategy, and the reverse power transmission of the distributed photovoltaic power station is predicted based on the optimal prevention and control strategy. The neural network combination model includes a photovoltaic power generation prediction sub-model and a load power prediction sub-model. The photovoltaic power generation prediction sub-model is constructed based on a long short-term memory (LSTM) network, and the load power prediction sub-model is constructed based on a temporal convolutional network (TCN). Optionally, acquiring the operational data of the distributed photovoltaic power station includes: Collect initial power generation data of each photovoltaic field in the distributed photovoltaic power station; wherein, the initial power generation data includes one or more of the following: real-time power generation, photovoltaic module operating temperature and combiner box output current; The initial power generation data of each photovoltaic field is collected and its status is monitored by the low-frequency switching station. The power generation power of each photovoltaic field is integrated to obtain the total initial power generation power of the station. The switching status and low-frequency input power of the low-frequency switching station are collected simultaneously to obtain the aggregated power generation data. The transmission process data of the aggregated power generation data is obtained by monitoring the transmission process through a low-frequency line. Power conversion data is obtained by monitoring the power conversion of the transmission process data using a frequency converter. The grid connection point monitoring device of the power frequency power grid connection line collects the real-time voltage, current and power flow direction of the grid connection point to obtain the grid connection interface data; The operating data of the distributed photovoltaic power station is obtained by integrating and verifying the summarized power generation data, transmission process data, power conversion data and grid connection interface data.

[0006] Optionally, the step of using a pre-built neural network combination model to predict the power of the distributed photovoltaic power station based on the operational data, to obtain the power prediction data of the distributed photovoltaic power station, includes: Based on the operational data, the photovoltaic power generation prediction sub-model is used to predict the power generation of the distributed photovoltaic power station, thereby obtaining the photovoltaic power generation prediction data of the distributed photovoltaic power station. Based on the operational data, the load power prediction sub-model is used to predict the load power of the distributed photovoltaic power station, thereby obtaining the load power prediction data of the distributed photovoltaic power station. The photovoltaic power generation forecast data and the load power forecast data are used as the power forecast data for the distributed photovoltaic power station.

[0007] Optionally, the photovoltaic power generation prediction sub-model includes the following construction process: The photovoltaic power curves and numerical weather forecast data of the distributed photovoltaic power stations over a historical period are used as input for training data. The output of training data is the photovoltaic power generation sequence of the distributed photovoltaic power station during the historical time period. Based on the input and output of the training data, the LSTM network is trained to obtain a photovoltaic power generation prediction sub-model.

[0008] Optionally, the load power prediction sub-model includes the following construction process: The load time-series correlation data and social behavior characteristic data of the distributed photovoltaic power station during the historical time period are used as training inputs; The regional load power sequence of the distributed photovoltaic power station during the historical time period is used as the training output. The TCN network is trained based on the training input and the training output to obtain a load power prediction sub-model.

[0009] Optionally, determining whether the distributed photovoltaic power station faces a risk of reverse power transmission based on the maximum transmission capacity and the power prediction data includes: Based on the photovoltaic power generation forecast data and load power forecast data in the power forecast data, the net power of the distributed photovoltaic power station is calculated; wherein, the net power is the difference between the photovoltaic power generation forecast value and the load power forecast value; The net power is compared with the maximum transmission capacity of the low-frequency system. If the net power is positive and greater than the maximum transmission capacity, it is determined that the distributed photovoltaic power station has a risk of reverse power transmission. If the net power is negative, or positive but less than or equal to the maximum transmission capacity, it is determined that the distributed photovoltaic power station does not have a risk of reverse power transmission.

[0010] Optionally, the objective function is set to minimize the sum of the total curtailment and the net peak power of the distributed photovoltaic power station; The expression for the objective function is as follows: ; in, Represent the objective function; Indicates minimization; Indicates the weight of the amount of light wasted; Indicates net power weight; Indicates the distributed photovoltaic power station The amount of light discarded at any given moment; Indicates the distributed photovoltaic power station The peak net power at any given moment.

[0011] Optionally, the step of forecasting the reverse power transmission of the distributed photovoltaic power station according to the optimal prevention and control strategy includes: The optimal prevention and control strategy is converted into a power upper limit setting command; According to the power upper limit setting instruction, the power forecasting of the distributed photovoltaic power station is performed using a feedforward feedback composite control mode.

[0012] Optionally, the operating data includes one or more of the following: photovoltaic power generation, load power, grid connection point voltage, grid connection point current, time and environmental information, and low-frequency system operating data; The low-frequency system operating data includes one or more of the following: real-time transmission power of low-frequency lines, low-frequency side voltage fluctuation data, low-frequency current fluctuation data, low-frequency switch station closing state duration, power frequency to low-frequency conversion efficiency of the frequency converter, and low-frequency side power margin.

[0013] Based on the same inventive concept, this invention also provides a distributed photovoltaic power forecasting system based on a low-frequency system, comprising: The data acquisition module is used to acquire the operating data of the distributed photovoltaic power station and the maximum transmission capacity of the low-frequency system connected to the grid with the distributed photovoltaic power station; The power prediction module is used to predict the power of the distributed photovoltaic power station based on the operating data using a pre-built neural network combination model, and obtain the power prediction data of the distributed photovoltaic power station. The risk assessment module is used to determine whether the distributed photovoltaic power station has a risk of reverse power transmission based on the maximum transmission capacity and the power prediction data. The reverse power prediction module is used to solve a pre-set objective function using an optimization control algorithm when the distributed photovoltaic power station has a risk of reverse power transmission, to obtain the optimal prevention and control strategy, and to predict the reverse power transmission of the distributed photovoltaic power station according to the optimal prevention and control strategy. The neural network combination model includes a photovoltaic power generation prediction sub-model and a load power prediction sub-model. The photovoltaic power generation prediction sub-model is constructed based on a long short-term memory (LSTM) network, and the load power prediction sub-model is constructed based on a temporal convolutional network (TCN).

[0014] Optionally, the data acquisition module includes: The initial data acquisition submodule is used to collect the initial power generation data of each photovoltaic field in the distributed photovoltaic power station; wherein, the initial power generation data includes one or more of the following: real-time power generation, photovoltaic module operating temperature and combiner box output current; The status monitoring submodule is used to summarize and monitor the initial power generation data of each photovoltaic field through the low-frequency switch station, integrate the power generation of each photovoltaic field to obtain the total initial power generation of the station, and simultaneously collect the switch on / off status and low-frequency side input power of the low-frequency switch station to obtain the summarized power generation data. The transmission monitoring submodule is used to monitor the transmission process of the aggregated power generation data through a low-frequency line to obtain transmission process data. The power monitoring submodule is used to monitor the power conversion of the transmission process data through the frequency converter and obtain power conversion data. The grid connection monitoring submodule is used to collect real-time voltage, current and power flow direction of the grid connection point through the grid connection point monitoring device of the power frequency grid connection line to obtain grid connection interface data; The integrated verification submodule is used to integrate and verify the summarized power generation data, transmission process data, power conversion data and grid connection interface data to obtain the operation data of the distributed photovoltaic power station.

[0015] Optionally, the power prediction module includes: The power generation prediction submodule is used to predict the power generation of the distributed photovoltaic power station based on the operating data and using the photovoltaic power generation prediction sub-model, so as to obtain the photovoltaic power generation prediction data of the distributed photovoltaic power station. The load power prediction submodule is used to predict the load power of the distributed photovoltaic power station based on the operating data and using the load power prediction sub-model, so as to obtain the load power prediction data of the distributed photovoltaic power station. The prediction output submodule is used to use the photovoltaic power generation prediction data and the load power prediction data as the power prediction data of the distributed photovoltaic power station.

[0016] Optionally, the power prediction module further includes: a power generation model construction module, specifically including: The first input setting submodule is used to use the photovoltaic power curves and numerical weather forecast data of the distributed photovoltaic power station over a historical period as input for training data. The first output setting submodule is used to output the photovoltaic power generation sequence of the distributed photovoltaic power station in the historical time period as training data. The first model training submodule is used to train the LSTM network based on the input and output of the training data to obtain a photovoltaic power generation prediction sub-model.

[0017] Optionally, the power prediction module further includes: a load model construction module, specifically including: The second input setting submodule is used to use the load time-series correlation data and social behavior characteristic data of the distributed photovoltaic power station in the historical time period as training input; The second output setting submodule is used to use the regional load power sequence of the distributed photovoltaic power station in the historical time period as the training output; The second model training submodule is used to train the TCN network based on the training input and the training output to obtain the load power prediction submodel.

[0018] Optionally, the risk assessment module includes: The net power calculation submodule is used to calculate the net power of the distributed photovoltaic power station based on the photovoltaic power generation prediction data and the load power prediction data in the power prediction data; wherein, the net power is the difference between the photovoltaic power generation prediction value and the load power prediction value; The data comparison submodule is used to compare the net power with the maximum transmission capacity of the low-frequency system. The risk output submodule is used to determine that the distributed photovoltaic power station has a risk of reverse power transmission when the net power is positive and greater than the maximum transmission capacity. The risk-free determination submodule is used to determine that there is no risk of reverse power transmission from the distributed photovoltaic power station when the net power is negative or positive but less than or equal to the maximum transmission capacity.

[0019] Optionally, the objective function is set to minimize the sum of the total curtailment and the net peak power of the distributed photovoltaic power station; The expression for the objective function is as follows: ; in, Represent the objective function; Indicates minimization; Indicates the weight of the amount of light wasted; Indicates net power weight; Indicates the distributed photovoltaic power station The amount of light discarded at any given moment; Indicates the distributed photovoltaic power station The peak net power at any given moment.

[0020] Optionally, the reverse forecast module includes: The instruction conversion submodule is used to convert the optimal prevention and control strategy into a power upper limit setting instruction; The power forecasting submodule is used to forecast the power output of the distributed photovoltaic power station using a feedforward-feedback composite control mode based on the power upper limit setting instruction.

[0021] Optionally, the operating data includes one or more of the following: photovoltaic power generation, load power, grid connection point voltage, grid connection point current, time and environmental information, and low-frequency system operating data; The low-frequency system operating data includes one or more of the following: real-time transmission power of low-frequency lines, low-frequency side voltage fluctuation data, low-frequency current fluctuation data, low-frequency switch station closing state duration, power frequency to low-frequency conversion efficiency of the frequency converter, and low-frequency side power margin.

[0022] In another aspect, the present invention also provides an electronic device, comprising: at least one processor and a memory; the memory and the processor are connected via a bus; The memory is used to store one or more programs; When the one or more programs are executed by the at least one processor, a method for predicting the back-feeding power of a distributed photovoltaic power station as described above is implemented.

[0023] In another aspect, the present invention also provides a computer device readable storage medium having an executable program stored thereon, wherein when the executable program is executed, it implements the method for predicting the back-feeding power of a distributed photovoltaic power station as described above.

[0024] Compared with the prior art, the beneficial effects of the present invention are as follows: This invention provides a method and system for predicting backfeed power from a distributed photovoltaic (PV) power plant based on a low-frequency system. The method includes: acquiring operational data of the distributed PV power plant and the maximum transmission capacity of the low-frequency system connected to the grid; predicting the power of the distributed PV power plant using a pre-built neural network combination model based on the operational data to obtain power prediction data; determining whether the distributed PV power plant has a backfeed power risk based on the maximum transmission capacity and the power prediction data; when the distributed PV power plant has a backfeed power risk, solving a pre-set objective function using an optimization control algorithm to obtain an optimal preventive control strategy, and predicting the backfeed power of the distributed PV power plant based on the optimal preventive control strategy; wherein, the neural network combination model includes: PV power generation capacity... The invention comprises a photovoltaic power generation prediction sub-model and a load power prediction sub-model. The photovoltaic power generation prediction sub-model is constructed based on a Long Short-Term Memory (LSTM) network, while the load power prediction sub-model is constructed based on a Temporal Convolutional Network (TCN). This invention utilizes a combined neural network model based on LSTM and TCN networks to perform collaborative power prediction for distributed photovoltaic power plants, which is beneficial for improving power prediction accuracy. Based on high-precision power prediction data, and combined with the maximum transmission capacity of the low-frequency system for backfeed risk assessment, it can achieve forward-looking and accurate identification of potential backfeed power risks, avoiding the lag in post-event response. Combined with an optimization control algorithm based on the objective function, it can prevent backfeed while taking into account economy and safety, reducing curtailment and ensuring grid stability. Ultimately, it achieves accurate backfeed power forecasting through the optimal strategy, which is beneficial for improving the reliability and economy of distributed photovoltaic grid-connected operation. Attached Figure Description

[0025] Figure 1 A flowchart illustrating a method for predicting the reverse power transmission of a distributed photovoltaic power station based on a low-frequency system, provided by the present invention. Figure 2 A schematic diagram illustrating the process of acquiring operational data of a distributed photovoltaic power station in a method for predicting the reverse power transmission of a distributed photovoltaic power station based on a low-frequency system provided by the present invention. Figure 3 A schematic diagram of the structural composition of a forecasting system for the reverse power transmission of a distributed photovoltaic power station provided by the present invention; Figure 4 This is a schematic diagram of the structure of an electronic device provided by the present invention. Detailed Implementation

[0026] This invention proposes a method, system, equipment, and medium for predicting the reverse power transmission of distributed photovoltaic power plants based on low-frequency systems. The specific embodiments of this invention will be further described in detail below with reference to the accompanying drawings.

[0027] Example 1: This invention provides a method for predicting the reverse power transmission of distributed photovoltaic power plants based on low-frequency systems, as illustrated in the flowchart below. Figure 1 As shown, it includes: Step 1: Obtain the operating data of the distributed photovoltaic power station and the maximum transmission capacity of the low-frequency system connected to the grid with the distributed photovoltaic power station; Step 2: Based on the operational data, use a pre-built neural network combination model to predict the power of the distributed photovoltaic power station, and obtain the power prediction data of the distributed photovoltaic power station; Step 3: Based on the maximum transmission capacity and the power prediction data, determine whether the distributed photovoltaic power station has a risk of reverse power transmission; Step 4: When the distributed photovoltaic power station has the risk of reverse power transmission, the optimal control algorithm is used to solve the pre-set objective function to obtain the optimal prevention and control strategy, and the reverse power transmission of the distributed photovoltaic power station is predicted according to the optimal prevention and control strategy. The neural network combination model includes a photovoltaic power generation prediction sub-model and a load power prediction sub-model. The photovoltaic power generation prediction sub-model is constructed based on a Long Short-Term Memory (LSTM) network, and the load power prediction sub-model is constructed based on a Temporal Convolutional Network (TCN). Generally, reverse power forecasting relies on single or separate models, failing to fully incorporate the long-term time-series dependence of photovoltaic power generation and the local characteristics of load fluctuations. This results in insufficient accuracy and poor adaptability of photovoltaic and load power forecasting. Furthermore, traditional methods often neglect the characteristics of low-frequency systems connected to distributed photovoltaic power plants (such as the maximum transmission capacity of low-frequency systems and key parameters like low-frequency power transmission status), relying solely on a simple comparison of the difference between photovoltaic output and local load to assess reverse power risk. This leads to risk assessments detached from the actual transmission capacity of low-frequency grid-connected scenarios, resulting in misjudgments or omissions due to the lack of consideration for low-frequency line carrying capacity limitations. Consequently, these methods fail to meet the accuracy and reliability requirements of reverse power forecasting for distributed photovoltaic power plants in low-frequency systems. To address these technical issues, this invention acquires operational data from distributed photovoltaic power plants and integrates multi-dimensional, scenario-based operational data. This provides comprehensive and reliable data support for subsequent accurate power forecasting based on neural network combined models and for assessing reverse power risk in conjunction with low-frequency system characteristics. For example, the operating data in step 1 above may include one or more of the following: photovoltaic power generation, load power, grid connection point voltage, grid connection point current, time and environmental information, and low-frequency system operating data; The low-frequency system operating data may include one or more of the following: real-time transmission power of low-frequency lines, low-frequency side voltage fluctuation data, low-frequency current fluctuation data, low-frequency switch station closing state duration, power frequency to low-frequency conversion efficiency of the inverter, and low-frequency side power margin.

[0028] To further clarify the specific acquisition path and integration logic of operational data, the data acquisition and processing process can be described step-by-step from the photovoltaic field, low-frequency switching station, and low-frequency line stages. Specifically: like Figure 2 As shown, in one implementation, the process of acquiring the operating data of a distributed photovoltaic power station may include: Collect initial power generation data for each photovoltaic field in the distributed photovoltaic power station (photovoltaic fields 1, 2, and 3 are taken as examples in the attached figure); wherein, the initial power generation data includes one or more of the following: real-time power generation, photovoltaic module operating temperature, and combiner box output current; The initial power generation data of each photovoltaic field is collected and its status is monitored by a low-frequency switching station (where low-frequency switching station 1 corresponds to photovoltaic field 1 and low-frequency switching station 2 corresponds to photovoltaic field 2). For example, the initial power generation data of each photovoltaic field can be collected by a collection line, that is, collection line 1 collects the initial power generation data of photovoltaic field 1 and collection line 2 collects the initial power generation data of photovoltaic field 2. The power generation power of each photovoltaic field is integrated to obtain the total initial power generation power of the field. The switch on / off status and low-frequency input power of the low-frequency switching station are collected simultaneously to obtain the aggregated power generation data. The transmission process data of the aggregated power generation data is obtained by monitoring the transmission process through a low-frequency line. The power conversion data is obtained by using a frequency converter (e.g., an AC-AC frequency converter, which can directly convert power frequency energy into low frequency energy) to monitor the power conversion of the transmission process data (e.g., AC-AC frequency converter 1 monitors the transmission process data of low frequency line 1, and AC-AC frequency converter 2 monitors the transmission process data of low frequency line 2). The grid connection point monitoring device of the power frequency grid connection line collects the real-time voltage, current and power flow direction of the grid connection point (e.g., line 1, 2, 3) to obtain the grid connection interface data; The operating data of the distributed photovoltaic power station is obtained by integrating and verifying the summarized power generation data, transmission process data, power conversion data and grid connection interface data. In this implementation, the micro-characteristics of photovoltaic output are captured by the initial power generation data of each photovoltaic field. By leveraging the aggregated monitoring of low-frequency switch stations, the transmission process monitoring of low-frequency lines, the power conversion monitoring of frequency converters, and the interface data acquisition of grid connection points, a correlation mapping is formed between photovoltaic power generation data and characteristic parameters such as the topology status (switch on / off), transmission capacity (line parameters), conversion efficiency (frequency converter operation), and grid connection interaction (power flow) of the low-frequency system. This correlation mapping enables the integrated operating data to not only possess the comprehensiveness and accuracy of conventional data acquisition, but also to identify hidden deviations (such as abnormal transmission loss of low-frequency lines and fluctuations in frequency converter conversion efficiency) that are difficult to detect in single-stage data acquisition through cross-stage data verification (such as deviation analysis between the initial photovoltaic power generation and the input power of the low-frequency switch station, and matching degree verification between transmission process data and power conversion data). This helps to overcome the limitations of traditional data acquisition, which can only reflect local states and is disconnected from the characteristics of the grid-connected system, thereby achieving a deep characterization and accurate support of the operating status of photovoltaic fields in low-frequency grid-connected scenarios.

[0029] By refining the data acquisition process for distributed photovoltaic (PV) power plants through the above steps, a full-link acquisition can be achieved, from initial data for each PV plant to complete operational data after integration and verification. This helps ensure the comprehensiveness and scenario adaptability of the data. To further develop accurate power prediction based on this operational data, a neural network ensemble model can be considered for prediction. Specifically: In one implementation, step 2 above, which involves using a pre-built neural network combination model to predict the power of the distributed photovoltaic power station based on the operational data, to obtain the power prediction data for the distributed photovoltaic power station, may include: Based on the operational data, the photovoltaic power generation prediction sub-model is used to predict the power generation of the distributed photovoltaic power station, thereby obtaining the photovoltaic power generation prediction data of the distributed photovoltaic power station. Based on the operational data, the load power prediction sub-model is used to predict the load power of the distributed photovoltaic power station, thereby obtaining the load power prediction data of the distributed photovoltaic power station. The photovoltaic power generation prediction data and the load power prediction data are used as the power prediction data for the distributed photovoltaic power station. In this implementation, the photovoltaic power generation prediction sub-model may include the following construction process: The photovoltaic power curves and numerical weather forecast data of the distributed photovoltaic power stations over a historical period are used as input for training data. The output of training data is the photovoltaic power generation sequence of the distributed photovoltaic power station during the historical time period. Based on the input and output of the training data, the LSTM network is trained to obtain a photovoltaic power generation prediction sub-model.

[0030] In this implementation, the load power prediction sub-model may include the following construction process: The load time-series correlation data and social behavior characteristic data of the distributed photovoltaic power station during the historical time period are used as training inputs; The regional load power sequence of the distributed photovoltaic power station during the historical time period is used as the training output. The TCN network is trained based on the training input and the training output to obtain a load power prediction sub-model. In the above implementation, the photovoltaic power generation prediction sub-model focuses on the long-term time-series dependence characteristics of photovoltaic output (such as the continuous influence of changes in irradiance and temperature), while the load power prediction sub-model accurately captures the local abrupt changes in load fluctuations (such as peak electricity consumption periods and load jumps caused by social behavior). This separate modeling avoids the compromise fitting of a single model to two types of heterogeneous characteristics, ensuring that the prediction results reflect both the smooth time-series trend of photovoltaic output and the instantaneous characteristics of load fluctuations. Furthermore, this combined prediction is not a simple superposition of results, but a collaborative output based on unified operational data (including low-frequency system characteristic parameters). This results in a natural correlation between photovoltaic power generation prediction data and load power prediction data with low-frequency grid connection scenarios (e.g., the photovoltaic prediction results implicitly match the low-frequency transmission capacity trend, and the load prediction results include the impact characteristics on low-frequency power absorption). Specifically, the photovoltaic power generation prediction sub-model, based on an LSTM network, correlates the long-term time series of historical photovoltaic power curves with numerical weather prediction data (including trends in meteorological factors such as irradiance and temperature). Using seasonal changes as input, the model can deeply explore the long-term dependence of photovoltaic output on meteorological conditions, overcoming the limitations of traditional models in capturing meteorological time-series characteristics. The load power prediction sub-model uses the TCN network as its core, incorporating historical load time-series correlation data and social behavior characteristic data (including load mutation factors caused by holidays, workdays, and time-based activities) into training. With the help of the TCN's strong ability to capture local time-series characteristics, it accurately depicts the instantaneous change patterns of load fluctuations caused by social behavior. This triple adaptation of network structure, data characteristics, and prediction object characteristics not only enables the prediction accuracy of the two types of sub-models to break through the performance bottleneck of single network or general data input, but also, through the fusion of historical patterns and specific influencing factors in the training data, enables the model to have the ability to generalize prediction of power fluctuations under extreme meteorological conditions (such as sudden strong sunlight) or special social behaviors (such as regional activities). As a result, the output photovoltaic power generation and load power prediction data are not only closer to reality in terms of numerical values, but also deeply consistent with the actual operating rules in low-frequency grid-connected scenarios in terms of change trends and fluctuation characteristics.

[0031] Through the power prediction process based on the neural network combination model described above, we can obtain photovoltaic power generation prediction data and load power prediction data for distributed photovoltaic power plants. To further determine whether there is a risk of reverse power transmission from distributed photovoltaic power plants based on this power prediction data, we can consider calculating the net power and combining it with net power characteristics for risk assessment. Specifically: In one implementation, step 3 above, which involves determining whether the distributed photovoltaic power station faces a risk of reverse power transmission based on the maximum transmission capacity and the power prediction data, may include: Based on the photovoltaic power generation forecast data and load power forecast data in the power forecast data, the net power of the distributed photovoltaic power station is calculated; wherein, the net power is the difference between the photovoltaic power generation forecast value and the load power forecast value; The net power is compared with the maximum transmission capacity of the low-frequency system. If the net power is positive and greater than the maximum transmission capacity, it is determined that the distributed photovoltaic power station has a risk of reverse power transmission. If the net power is negative, or positive but less than or equal to the maximum transmission capacity, it is determined that the distributed photovoltaic power station does not have a risk of reverse power transmission.

[0032] This implementation method, by comparing net power with the maximum transmission capacity of the low-frequency system in a hierarchical manner, overcomes the one-sidedness of judging risk solely based on the positive or negative net power. It upgrades risk assessment from a qualitative analysis of whether a backfeeding trend exists to a quantitative assessment of whether the backfeeding power exceeds the system's carrying capacity. This clearly defines the boundary between the tolerable positive net power and the excess backfeeding power that needs to be controlled. This quantitative judgment logic avoids blindly suppressing all positive net power, only formulating control measures for the portion exceeding the capacity threshold. This not only helps ensure grid security but also reserves space for improving photovoltaic absorption efficiency. Simultaneously, the strong binding of this implementation method with the characteristics of the low-frequency system makes the risk assessment results naturally adaptable to the equipment operating logic, providing a precise basis for translating control strategies into specific instructions such as inverter power adjustment and switch station status adjustment. This facilitates seamless integration of risk identification and equipment control, ultimately improving the targeting and economy of backfeeding power control in low-frequency grid-connected scenarios.

[0033] Through the above steps of assessing the risk of reverse power transmission based on power prediction data, it can be clearly determined whether a distributed photovoltaic power station faces reverse power transmission risk. To further achieve precise and effective prevention and control when such risk exists, it is possible to consider setting an objective function and using an optimization control algorithm to solve for the optimal strategy, thereby completing the reverse power transmission forecast. Specifically: In one implementation, the objective function in step 4 above can be set to minimize the sum of the total curtailment and the net peak power of the distributed photovoltaic power station; For example, the expression for the objective function can be as follows: ; in, Represent the objective function; Indicates minimization; Indicates the weight of the amount of light wasted; Indicates net power weight; Indicates the distributed photovoltaic power station The amount of light discarded at any given moment; Indicates the distributed photovoltaic power station The net power peak at any given moment; in this example, the inclusion of total curtailment is directly related to the utilization efficiency of photovoltaic resources (the smaller the curtailment, the more fully photovoltaics are absorbed, and the higher the economic efficiency), while the constraint of net power peak directly addresses the source of reverse power risk (the lower the peak value, the lower the possibility of exceeding the transmission capacity of low-frequency systems, and the better the safety), while the weighting coefficient and The introduction of this feature allows for dynamic adjustment of scenario parameters such as the carrying capacity of low-frequency systems and grid dispatch requirements, enabling the objective function to both reduce curtailment (increase power consumption) and mitigate solar curtailment when photovoltaic power absorption pressure is high. Furthermore, it can focus on controlling the net power peak when the system security risk is high (increasing) This multi-objective weighted optimization logic not only avoids the contradiction of excessively allowing the peak net power to be used in pursuit of economy under a single objective, which leads to the risk of back transmission, or excessive curtailment of light to ensure safety, which leads to the waste of resources, but also implicitly links the peak net power with the characteristics of the low-frequency system (the constraint of the maximum transmission capacity of the low-frequency system on the net power needs to be adapted during the solution of the objective function). This makes the optimization result naturally adapt to the actual operating boundary of the low-frequency grid connection scenario, and the final output objective function solution has a quantitative basis for balancing economy and safety. In one implementation, when the distributed photovoltaic power station faces the risk of reverse power transmission, the process of solving a pre-set objective function using an optimization control algorithm to obtain the optimal prevention and control strategy may include: Based on power forecast data, the photovoltaic power generation and load power forecast sequences for a future preset period are extracted to determine the initial range of curtailment and the control threshold of net power peak. Simultaneously, considering the current transmission load status of the low-frequency system, the weight of curtailment in the objective function is dynamically calibrated using the analytic hierarchy process (AHP). α With net power weight βThis ensures that the weight parameters are adapted to the system's operating status in real time; and it also constructs multi-dimensional constraints, including low-frequency system transmission constraints (line power does not exceed the maximum capacity, inverter conversion power matches transmission power), photovoltaic regulation constraints (upper limit of curtailment and change rate threshold), and grid connection interface constraints (net power flow direction and voltage and current fluctuation range), forming the boundary conditions for solving the problem. An improved particle swarm optimization algorithm is used for iterative solution. The amount of light wasted in each time period is used as the optimization variable. The objective function with constraint penalty term is used as the fitness function. The search range and convergence accuracy are balanced by dynamically adjusting the inertial weight. The iteration stops when the fitness value continuously meets the convergence condition, and the corresponding optimal light wasted sequence is output. The sequence is transformed into the photovoltaic inverter power limit and inverter regulation command, which are then substituted into the photovoltaic-low-frequency system joint simulation model. The effectiveness of the strategy is verified through extreme scenario data. If constraint violations are found, the parameters are adjusted and re-optimized. Finally, the optimal preventive control strategy that meets both safety and economic requirements is output. In this implementation, a precise balance between safety and economy is achieved through dynamic weight calibration. Multi-dimensional constraint fusion makes the optimization process naturally adaptable to the transmission characteristics of the low-frequency system and the equipment regulation boundary. The iterative mechanism of the improved particle swarm algorithm takes into account both global optimization and local convergence accuracy. The strategy verification stage ensures the engineering reliability of the output results under extreme scenarios. This ultimately achieves a leap from passive response to active adaptation in the prevention and control of backfeeding power. It is beneficial to avoid resource waste caused by excessive curtailment of solar power and prevent system risks caused by net power exceeding limits. In the end, the optimal preventive control strategy has both theoretical optimality and practical operational feasibility.

[0034] In one implementation, the process of forecasting the reverse power transmission of the distributed photovoltaic power station based on the optimal prevention and control strategy may include: The optimal prevention and control strategy is converted into a power upper limit setting command; According to the power upper limit setting instruction, the power forecast of the distributed photovoltaic power station is performed by adopting a feedforward feedback composite control mode. In this implementation, the power upper limit setting command is directly linked to the adjustment parameters of the core equipment of the low-frequency system (such as the power frequency to low frequency conversion device), so that the abstract optimization strategy is transformed into a quantitative control benchmark that the equipment can execute (such as the conversion power upper limit and frequency fine-tuning threshold), ensuring the consistency between the strategy implementation and the operating logic of the low-frequency equipment; the feedforward control issues commands in advance based on power prediction data, which can predict the trend changes of photovoltaic output and load fluctuations and avoid sudden back-feeding power due to lagging adjustment; the feedback control collects actual operating data such as power flow at the grid connection point and transmission status of low-frequency lines in real time, and dynamically corrects the power upper limit command to offset the deviation caused by prediction errors and system disturbances (such as sudden changes in irradiance and load jumps). This composite control mode not only realizes the closed loop of prediction-decision-execution-correction, but also, through the synergy of feedforward and feedback, enables the control process to adapt to the transmission delay characteristics of low-frequency systems (feedforward early response) and be compatible with the equipment adjustment accuracy limitations (feedback dynamic calibration). Thus, while ensuring the accuracy of the reverse power forecast, it reduces the mechanical losses and power fluctuations caused by frequent adjustments of low-frequency equipment, and improves the stability and economy of the coordinated operation of distributed photovoltaic power plants and low-frequency systems.

[0035] In summary, this invention addresses the technical problems of insufficient accuracy and poor adaptability of traditional single or separate models in photovoltaic and load power prediction. Furthermore, these models fail to consider the characteristics of low-frequency systems connected to distributed photovoltaic (PV) power plants (including the maximum transmission capacity of the low-frequency system and low-frequency side operating data), leading to incomplete judgment of backfeed power risk and an inability to meet the backfeed power forecasting requirements in low-frequency grid-connected scenarios. This invention proposes a backfeed power forecasting method for distributed PV power plants based on low-frequency systems. By constructing a combined neural network model of LSTM and TCN, it facilitates high-precision prediction of PV power generation and load power. Based on this prediction data and the maximum transmission capacity of the low-frequency system, backfeed power risk can be identified. After identifying the risk, an optimal preventive control strategy is generated by solving an optimized objective function, and the strategy is ultimately transformed into control commands. Moreover, this invention can use the frequency converter as the core control object, directly and quickly adjusting its output power upper limit, forming a closed-loop execution method from accurate prediction and risk warning to source control. This effectively suppresses backfeed power at the source, thereby improving the reliability and economy of distributed PV grid-connected operation.

[0036] Example 2: Based on the same inventive concept, this invention also provides a distributed photovoltaic power forecasting system based on a low-frequency system, the structural composition of which is shown in the schematic diagram below. Figure 3 As shown, it includes: The data acquisition module is used to acquire the operating data of the distributed photovoltaic power station and the maximum transmission capacity of the low-frequency system connected to the grid with the distributed photovoltaic power station; The power prediction module is used to predict the power of the distributed photovoltaic power station based on the operating data using a pre-built neural network combination model, and obtain the power prediction data of the distributed photovoltaic power station. The risk assessment module is used to determine whether the distributed photovoltaic power station has a risk of reverse power transmission based on the maximum transmission capacity and the power prediction data. The reverse power prediction module is used to solve a pre-set objective function using an optimization control algorithm when the distributed photovoltaic power station has a risk of reverse power transmission, to obtain the optimal prevention and control strategy, and to predict the reverse power transmission of the distributed photovoltaic power station according to the optimal prevention and control strategy. The neural network combination model includes a photovoltaic power generation prediction sub-model and a load power prediction sub-model. The photovoltaic power generation prediction sub-model is constructed based on a Long Short-Term Memory (LSTM) network, and the load power prediction sub-model is constructed based on a Temporal Convolutional Network (TCN). For example, the operating data may include one or more of the following: photovoltaic power generation, load power, grid connection voltage, grid connection current, and time and environmental information.

[0037] In one implementation, the data acquisition module may include: The initial acquisition submodule is used to acquire the initial power generation data of each photovoltaic field in the distributed photovoltaic power station; wherein, the initial power generation data includes one or more of the following: real-time power generation, photovoltaic module operating temperature and combiner box output current; The status monitoring submodule is used to summarize and monitor the initial power generation data of each photovoltaic field through the low-frequency switch station, integrate the power generation of each photovoltaic field to obtain the total initial power generation of the station, and simultaneously collect the switch on / off status and low-frequency side input power of the low-frequency switch station to obtain the summarized power generation data. The transmission monitoring submodule is used to monitor the transmission process of the aggregated power generation data through a low-frequency line to obtain transmission process data. The power monitoring submodule is used to monitor the power conversion of the transmission process data through the frequency converter and obtain power conversion data. The grid connection monitoring submodule is used to collect real-time voltage, current and power flow direction of the grid connection point through the grid connection point monitoring device of the power frequency grid connection line to obtain grid connection interface data; The integrated verification submodule is used to integrate and verify the summarized power generation data, transmission process data, power conversion data and grid connection interface data to obtain the operation data of the distributed photovoltaic power station.

[0038] In one implementation, the power prediction module may include: The power generation prediction submodule is used to predict the power generation of the distributed photovoltaic power station based on the operating data and using the photovoltaic power generation prediction sub-model, so as to obtain the photovoltaic power generation prediction data of the distributed photovoltaic power station. The load power prediction submodule is used to predict the load power of the distributed photovoltaic power station based on the operating data and using the load power prediction sub-model, so as to obtain the load power prediction data of the distributed photovoltaic power station. The prediction output submodule is used to use the photovoltaic power generation prediction data and the load power prediction data as the power prediction data of the distributed photovoltaic power station.

[0039] In one implementation, the power prediction module may further include: a power generation model construction module, which may specifically include: The first input setting submodule is used to use the photovoltaic power curves and numerical weather forecast data of the distributed photovoltaic power station over a historical period as input for training data. The first output setting submodule is used to output the photovoltaic power generation sequence of the distributed photovoltaic power station in the historical time period as training data. The first model training submodule is used to train the LSTM network based on the input and output of the training data to obtain a photovoltaic power generation prediction sub-model.

[0040] In one implementation, the power prediction module may further include: a load model construction module, which may specifically include: The second input setting submodule is used to use the load time-series correlation data and social behavior characteristic data of the distributed photovoltaic power station in the historical time period as training input; The second output setting submodule is used to use the regional load power sequence of the distributed photovoltaic power station in the historical time period as the training output; The second model training submodule is used to train the TCN network based on the training input and the training output to obtain the load power prediction submodel.

[0041] In one implementation, the risk assessment module may include: The net power calculation submodule is used to calculate the net power of the distributed photovoltaic power station based on the photovoltaic power generation prediction data and the load power prediction data in the power prediction data; wherein, the net power is the difference between the photovoltaic power generation prediction value and the load power prediction value; The data comparison submodule is used to compare the net power with the maximum transmission capacity of the low-frequency system. The risk output submodule is used to determine that the distributed photovoltaic power station has a risk of reverse power transmission when the net power is positive and greater than the maximum transmission capacity. The risk-free determination submodule is used to determine that there is no risk of reverse power transmission from the distributed photovoltaic power station when the net power is negative or positive but less than or equal to the maximum transmission capacity.

[0042] In one implementation, the objective function can be set to minimize the sum of the total curtailment and the net peak power of the distributed photovoltaic power station; For example, the expression for the objective function can be as follows: ; in, Represent the objective function; Indicates minimization; Indicates the weight of the amount of light wasted; Indicates net power weight; Indicates the distributed photovoltaic power station The amount of light discarded at any given moment; Indicates the distributed photovoltaic power station The peak net power at any given moment.

[0043] In one implementation, the reverse forecast module may include: The instruction conversion submodule is used to convert the optimal prevention and control strategy into a power upper limit setting instruction; The power forecasting submodule is used to forecast the power output of the distributed photovoltaic power station using a feedforward-feedback composite control mode based on the power upper limit setting instruction.

[0044] Example 3: like Figure 4 As shown, the present invention also provides an electronic device, which may be a computer device, a microcontroller device, a smart mobile device, etc. The electronic device in this embodiment may include a processor, a memory, a transceiver component, etc. The memory, processor, and transceiver component are connected via a bus; the memory can be used to store executable programs, and an exemplary executable program may include instructions; the processor is used to execute the instructions stored in the memory. The memory can also be used to store data, which can be accessed and / or modified when instructions are executed.

[0045] The processor may be a Central Processing Unit (CPU), or it may be other general-purpose processors, digital signal processors (DSPs), application-specific integrated circuits (ASICs), field-programmable gate arrays (FPGAs), or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. It is the computing and control core of the terminal, and it is suitable for implementing one or more instructions. Specifically, it is suitable for loading and executing one or more instructions in the storage medium to realize the corresponding method flow or corresponding function, so as to realize the steps of the distributed photovoltaic power forecasting method based on a low-frequency system in the above embodiments.

[0046] Example 4: Based on the same inventive concept, this invention also provides a readable storage medium, specifically an electronic device readable storage medium (Memory). This readable storage medium is a memory device within an electronic device used to store programs and data. It is understood that the storage medium here can include both built-in storage media within the electronic device and extended storage media supported by the electronic device. The storage medium provides storage space, which stores the terminal's operating system. Furthermore, this storage space also stores one or more instructions suitable for loading and execution by a processor. These instructions can be one or more executable programs (including program code). It should be noted that the storage medium here can be high-speed RAM or non-volatile memory, such as at least one disk storage device. Loading and executing one or more instructions stored in the storage medium by the processor can implement the steps of the distributed photovoltaic power forecasting method based on a low-frequency system in the above embodiments.

[0047] Those skilled in the art will understand that embodiments of the present invention can be provided as methods, systems, or computer program products. Therefore, the present invention can take the form of a completely hardware embodiment, a completely software embodiment, or an embodiment combining software and hardware aspects. Furthermore, the present invention can take the form of a computer program product embodied on one or more computer-usable storage media (including, but not limited to, disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0048] This invention is described with reference to flowchart illustrations and / or block diagrams of methods, apparatus (systems), and computer program products according to embodiments of the invention. It will be understood that each block of the flowchart illustrations and / or block diagrams, and combinations of blocks in the flowchart illustrations and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, special-purpose computer, embedded processor, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, generate instructions for implementing the flowchart illustrations and / or block diagrams. Figure 1 One or more processes and / or boxes Figure 1 A device that provides the functions specified in one or more boxes.

[0049] These computer program instructions may also be stored in a computer-readable storage medium that can direct a computer or other programmable data processing device to function in a particular manner, such that the instructions stored in the computer-readable storage medium produce an article of manufacture including instruction means, which are implemented in a process Figure 1 One or more processes and / or boxes Figure 1 The function specified in one or more boxes.

[0050] These computer program instructions may also be loaded onto a computer or other programmable data processing equipment to cause a series of operational steps to be performed on the computer or other programmable equipment to produce a computer-implemented process, thereby providing instructions that execute on the computer or other programmable equipment for implementing the process. Figure 1 One or more processes and / or boxes Figure 1 The steps of the function specified in one or more boxes.

[0051] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and not to limit its scope of protection. Although the present invention has been described in detail with reference to the above embodiments, those skilled in the art should understand that after reading the present invention, they can still make various changes, modifications or equivalent substitutions to the specific implementation methods of the application, but these changes, modifications or equivalent substitutions are all within the scope of protection of the claims pending approval.

Claims

1. A method for predicting the reverse power transmission of distributed photovoltaic power plants based on low-frequency systems, characterized in that, include: Obtain the operating data of the distributed photovoltaic power station and the maximum transmission capacity of the low-frequency system connected to the grid with the distributed photovoltaic power station; Based on the operational data, a pre-built neural network combination model is used to predict the power of the distributed photovoltaic power station, thereby obtaining the power prediction data of the distributed photovoltaic power station. Based on the maximum transmission capacity and the power prediction data, it is determined whether the distributed photovoltaic power station has the risk of reverse power transmission; When the distributed photovoltaic power station faces the risk of reverse power transmission, an optimization control algorithm is used to solve the pre-set objective function to obtain the optimal prevention and control strategy, and the reverse power transmission of the distributed photovoltaic power station is predicted based on the optimal prevention and control strategy. The neural network combination model includes a photovoltaic power generation prediction sub-model and a load power prediction sub-model. The photovoltaic power generation prediction sub-model is constructed based on a long short-term memory (LSTM) network, and the load power prediction sub-model is constructed based on a temporal convolutional network (TCN).

2. The method as described in claim 1, characterized in that, The acquisition of operational data from distributed photovoltaic power stations includes: Collect initial power generation data of each photovoltaic field in the distributed photovoltaic power station; wherein, the initial power generation data includes one or more of the following: real-time power generation, photovoltaic module operating temperature and combiner box output current; The initial power generation data of each photovoltaic field is collected and its status is monitored by the low-frequency switching station. The power generation power of each photovoltaic field is integrated to obtain the total initial power generation power of the station. The switching status and low-frequency input power of the low-frequency switching station are collected simultaneously to obtain the aggregated power generation data. The transmission process data of the aggregated power generation data is obtained by monitoring the transmission process through a low-frequency line. Power conversion data is obtained by monitoring the power conversion of the transmission process data using a frequency converter. The grid connection point monitoring device of the power frequency power grid connection line collects the real-time voltage, current and power flow direction of the grid connection point to obtain the grid connection interface data; The operating data of the distributed photovoltaic power station is obtained by integrating and verifying the summarized power generation data, transmission process data, power conversion data and grid connection interface data.

3. The method as described in claim 1, characterized in that, The step of using a pre-built neural network combination model to predict the power of the distributed photovoltaic power station based on the operational data to obtain the power prediction data of the distributed photovoltaic power station includes: Based on the operational data, the photovoltaic power generation prediction sub-model is used to predict the power generation of the distributed photovoltaic power station, thereby obtaining the photovoltaic power generation prediction data of the distributed photovoltaic power station. Based on the operational data, the load power prediction sub-model is used to predict the load power of the distributed photovoltaic power station, thereby obtaining the load power prediction data of the distributed photovoltaic power station. The photovoltaic power generation forecast data and the load power forecast data are used as the power forecast data for the distributed photovoltaic power station.

4. The method as described in claim 3, characterized in that, The photovoltaic power generation prediction sub-model includes the following construction process: The photovoltaic power curves and numerical weather forecast data of the distributed photovoltaic power stations over a historical period are used as input for training data. The output of training data is the photovoltaic power generation sequence of the distributed photovoltaic power station during the historical time period. Based on the input and output of the training data, the LSTM network is trained to obtain a photovoltaic power generation prediction sub-model.

5. The method as described in claim 4, characterized in that, The load power prediction sub-model includes the following construction process: The load time-series correlation data and social behavior characteristic data of the distributed photovoltaic power station during the historical time period are used as training inputs; The regional load power sequence of the distributed photovoltaic power station during the historical time period is used as the training output. The TCN network is trained based on the training input and the training output to obtain a load power prediction sub-model.

6. The method as described in claim 3, characterized in that, The step of determining whether the distributed photovoltaic power station faces a risk of reverse power transmission based on the maximum transmission capacity and the power prediction data includes: Based on the photovoltaic power generation forecast data and load power forecast data in the power forecast data, the net power of the distributed photovoltaic power station is calculated; wherein, the net power is the difference between the photovoltaic power generation forecast value and the load power forecast value; The net power is compared with the maximum transmission capacity of the low-frequency system. If the net power is positive and greater than the maximum transmission capacity, it is determined that the distributed photovoltaic power station has a risk of reverse power transmission. If the net power is negative, or positive but less than or equal to the maximum transmission capacity, it is determined that the distributed photovoltaic power station does not have a risk of reverse power transmission.

7. The method as described in claim 1, characterized in that, The objective function is set to minimize the sum of the total curtailment and the net peak power of the distributed photovoltaic power station. The expression for the objective function is as follows: ； in, Represent the objective function; Indicates minimization; Indicates the weight of the amount of light wasted; Indicates net power weight; Indicates the distributed photovoltaic power station The amount of light discarded at any given moment; Indicates the distributed photovoltaic power station The peak net power at any given moment.

8. The method as described in claim 1, characterized in that, The step of forecasting the reverse power transmission of the distributed photovoltaic power station according to the optimal prevention and control strategy includes: The optimal prevention and control strategy is converted into a power upper limit setting command; According to the power upper limit setting instruction, the power forecasting of the distributed photovoltaic power station is performed using a feedforward feedback composite control mode.

9. The method as described in claim 1, characterized in that, The operating data includes one or more of the following: photovoltaic power generation, load power, grid connection point voltage, grid connection point current, time and environmental information, and low-frequency system operating data; The low-frequency system operating data includes one or more of the following: real-time transmission power of low-frequency lines, low-frequency side voltage fluctuation data, low-frequency current fluctuation data, low-frequency switch station closing state duration, power frequency to low-frequency conversion efficiency of the frequency converter, and low-frequency side power margin.

10. A distributed photovoltaic power forecasting system based on a low-frequency system, characterized in that, include: The data acquisition module is used to acquire the operating data of the distributed photovoltaic power station and the maximum transmission capacity of the low-frequency system connected to the grid with the distributed photovoltaic power station; The power prediction module is used to predict the power of the distributed photovoltaic power station based on the operating data using a pre-built neural network combination model, and obtain the power prediction data of the distributed photovoltaic power station. The risk assessment module is used to determine whether the distributed photovoltaic power station has a risk of reverse power transmission based on the maximum transmission capacity and the power prediction data. The reverse power prediction module is used to solve a pre-set objective function using an optimization control algorithm when the distributed photovoltaic power station has a risk of reverse power transmission, to obtain the optimal prevention and control strategy, and to predict the reverse power transmission of the distributed photovoltaic power station according to the optimal prevention and control strategy. The neural network combination model includes a photovoltaic power generation prediction sub-model and a load power prediction sub-model. The photovoltaic power generation prediction sub-model is constructed based on a long short-term memory (LSTM) network, and the load power prediction sub-model is constructed based on a temporal convolutional network (TCN).

Images