An industrial control system for photovoltaic power generation energy saving management
Through cross-domain collaborative control of the perception layer, edge collaboration layer, and central decision-making layer, the problems of energy consumption of auxiliary equipment, IoT energy consumption, and heat loss of the power collection network in photovoltaic power plants are solved, thereby improving the overall power generation efficiency of photovoltaic power plants.
Patent Information
- Application Number
- CN202611457241.8
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2026-09-18
- Publication Date
- 2026-11-17
AI Technical Summary
Existing industrial automation control systems for photovoltaic power plants have failed to effectively reduce the mechanical and thermal energy consumption of auxiliary equipment, the communication and computing energy consumption of industrial IoT, and the heat loss caused by uneven power distribution in the power collection network, resulting in low overall power generation efficiency.
The architecture consists of a perception layer, an edge collaboration layer, and a central decision-making layer. It achieves cross-domain collaborative closed-loop control by using a digital twin auxiliary machine optimization module, a network adaptive sleep scheduling module, and a spatial impedance matrix heat loss suppression module, thereby optimizing the energy consumption of auxiliary machines, the energy consumption of information networks, and the heat loss of power collection lines.
Without compromising monitoring accuracy, this approach aims to reduce waste of plant power caused by ineffective solar tracking and heat dissipation, decrease fatigue wear of auxiliary mechanical components, reduce IoT energy consumption, optimize heat loss in the power collection network, and improve the net power generation efficiency of photovoltaic power plants.
Smart Images

Figure CN123113579A_ABST
Abstract
Description
Technical Field
[0001] This invention relates to the field of industrial automation control technology for photovoltaic power plants, specifically to an industrial control system for energy-saving management of photovoltaic power generation. It is used to coordinate and manage the mechanical and thermal energy consumption of auxiliary equipment in large photovoltaic power plants, the communication and computing power consumption of the industrial Internet of Things itself, and the heat loss caused by uneven power distribution in the power collection network. It belongs to the cross-technology direction of industrial automation control of photovoltaic power plants, industrial Internet of Things computing power and green communication industry, and energy-saving and carbon-reduction control of new energy smart power plants. Background Technology
[0002] Large-scale photovoltaic power plants typically employ distributed control systems or supervisory control and data acquisition systems as the core of the station control layer. These systems connect to weather stations, tracking bracket controllers, inverters, edge computing gateways, and temperature and current measurement devices on the collector lines via industrial Ethernet to manage array power generation. In existing, relatively similar solutions, the tracking bracket controller triggers motor tracking actions solely based on changes in the local irradiance sensor signal, without calculating the auxiliary power consumed by the motor's actions. The inverter cooling fan starts and stops in stages according to preset temperature thresholds, constituting local closed-loop passive control. The edge computing gateway and sensor nodes maintain a fixed sampling frequency and communication topology, failing to adjust according to different operating conditions such as array operation, steady state, or nighttime sleep. The automatic power generation control module receives power rationing signals. After receiving the instruction, power is allocated according to the rated capacity of each inverter, without considering the impedance differences between the inverters and the booster station. In the above scheme, the physical equipment control, industrial IoT communication and computing resource scheduling, and power allocation of the collection network operate independently and are unaware of each other. This may result in negative net benefits from auxiliary machine operation under conditions of severe light fluctuations. The energy consumption of the industrial IoT itself remains at a high level for a long time, and the high impedance lines at the far end will generate significant Joule heat loss under the condition of power generation limitation. Overall, this restricts the improvement of the power station's net power generation efficiency. In addition, each control subsystem operates with its own independent closed-loop logic, lacking a cross-domain data sharing mechanism, making it difficult to comprehensively optimize the power station's net energy efficiency from the three dimensions of auxiliary machine energy consumption, information network energy consumption, and collection line heat loss. Summary of the Invention
[0003] The technical problem this invention aims to solve is that existing photovoltaic power plant industrial automation control systems only focus on increasing array power generation, neglecting the mechanical and thermal energy consumption of auxiliary equipment such as tracking brackets and cooling fans, the communication and computing power consumption of the massive industrial IoT itself, and the line heat loss caused by the failure to consider line impedance differences in power distribution under limited or uneven lighting conditions. This results in a low overall net power generation efficiency of the power plant. Specifically, the following three issues need to be addressed: First, how to determine whether the auxiliary equipment operation has a positive net benefit under fluctuating lighting conditions to avoid wasting plant power due to ineffective tracking; second, how to make the communication and computing resources of the industrial IoT adaptively adjust according to the operating conditions of physical equipment to reduce the energy consumption of the information system itself; and third, how to perform differentiated scheduling based on the actual electrical impedance differences of the collection lines under limited or uneven lighting conditions to reduce Joule heat loss caused by high current long-distance transmission.
[0004] To solve the above-mentioned technical problems, the present invention provides an industrial control system for photovoltaic power generation energy-saving management, including a sensing layer, an edge collaboration layer and a central decision layer. The sensing layer includes a meteorological irradiance sensing unit, a tracking bracket and its drive motor, an inverter and its cooling fan, a collector line temperature and current measurement sensor, an edge computing gateway and a sensing node.
[0005] The edge collaboration layer includes a software-defined network controller and an edge computing resource pool;
[0006] The central decision-making layer is deployed on the station control center server. It interfaces with the existing industrial control system and automatic power generation control module in the station. It is equipped with a digital twin auxiliary machine optimization module, a network adaptive sleep scheduling module, and a spatial impedance matrix heat loss suppression module. The digital twin auxiliary machine optimization module, based on micro-meteorological forecast data and auxiliary equipment energy consumption model, performs joint simulation of the power generation gain and power consumption generated by the tracking and heat dissipation actions performed by the auxiliary equipment to obtain the net power generation index. Based on this, it determines whether to issue an action command or a sleep command to the corresponding auxiliary equipment.
[0007] The network adaptive sleep scheduling module drives the software-defined network controller to reconstruct the data acquisition network topology and schedule the computing power allocation of the edge computing resource pool based on the operating conditions of the physical devices in the perception layer.
[0008] The spatial impedance matrix heat loss suppression module is based on the dynamic electrical impedance topology matrix of the entire station's collection lines. When receiving the grid's power limiting command, it constructs a power allocation optimization model aimed at reducing the heat loss of the collection network, solves the power allocation command for each inverter, and issues it. The above three modules realize data interaction through a unified data bus, forming a cross-domain collaborative closed loop.
[0009] The beneficial effects of this invention are as follows:
[0010] Firstly, by jointly simulating the power generation gain and power consumption before the auxiliary machine performs its actions and determining the net power generation, the auxiliary machine can be put into hibernation when the net benefit is insufficient, thereby reducing the waste of plant power caused by ineffective tracking and ineffective heat dissipation, and reducing the fatigue wear of auxiliary machine mechanical parts caused by frequent start-stop.
[0011] Secondly, by adaptively reconstructing the network topology and computing power scheduling based on the operating conditions of physical equipment, the communication and computing power energy consumption of the Industrial Internet of Things can be reduced without reducing the necessary monitoring accuracy, thus avoiding the energy consumption of the information system offsetting the energy-saving results on the physical side.
[0012] Third, by constructing a dynamic electrical impedance topology matrix and redistributing the limited power according to the principle of electrical distance priority, the heat loss of the entire power collection network can be reduced and the power conversion efficiency improved while meeting the total power constraint of the power grid. The three modules operate in coordination through a unified data bus, enabling the overall optimization of three previously overlooked dimensions: auxiliary machine energy consumption, information network energy consumption, and heat loss of the collection lines, thereby improving the net power generation efficiency of the photovoltaic power station as a whole. Attached Figure Description
[0013] Figure 1 This is a flowchart of the digital twin auxiliary machine optimization module of the present invention;
[0014] Figure 2 This is a flowchart illustrating the network adaptive sleep scheduling module of the present invention.
[0015] Figure 3 This is a flowchart illustrating the operation of the thermal loss suppression module of the spatial impedance matrix in this invention. Detailed Implementation
[0016] The following is in conjunction with the appendix Figure 1 -Appendix Figure 3The technical solution of the present invention will be further explained in detail with specific implementation methods. The industrial control system for photovoltaic power generation energy-saving management provided by the present invention is divided into three layers in its overall architecture: a perception layer, an edge collaboration layer, and a central decision-making layer. The perception layer is deployed at the photovoltaic power station site and includes the irradiance sensing unit and cloud image acquisition device of the meteorological station, the bifacial or flat single-axis tracking bracket and its drive motor, the inverter and its cooling fan, the temperature and current measurement sensors of the collector line, the edge computing gateway and sensing nodes, which are used to collect raw data reflecting the array power generation, auxiliary machine operation and line status. The edge collaboration layer includes a software-defined network controller and an edge computing resource pool, which is responsible for the central decision-making layer. The decision-making layer executes network topology reconstruction and computing task offloading under the scheduling instructions. The central decision-making layer is deployed on the station control center server and contains a digital twin model library, a micro-meteorological prediction sub-model, a model predictive control optimization engine, a network hibernation scheduling engine, and a spatial impedance matrix optimization solver. It also interfaces with the existing industrial control system and automatic power generation control module in the station through a standard interface. There are both bottom-up data acquisition paths and top-down control command execution paths between the three layers. The three major functional modules in the central decision-making layer share array operating conditions, meteorological forecasts, network status, and electrical topology data through a unified data bus, forming a cross-domain collaborative closed loop.
[0017] (a) Digital twin-driven optimization mechanism for auxiliary machine net power generation
[0018] The digital twin referred to in this invention refers to a virtual simulation model constructed in a central server that is mapped in real time to the physical entities of a photovoltaic power station (including arrays, tracking brackets, inverters, and collector lines). It is used to synchronously reflect the operating status of the physical equipment and support predictive simulation. The model predictive control refers to a closed-loop control method that predicts the output over a future period based on the system model and solves the optimal control sequence through rolling optimization. The net power generation refers to the difference between the power generation gain brought about by the auxiliary equipment performing a certain action and the auxiliary power consumed by the action itself.
[0019] In this embodiment, the micro-meteorological prediction sub-model receives real-time irradiance sequences and cloud image data uploaded by the meteorological station, identifies the movement trajectory and speed of fragmented clouds, and outputs a progressively longer irradiance prediction sequence for the future prediction time domain. The digital twin model, based on the current tracking bracket tilt angle, inverter operating status, and ambient temperature, combines the aforementioned irradiance prediction sequence to simulate and calculate the power generation gain that can be obtained by performing tracking or cooling actions at each step within the prediction time domain, denoted as... And the auxiliary power consumed in performing this action, denoted as Based on this, the model predictive control optimization engine constructs the following net power generation objective function, and solves for the control sequence that increases the value of the objective function in the prediction time domain:
[0020] ;
[0021] In the formula, The objective function value, To predict the number of steps in the time domain, For the prediction step index, For the first The power generation gain from the execution of each step, whether for beam tracking or heat dissipation. For the first The auxiliary power consumed in performing this action. This is the weighting factor for net power generation gain. The penalty weighting coefficient for controlling motion smoothness is used to suppress frequent start-stop of auxiliary equipment. For the first The change in the step control quantity (tracking angle adjustment quantity or fan speed adjustment quantity) relative to the previous step can be mitigated by introducing a penalty term for the change in the control quantity of adjacent predicted steps into the objective function. This can suppress frequent mechanical actions of auxiliary equipment caused by short-term irradiance fluctuations while ensuring net benefits.
[0022] Subsequently, the central decision-making body calculated the net power generation indicator for the current step:
[0023] ;
[0024] In the formula, For the first The net gain power of the step will Compared with the preset economic threshold Comparison: In Less than In this situation, the central decision-making level issues a sleep command to the corresponding tracking motor or cooling fan, suspending the optimization of the auxiliary equipment's operation. Not less than In the case of [missing information], the corresponding action command is issued according to the control sequence obtained by model predictive control solution. After the action is executed, the system collects actual power generation and energy consumption feedback data to correct the relevant parameters of the digital twin model. The above joint simulation and judgment process is repeated in the next sampling period to form rolling time-domain closed-loop optimization.
[0025] In another embodiment of the present invention, the digital twin auxiliary machine optimization module can also use a machine learning regression model trained based on historical operating data to directly use real-time irradiance, cloud movement speed and current tracking angle as input features to predict the net power generation index after the execution of the action, and use the prediction result to replace the output results of the aforementioned joint simulation and rolling optimization for the determination of the net benefit of the auxiliary machine action and the hibernation decision, as an alternative implementation of the aforementioned digital twin joint simulation scheme.
[0026] (II) Adaptive Sleep Scheduling Mechanism of Edge Computing Power and Control Network for Physical Operating Condition Awareness
[0027] The edge computing gateway referred to in this invention refers to an industrial Internet of Things (IoT) terminal device deployed at a photovoltaic power station site that has the ability to collect, preprocess, and offload computing power. The software-defined network refers to a network architecture that separates the network control plane from the data forwarding plane and dynamically orchestrates the network topology, routing, and bandwidth resources through a centralized controller.
[0028] In this embodiment, the central decision-making layer periodically collects the operating status flags of each photovoltaic array tracking bracket and inverter, as well as the operating condition judgment results output by the digital twin model, through the existing industrial control system interface. These are divided into three states: operating state, steady state, and nighttime sleep state. For each operating condition state, the central decision-making layer determines the minimum necessary sensor sampling frequency corresponding to the current operating condition state based on the pre-established state-to-network topology mapping relationship. and minimum necessary edge computing power requirements The central decision-making level will , The sampling frequency currently configured in the network Computing power configuration If, during the comparison, at least one of the sampling frequency and computing power configurations currently configured in the network exceeds the requirements corresponding to the current operating conditions, it is determined that there are redundant resources that can be put into hibernation. Based on the comparison results, the software-defined network controller generates a topology reconfiguration instruction, puts the aggregation gateway and sensor nodes in the redundant state into hibernation mode, and dynamically migrates the edge computing tasks originally undertaken by the redundant nodes to nearby edge nodes or central servers that are still in an active state for execution.
[0029] The power savings in information systems resulting from hibernation scheduling can be calculated using the following formula:
[0030] ;
[0031] In the formula, This refers to the amount of electricity saved by the IoT system due to sleep scheduling during the statistical period. When no hibernation scheduling is performed The baseline power consumption of the network and computing power at any given moment. After hibernation scheduling The actual power consumption of the network and computing power at any given time. To ensure statistical time intervals, the system further incorporates a periodic wake-up detection mechanism to periodically verify whether the operating conditions of physical devices have changed. When the array is detected to transition from a steady state or a nighttime sleep state to an operating state, the system immediately restores the normal topology and sampling frequency of the corresponding gateway and sensor nodes, and relocates the migrated computational tasks back, thereby ensuring that the monitoring and control accuracy is not affected.
[0032] (III) Dynamic suppression mechanism of heat loss in collector networks based on spatial impedance matrix
[0033] The spatial impedance topology matrix referred to in this invention refers to a matrix-based data structure that describes the electrical impedance relationship between each inverter and the step-up substation, constructed based on the physical length, wire diameter, material, and real-time temperature of the entire station's collection lines. The automatic generation control referred to in this invention refers to the automated control mechanism in which the power grid dispatching system issues active power adjustment commands to the power station and requires the power station to track and execute them in real time.
[0034] In this embodiment, the central decision-making layer constructs an initial collector line impedance topology matrix based on the cable length, diameter, and material parameters of each array to the step-up substation recorded in the power plant construction drawings. The matrix contains the first... The element corresponds to the first element. Equivalent resistance from inverter to step-up substation The system continuously collects real-time output current of each inverter and measured temperature of the corresponding cables. The impedance matrix is dynamically temperature corrected according to the following formula:
[0035] ;
[0036] In the formula, For the first The current resistance of the cable after correction. For the first The resistance of the cable at the reference temperature. The temperature coefficient of resistance of the cable conductor material. For the first Current measured temperature of the cable. This is a reference temperature.
[0037] Upon receiving the total power dispatch instruction for the entire station issued by the power grid Subsequently, the central decision-making layer uses the corrected impedance topology matrix as the weighting basis to construct a power allocation optimization model aimed at reducing the heat loss of the entire station's power collection network:
[0038] ;
[0039] The constraints are:
[0040] ;
[0041] In the formula, The total number of inverters participating in the scheduling. For the first The active power output of the inverter to be solved. For the first The maximum allowable output of the inverter For the first Cable current, and The relationship is:
[0042] ;
[0043] In the formula, For the first Roadside operating voltage For the first The power factor of the circuit is optimized by using quadratic programming or Lagrange multiplier method, commonly used in power flow calculations, to solve the constrained optimization model. This yields the output allocation results for each inverter that meet the total power constraint and have low heat loss. The solution results reflect the principle of prioritizing electrical distance, meaning that inverters with smaller impedance values are allocated a higher proportion of output, while the output of far-end inverters with larger impedance values is correspondingly reduced. The central decision-making layer sends the power allocation instructions obtained from the solution to each inverter for execution through the existing communication interface, and collects current and temperature feedback data after execution in real time. The process returns to the aforementioned temperature correction and optimization solution process at fixed or event-triggered intervals, forming a dynamic closed-loop optimization that changes with line temperature and scheduling instructions.
[0044] In another embodiment of the present invention, the space impedance matrix heat loss suppression module can also use an impedance sorting-based heuristic allocation method to replace the solution process of the aforementioned constrained optimization model: each inverter is sorted in advance according to the impedance value of its corresponding cable from small to large. After receiving the total power limit command, the output of each inverter is increased to its allowed upper limit in the sorting order until the total power constraint is met. The output of the remaining inverters in the lower order is reduced to the allowed lower limit. The simplified sorting and allocation algorithm realizes the differentiated scheduling effect of electrical distance priority, which is suitable for field controllers with limited computing resources.
[0045] (iv) The principle of coordinated operation of the three major functional modules
[0046] The digital twin auxiliary optimization module, the network adaptive sleep scheduling module, and the spatial impedance matrix heat loss suppression module share information through a unified data bus at the central decision-making level. Specifically, the array operating condition judgment result output by the digital twin auxiliary optimization module serves as the status input of the network adaptive sleep scheduling module, used to determine the network topology and computing power configuration to be executed. The edge node computing power surplus information fed back by the network adaptive sleep scheduling module provides data support for solving the power allocation optimization model of the spatial impedance matrix heat loss suppression module. The output allocation results of each inverter output by the spatial impedance matrix heat loss suppression module are fed back to the digital twin auxiliary optimization module to correct the power generation gain benchmark used in its joint simulation. Through the data flow in the above three directions, this invention incorporates the three previously independently managed dimensions of physical equipment energy consumption, information network energy consumption, and collector line heat loss into a unified collaborative optimization closed loop, achieving a systematic improvement in the net energy efficiency of the power plant without changing the basic functions of the existing industrial control system and automatic power generation control module.
Claims
1. An industrial control system for photovoltaic power generation energy-saving management, comprising a sensing layer, an edge collaboration layer, and a central decision-making layer, wherein the sensing layer includes a meteorological irradiance sensing unit, a tracking bracket and its drive motor, an inverter and its cooling fan, temperature and current sensors for collector lines, an edge computing gateway, and sensing nodes; the edge collaboration layer includes a software-defined network controller and an edge computing resource pool; and the central decision-making layer is deployed on a station control center server and interfaces with existing industrial control systems and automatic power generation control modules within the station. The system is characterized in that... The central decision-making layer is equipped with a digital twin auxiliary machine optimization module, a network adaptive sleep scheduling module, and a spatial impedance matrix heat loss suppression module. The digital twin auxiliary equipment optimization module is configured to perform joint simulation of the power generation gain and power consumption generated by the tracking action and heat dissipation action performed by the auxiliary equipment based on micro-meteorological forecast data and auxiliary equipment energy consumption model, so as to obtain a net power generation index. When the net power generation index is not less than a preset economic threshold, an action command is issued to the corresponding auxiliary equipment. When the net power generation index is less than the preset economic threshold, a sleep command is issued to the corresponding auxiliary equipment. The auxiliary equipment includes the drive motor of the tracking bracket and the cooling fan of the inverter. The network adaptive sleep scheduling module is configured to drive the software-defined network controller to reconstruct the data acquisition network topology and schedule the computing power allocation of the edge computing resource pool based on the operating conditions of the physical devices in the perception layer. The space impedance matrix heat loss suppression module is configured as a dynamic electrical impedance topology matrix based on the entire station's collection lines. When it receives a power limiting command from the power grid, it constructs a power allocation optimization model with the goal of reducing the heat loss of the collection network, solves for the power allocation command of each inverter, and sends it out. The digital twin auxiliary machine optimization module, the network adaptive sleep scheduling module, and the spatial impedance matrix heat loss suppression module achieve data interaction through a unified data bus, forming a collaborative closed loop.
2. The system according to claim 1, characterized in that, The digital twin auxiliary equipment optimization module is configured to calculate the difference between the power generation gain and the power consumption for multiple consecutive prediction steps in the prediction time domain, and to penalize the change amplitude of the control quantity between adjacent prediction steps during the calculation process in order to suppress the operation frequency of the auxiliary equipment. The digital twin auxiliary machine optimization module is also configured to correct the model parameters used in the co-simulation based on the actual power generation data and energy consumption feedback data collected after the execution of the action.
3. The system according to claim 1, characterized in that, The network adaptive sleep scheduling module is configured to divide the operating conditions of the physical devices in the perception layer into three states: running state, steady state, and nighttime sleep state, and determine the sampling frequency and computing power requirements corresponding to the current operating condition state based on the pre-established state-to-network topology mapping relationship. When at least one of the sampling frequency and computing power configurations currently configured in the network exceeds the requirements corresponding to the current operating state, it is determined that there are redundant resources, and the software-defined network controller is driven to put the aggregation gateway and sensor node in the redundant state into sleep mode, while migrating the computing tasks undertaken by the redundant edge nodes to the edge nodes in the active state.
4. The system according to claim 3, characterized in that, The network adaptive sleep scheduling module is also configured to periodically check whether the operating conditions of the physical devices in the sensing layer have changed. When it detects a change from either steady state or nighttime sleep state to running state, it restores the sampling frequency of the aggregation gateway and sensor nodes in sleep mode and migrates the migrated computing tasks back.
5. The system according to claim 1, characterized in that, The spatial impedance matrix heat loss suppression module is configured to construct an initial impedance topology matrix based on the length, wire diameter, and material parameters of the collector lines from each inverter to the booster station, and to dynamically correct the initial impedance topology matrix based on the measured temperature of the cables corresponding to the real-time output current of each inverter, so as to obtain the corrected impedance topology matrix. The spatial impedance matrix heat loss suppression module is further configured to, after receiving the total power dispatch instruction for the entire station issued by the power grid, construct a power allocation optimization model with the goal of reducing the heat loss of the entire station's power collection network based on the modified impedance topology matrix as the weight, and solve for the power allocation instruction of each inverter under the conditions of satisfying the total power constraint of the entire station and the upper limit constraint of the output of each inverter.
6. The system according to claim 5, characterized in that, The solution results of the power allocation optimization model satisfy: In the modified impedance topology matrix, inverters with smaller impedance values are allocated a higher proportion of output power, while inverters with larger impedance values are allocated a lower proportion of output power.
7. The system according to claim 1, characterized in that, The physical equipment condition determination result output by the digital twin auxiliary machine optimization module serves as the status input of the network adaptive sleep scheduling module. The edge node computing power surplus information fed back by the network adaptive sleep scheduling module provides data support for solving the power allocation optimization model of the space impedance matrix heat loss suppression module. The power allocation commands of each inverter output by the space impedance matrix heat loss suppression module are fed back to the digital twin auxiliary optimization module to correct the power generation gain benchmark used in the co-simulation.
8. The system according to claim 1, characterized in that, The digital twin auxiliary machine optimization module is further configured to use a machine learning regression model trained based on historical operating data, with real-time irradiance, cloud movement speed and current tracking angle as input features, to predict the net power generation index after the auxiliary machine performs the action, and use the predicted net power generation index to replace the net power generation index obtained by the co-simulation for the determination of the action command and the sleep command.
9. The system according to claim 5, characterized in that, The space impedance matrix heat loss suppression module is also configured to replace the solution process of the power allocation optimization model with an impedance sorting allocation method: sort each inverter according to the impedance value of its corresponding cable from small to large; after receiving the total power scheduling command of the whole station, increase the output of each inverter to its upper limit in the sorting order until the total power constraint of the whole station is met, and reduce the output of the remaining inverters in the sorting order to the lower limit.
