Gigabit passive optical network (GPON) all-optical network energy consumption intelligent monitoring and energy-saving scheduling system integrated with AI

Abstract

The invention discloses an AI-fused gigabit passive optical network (GPON) all-optical network energy consumption intelligent monitoring and energy-saving scheduling system, and relates to the technical field of crossing of optical communication networks and artificial intelligence, the scheduling system comprises a data acquisition module, an edge AI gateway, a cloud decision platform and an execution terminal, the method has the advantages that dynamic space-time analysis of whole network flow, environmental parameters and topological characteristics is achieved, compared with traditional cloud centralized prediction and edge calculation, data transmission delay is remarkably reduced, decision lag caused by network congestion is avoided, the system combines time stamps and spatial topology, and the system is more efficient and reliable. According to the method, energy consumption rules of different scenes such as dormitory areas and office areas are actively recognized, a prediction basis of millisecond-level response is provided for dynamic energy-saving scheduling, meanwhile, the calculation power limitation of edge equipment is overcome through model lightweight design, it is ensured that efficient operation can still be achieved in a resource-limited environment, and the real-time performance and accuracy of energy efficiency management are improved from the source.

Description

Technical Field

[0001] This invention relates to the field of interdisciplinary technology of optical communication networks and artificial intelligence, specifically to an AI-integrated GPON all-optical network energy consumption intelligent monitoring and energy-saving scheduling system. Background Technology

[0002] With the widespread application of fiber optic access technology, GPON (Gigabit Passive Optical Network) has become the mainstream solution. Its network scale and user capacity have continued to expand. The surge in the number of network devices and the increasing variety of service types have led to a significant increase in overall energy consumption. This not only brings high operating costs but also goes against the goal of green and low-carbon development. Traditional energy management methods mainly rely on static configuration or simple threshold control, which makes it difficult to achieve refined energy-saving goals under complex and ever-changing network traffic and service demands, while ensuring that the quality of user services is not compromised.

[0003] Faced with this challenge, a more intelligent management approach is urgently needed. The key lies in how to achieve real-time and accurate monitoring of the entire network's energy consumption and make dynamic energy-saving scheduling decisions based on this. This requires the system to proactively sense changes in network status, predict trends in key parameters, and, under the premise of meeting strict Quality of Service (QoS) requirements, to conduct coordinated and adaptive power consumption control of core devices such as Optical Line Terminals (OLTs) and Optical Network Units (ONUs) to achieve a dynamic balance between energy-saving goals and QoS assurance. This has become a technical challenge that urgently needs to be overcome to improve the energy efficiency of GPON networks. Summary of the Invention

[0004] The purpose of this invention is to provide an AI-integrated GPON all-optical network energy consumption intelligent monitoring and energy-saving scheduling system.

[0005] To achieve the above objectives, the present invention provides the following technical solution: an AI-integrated GPON all-optical network energy consumption intelligent monitoring and energy-saving scheduling system, comprising a scheduling system, wherein the scheduling system includes modules:

[0006] The data acquisition module is configured to collect real-time traffic data, device energy consumption data, and ambient temperature and humidity data in the GPON network, and to mark time dimension information and spatial topology information.

[0007] The edge AI gateway, deployed on the OLT side, includes a lightweight LSTM-1DCNN fusion model, configured to extract spatiotemporal features based on historical traffic data, device location topology, environmental parameters and time stamps, and output energy consumption prediction results with a granularity of the next 5 minutes.

[0008] The cloud-based decision-making platform includes a multi-objective optimization decision engine and a digital twin verification module. The multi-objective optimization decision engine is configured to generate energy-saving strategies by taking the minimization of total network power consumption, QoS default duration and fault recovery time as objective functions and combining knowledge graph rules. The digital twin verification module is configured to verify the QoS default probability of the strategy by simulating the injected traffic load.

[0009] The execution terminal includes an OLT optical power dynamic adjustment unit and an ONU multi-level sleep control unit. The OLT optical power dynamic adjustment unit is configured to adjust the transmission power using adaptive bias current control based on the traffic prediction results. The ONU multi-level sleep control unit is configured to switch power consumption states based on a three-level sleep mode extension of the IEEE 1904.1 standard.

[0010] As a further aspect of the present invention: the data acquisition module includes:

[0011] A flow sensor is configured to collect historical service traffic from the OLT and ONU in 5-minute granularities.

[0012] Temperature and humidity sensors are configured to monitor environmental parameters of the optical module.

[0013] Topology identification unit, configured to mark the location attributes and region type labels of ONUs.

[0014] As a further aspect of the present invention: the computation process of the lightweight LSTM-1DCNN fusion model satisfies:

[0015] h t =σ(W h [h t-1 ,x t ]+b h );

[0016] y=ReLU(ConvID(k=3,s=1)(h t ));

[0017] Among them, h t Let x be the hidden state at time t. t For the input feature vector, W h Let b be the weight matrix. h is the bias term, k is the kernel size, and s is the stride.

[0018] As a further aspect of the present invention: the OLT optical power dynamic adjustment unit performs adaptive bias current control.

[0019]

[0020] Among them, I bias For bias current, Imin I max Here, k is the current boundary value, k is the temperature compensation coefficient, and ΔT is the difference between the ambient temperature and the reference temperature. To predict the flow rate, Q th This is the traffic threshold.

[0021] As a further aspect of the present invention: the ONU multi-level sleep control unit includes:

[0022] Active mode: Power consumption 15.0W, wake-up latency 0ms;

[0023] Shallow sleep mode: power consumption 5.2W, wake-up latency 20ms;

[0024] Deep sleep mode: power consumption 0.8W, wake-up latency 500ms.

[0025] As a further aspect of the present invention: the objective function of the multi-objective optimization decision engine is:

[0026] min(α·P total +β·Violation QoS +γ·T recovery );

[0027] Where α, β, and γ are weighting coefficients, P total For the total power consumption of the entire network, Violation QoS For QoS violation duration, T recovery This refers to the fault recovery time.

[0028] As a further aspect of the present invention: the knowledge graph rules include conditional execution logic:

[0029] If the ONU area type is dormitory area and the time is >23:00, then deep sleep mode is allowed;

[0030] If a QoS priority tag is detected, the sleeping ONU will be forcibly woken up.

[0031] As a further aspect of the present invention: the operation process of the digital twin verification module includes:

[0032] Receive energy-saving strategy parameters and inject simulated traffic load;

[0033] Calculate the QoS violation probability, and approve policy execution when the violation probability is <5%;

[0034] When the probability of default is ≥5%, the strategy is re-optimized.

[0035] Compared with the prior art, the beneficial effects of the present invention by adopting the above technical solution are as follows:

[0036] 1. This invention utilizes a lightweight LSTM-1DCNN fusion model deployed directly at the edge to achieve dynamic spatiotemporal analysis of network traffic, environmental parameters, and topology features. It accurately predicts device energy consumption trends with a 5-minute granularity. Compared to traditional centralized cloud-based prediction, edge computing significantly reduces data transmission latency and avoids decision lag caused by network congestion. The system combines time stamps and spatial topology to proactively identify energy consumption patterns in differentiated scenarios such as dormitory areas and office areas, providing a millisecond-level predictive basis for dynamic energy-saving scheduling. At the same time, the lightweight model design overcomes the computing power limitations of edge devices, ensuring efficient operation even in resource-constrained environments, thereby improving the real-time performance and accuracy of energy efficiency management from the source.

[0037] 2. This invention utilizes a cloud-based multi-objective optimization decision engine to incorporate three key indicators—total network power consumption, QoS default duration, and fault recovery time—into a unified objective function for the first time. It also introduces a knowledge graph rule base for constraint. For example, it automatically identifies when ONUs are allowed to go into deep sleep during late-night hours in dormitory areas, while forcing high-priority services to remain active, avoiding service interruptions caused by "one-size-fits-all" energy saving. The optimization process considers the impact of temperature on optical module power consumption, ensuring that the energy-saving strategy conforms to actual physical constraints. The digital twin module further simulates and verifies the strategy, approving execution only when the QoS default probability is below 5%. This fundamentally resolves the conflict between energy saving and service quality, achieving a dual improvement in network energy efficiency and user experience.

[0038] 3. This invention breaks through the traditional static energy-saving mode and establishes a closed-loop dynamic control across the entire chain: the data acquisition module senses changes in the environment and traffic in real time; edge prediction results drive cloud-based strategy generation; the digital twin verification module simulates strategy risks through traffic load injection; and finally, the execution terminal realizes adaptive adjustment of OLT optical power and three-level sleep mode of ONU. This closed-loop design enables the system to have self-learning capabilities—when simulation verification finds that the probability of default exceeds the standard, it automatically triggers strategy re-optimization and iteratively updates the knowledge graph rules, without the need for manual intervention throughout the process. It continuously optimizes energy efficiency in complex and ever-changing network environments, forming a sustainable and evolving energy-saving mechanism. Attached Figure Description

[0039] Figure 1 This is a system flowchart of the present invention. Detailed Implementation

[0040] The specific embodiments of the present invention will be further described below with reference to the accompanying drawings. It should be noted that the description of these embodiments is for the purpose of helping to understand the present invention, but does not constitute a limitation of the present invention.

[0041] Furthermore, the technical features involved in the various embodiments of the present invention described below can be combined with each other as long as they do not conflict with each other.

[0042] Please see the appendix Figure 1 This invention relates to an AI-integrated GPON all-optical network energy consumption intelligent monitoring and energy-saving scheduling system, comprising a scheduling system, which includes modules:

[0043] The data acquisition module is configured to collect real-time traffic data, device energy consumption data, and ambient temperature and humidity data in the GPON network, and to mark time dimension information and spatial topology information.

[0044] The edge AI gateway, deployed on the OLT side, includes a lightweight LSTM-1DCNN fusion model, configured to extract spatiotemporal features based on historical traffic data, device location topology, environmental parameters and time stamps, and output energy consumption prediction results with a granularity of the next 5 minutes.

[0045] The cloud-based decision-making platform includes a multi-objective optimization decision engine and a digital twin verification module. The multi-objective optimization decision engine is configured to minimize the total power consumption of the entire network, the QoS default duration, and the fault recovery time as objective functions, and generate energy-saving strategies by combining knowledge graph rules. The digital twin verification module is configured to verify the QoS default probability of the strategy by simulating the injected traffic load.

[0046] The execution terminal includes an OLT optical power dynamic adjustment unit and an ONU multi-level sleep control unit. The OLT optical power dynamic adjustment unit is configured to adjust the transmission power using adaptive bias current control based on the traffic prediction results. The ONU multi-level sleep control unit is configured to switch power consumption states in a three-level sleep mode based on the IEEE 1904.1 standard extension.

[0047] In one embodiment of the present invention: the data acquisition module includes:

[0048] A flow sensor is configured to collect historical service traffic from the OLT and ONU in 5-minute granularities.

[0049] Temperature and humidity sensors are configured to monitor environmental parameters of the optical module.

[0050] Topology identification unit, configured to mark the location attributes and region type labels of ONUs.

[0051] In one embodiment of the present invention: the computation process of the lightweight LSTM-1DCNN fusion model satisfies:

[0052] h t =σ(W h [h t-1 ,x t ]+b h );

[0053] y=ReLU(ConvID(k=3,s=1)(h t ));

[0054] Among them, h t Let x be the hidden state at time t. t For the input feature vector, W h Let b be the weight matrix. h is the bias term, k is the kernel size, and s is the stride.

[0055] In one embodiment of the present invention: the OLT optical power dynamic adjustment unit performs adaptive bias current control:

[0056]

[0057] Among them, I bias For bias current, I min I max Here, k is the current boundary value, k is the temperature compensation coefficient, and ΔT is the difference between the ambient temperature and the reference temperature. To predict the flow rate, Q th This is the traffic threshold.

[0058] In one embodiment of the present invention: the ONU multi-level sleep control unit includes:

[0059] Active mode: Power consumption 15.0W, wake-up latency 0ms;

[0060] Shallow sleep mode: power consumption 5.2W, wake-up latency 20ms;

[0061] Deep sleep mode: power consumption 0.8W, wake-up latency 500ms.

[0062] In one embodiment of the present invention: the objective function of the multi-objective optimization decision engine is:

[0063] min(α·P total +β·Violation QoS +γ·T recovery );

[0064] Where α, β, and γ are weighting coefficients, P total For the total power consumption of the entire network, Violation QoS For QoS violation duration, T recovery This refers to the fault recovery time.

[0065] In one embodiment of the present invention: the knowledge graph rule includes conditional execution logic:

[0066] If the ONU area type is dormitory area and the time is >23:00, then deep sleep mode is allowed;

[0067] If a QoS priority tag is detected, the sleeping ONU will be forcibly woken up.

[0068] In one embodiment of the present invention, the operation flow of the digital twin verification module includes:

[0069] Receive energy-saving strategy parameters and inject simulated traffic load;

[0070] Calculate the QoS violation probability, and approve policy execution when the violation probability is <5%;

[0071] When the probability of default is ≥5%, the strategy is re-optimized.

[0072] Example 1: Nighttime Deep Energy-Saving Scheduling in Campus Dormitory Area

[0073] After a university deployed this system, the data acquisition module marked 300 ONUs (area type: student dormitory) in the dormitory area through topology identification units. The flow sensor monitored the flow rate at a 5-minute granularity, which dropped to 10% of the peak after 23:00 every day. The lightweight LSTM-1DCNN fusion model of the edge AI gateway combined historical data (including the flow rate fluctuation characteristics at the end of the semester) and the ambient temperature (28℃) to predict that the energy consumption would remain low for the next 30 minutes.

[0074] The cloud-based decision engine activates the knowledge graph rule: "Deep sleep is allowed when dormitory time is > 23:00". The generated policy is: disable the broadcast function of non-emergency ONUs, schedule 95% of ONUs to enter deep sleep mode (power consumption 0.8W), inject simulated load (including burst video requests) into the digital twin module, verify that the QoS violation probability is only 2.1% (below the 5% threshold), and approve the execution.

[0075] Perform terminal operations:

[0076] OLT optical power unit based on flow prediction value (below the threshold Q) th =100Mbps), activate adaptive bias current control, and reduce the transmit power from I max Down to I min +k·ΔT (ΔT=5℃), power consumption is reduced by 40%;

[0077] The ONU sleep unit switches to deep sleep according to the IEEE 1904.1 standard, with a wake-up latency tolerance of 500ms (to meet the low real-time requirements in the middle of the night);

[0078] Results: Daily power saving reached 62 kWh, and the student morning wake-up online rate was 100%, verifying the balance between energy saving and service quality in low-traffic scenarios.

[0079] Example 2: Dynamic Optimization of Hybrid Business in Enterprise Parks

[0080] A technology park includes an office area (high real-time business), an IDC area (high priority business), and public areas. The data acquisition module detected that the traffic in the office area decreased by 60% during the lunch break, but the IDC area maintained a high load. The edge model identified the spatial topology differences: the ONUs in the office area were marked as "normal business", while those in the IDC area were marked as "QoS priority".

[0081] The cloud-based decision engine uses a multi-objective function min(α·P) total +β·Violation QoS +γ·T recovery )optimization:

[0082] • The ONU in the office area switches to shallow sleep mode (power consumption 5.2W, wake-up latency 20ms) to meet the needs of sudden meetings;

[0083] • The IDC area is forced to maintain an active mode to avoid the risk of business interruption;

[0084] • The OLT optical power unit dynamically increases the bias current compensation (k·ΔT term) based on the ambient temperature of 32℃ to avoid high-temperature optical decay;

[0085] Digital twin verification: Injecting burst traffic load into the IDC area, the QoS default probability is 3.2% < 5%, the policy is approved, and the execution terminal achieves differentiated control by zone: the power saving rate in the office area reaches 51%, the IDC area has zero service interruption, the total power consumption of the entire network decreases by 38%, and the fault recovery time is shortened to 120 seconds (originally an average of 300 seconds).

[0086] Example 3: Resilient Energy-Saving Recovery under High-Temperature Fault Scenarios

[0087] A city experienced an extreme high temperature (38℃). The data acquisition module detected that the temperature of a certain OLT optical module exceeded the standard and the flow rate fluctuated abnormally. The edge model combined with temperature and humidity data to predict that the area may fail within 5 minutes and issued a real-time alarm.

[0088] Cloud-based decision engine emergency response:

[0089] 1. Activate the knowledge graph rule: "Force wake up the ONU if a QoS priority tag is detected" to ensure uninterrupted emergency communication;

[0090] 2. Adjust the weights of the objective function (increase γ·T) recovery Prioritize and generate energy-saving strategies with rapid recovery as the primary focus:

[0091] The faulty OLT and associated ONU switch to shallow sleep (maintaining the basic link);

[0092] Health OLT takeover business, optical power unit according to I max Operational support signal strength;

[0093] The digital twin module simulates the takeover process: the probability of default is 4.7% < 5%. The strategy is executed, and the execution terminal completes the optical power reconfiguration and ONU mode switching within 90 seconds. The fault recovery time is shortened by 65% ​​compared with traditional methods. Afterwards, the system automatically updates the knowledge graph: a new rule of "disable deep sleep when the temperature is > 35℃" is added to form a closed-loop optimization.

[0094] While the present invention has been disclosed above with reference to preferred embodiments, it is not intended to limit the invention. Any variations and modifications can be made by those skilled in the art without departing from the spirit and scope of the invention. Therefore, any modifications, equivalent changes, and alterations made to the above embodiments based on the technical essence of the present invention, without departing from the scope of the invention, fall within the protection scope defined by the claims of the present invention.

Claims

1. An AI-integrated GPON all-optical network energy consumption intelligent monitoring and energy-saving scheduling system, including a scheduling system, characterized in that: The scheduling system includes modules: The data acquisition module is configured to collect real-time traffic data, device energy consumption data, and ambient temperature and humidity data in the GPON network, and to mark time dimension information and spatial topology information. The edge AI gateway, deployed on the OLT side, includes a lightweight LSTM-1DCNN fusion model, configured to extract spatiotemporal features based on historical traffic data, device location topology, environmental parameters and time stamps, and output energy consumption prediction results with a granularity of the next 5 minutes. The cloud-based decision-making platform includes a multi-objective optimization decision engine and a digital twin verification module. The multi-objective optimization decision engine is configured to generate energy-saving strategies by taking the minimization of total network power consumption, QoS default duration and fault recovery time as objective functions and combining knowledge graph rules. The digital twin verification module is configured to verify the QoS default probability of the strategy by simulating the injected traffic load. The execution terminal includes an OLT optical power dynamic adjustment unit and an ONU multi-level sleep control unit. The OLT optical power dynamic adjustment unit is configured to adjust the transmission power using adaptive bias current control based on the traffic prediction results. The ONU multi-level sleep control unit is configured to switch power consumption states based on a three-level sleep mode extension of the IEEE 1904.1 standard.

2. The AI-integrated GPON all-optical network energy consumption intelligent monitoring and energy-saving scheduling system according to claim 1, characterized in that: The data acquisition module includes: A flow sensor is configured to collect historical service traffic from the OLT and ONU in 5-minute granularities. Temperature and humidity sensors are configured to monitor environmental parameters of the optical module. Topology identification unit, configured to mark the location attributes and region type labels of ONUs.

3. The AI-integrated GPON all-optical network energy consumption intelligent monitoring and energy-saving scheduling system according to claim 2, characterized in that: The computation process of the lightweight LSTM-1DCNN fusion model satisfies: h t =σ(W h [h t-1 ,x t ]+b h ); y=ReLU(ConvID(k=3,s=1)(h t )); Among them, h t Let x be the hidden state at time t. t For the input feature vector, W h Let b be the weight matrix. h is the bias term, k is the kernel size, and s is the stride.

4. The AI-integrated GPON all-optical network energy consumption intelligent monitoring and energy-saving scheduling system according to claim 3, characterized in that: The OLT optical power dynamic adjustment unit performs adaptive bias current control: Among them, I bias For bias current, I min I max Here, k is the current boundary value, k is the temperature compensation coefficient, and ΔT is the difference between the ambient temperature and the reference temperature. To predict the flow rate, Q th This is the traffic threshold.

5. The AI-integrated GPON all-optical network energy consumption intelligent monitoring and energy-saving scheduling system according to claim 4, characterized in that: The ONU multi-level sleep control unit includes: Active mode: Power consumption 15.0W, wake-up latency 0ms; Shallow sleep mode: power consumption 5.2W, wake-up latency 20ms; Deep sleep mode: power consumption 0.8W, wake-up latency 500ms.

6. The AI-integrated GPON all-optical network energy consumption intelligent monitoring and energy-saving scheduling system according to claim 5, characterized in that: The objective function of the multi-objective optimization decision engine is: min(α·P total +b·Violation QoS +γ·T recovery ); Where α, β, and γ are weighting coefficients, P total For the total power consumption of the entire network, Violation QoS For QoS violation duration, T recovery This refers to the fault recovery time.

7. The AI-integrated GPON all-optical network energy consumption intelligent monitoring and energy-saving scheduling system according to claim 6, characterized in that: The knowledge graph rules include conditional execution logic: If the ONU area type is dormitory area and the time is >23:00, then deep sleep mode is allowed; If a QoS priority tag is detected, the sleeping ONU will be forcibly woken up.

8. The AI-integrated GPON all-optical network energy consumption intelligent monitoring and energy-saving scheduling system according to claim 7, characterized in that: The operation process of the digital twin verification module includes: Receive energy-saving strategy parameters and inject simulated traffic load; Calculate the QoS violation probability, and approve policy execution when the violation probability is <5%; When the probability of default is ≥5%, the strategy is re-optimized.

Images