[{"data":1,"prerenderedAt":195},["ShallowReactive",2],{"content-query-vKHCKFwqs1":3},{"_path":4,"_dir":5,"_draft":6,"_partial":6,"_locale":7,"title":8,"description":9,"date":10,"cover":11,"type":12,"body":13,"_type":189,"_id":190,"_source":191,"_file":192,"_stem":193,"_extension":194},"/news/zh/3755","zh",false,"","使能油气勘探10倍级降本增效，东方物探联合人大高瓴学院发布基于昇思打造的全波形反演模型","随着油气勘探目标日益复杂，地震勘探技术亟待升级。东方物探全球首创的“两宽一高”技术成为行业引领，但其核心环节——全波形反演（FWI）因依赖高算力数值求解（如声波/弹性波方程），面临巨大成本挑战。","2025-06-10","https://obs-mindspore-file.obs.cn-north-4.myhuaweicloud.com/file/2025/06/13/6b5c1e5b141e4eea881477a0a3be2ce9.png","news",{"type":14,"children":15,"toc":186},"root",[16,24,30,35,55,60,68,76,84,89,94,102,107,115,120,128,133,141,149,156,161,166,181],{"type":17,"tag":18,"props":19,"children":21},"element","h1",{"id":20},"使能油气勘探10倍级降本增效东方物探联合人大高瓴学院发布基于昇思打造的全波形反演模型",[22],{"type":23,"value":8},"text",{"type":17,"tag":25,"props":26,"children":27},"p",{},[28],{"type":23,"value":29},"随着油气勘探目标日益复杂，地震勘探技术亟待升级。东方物探全球首创的“两宽一高”技术成为行业引领，但其核心环节——**全波形反演（FWI）**因依赖高算力数值求解（如声波/弹性波方程），面临巨大成本挑战。",{"type":17,"tag":25,"props":31,"children":32},{},[33],{"type":23,"value":34},"2025年5月28日，3000亿参数昆仑大模型发布会在京举办，全面展示昆仑大模型建设应用成果，见证人工智能对行业发展的变革影响。",{"type":17,"tag":25,"props":36,"children":37},{},[38,40,46,48,53],{"type":23,"value":39},"2024年，中石油启动“人工智能+”行动，依托昆仑大模型聚焦FWI科学计算攻关。由中石油东方物探、人大高瓴人工智能学院、华为组成的产学研团队，突破FWI算力瓶颈，解决FWI正反演过程算力成本高、技术发展推广难的问题。研发",{"type":17,"tag":41,"props":42,"children":43},"strong",{},[44],{"type":23,"value":45},"AI驱动的FWI新一代解决方案",{"type":23,"value":47},"，实现地震波求解效率",{"type":17,"tag":41,"props":49,"children":50},{},[51],{"type":23,"value":52},"量级提升",{"type":23,"value":54},"，推动油气勘探“AI换芯”进程。",{"type":17,"tag":25,"props":56,"children":57},{},[58],{"type":23,"value":59},"该研究基于昇思MindSpore AI4S使能套件，通过数据与机理双驱动的解决方案、应用昇思自动微分技术，并基于扩散模型的三维地质数据生成技术提升求解效率。",{"type":17,"tag":25,"props":61,"children":62},{},[63],{"type":17,"tag":41,"props":64,"children":65},{},[66],{"type":23,"value":67},"# 01",{"type":17,"tag":25,"props":69,"children":70},{},[71],{"type":17,"tag":41,"props":72,"children":73},{},[74],{"type":23,"value":75},"技术特点",{"type":17,"tag":25,"props":77,"children":78},{},[79],{"type":17,"tag":80,"props":81,"children":83},"img",{"alt":7,"src":82},"https://obs-mindspore-file.obs.cn-north-4.myhuaweicloud.com/file/2025/06/13/0c4e02f2db5e4ef682402ce96270fb4d.png",[],{"type":17,"tag":25,"props":85,"children":86},{},[87],{"type":23,"value":88},"图1. 模型推理过程示意",{"type":17,"tag":25,"props":90,"children":91},{},[92],{"type":23,"value":93},"项目融合三大核心技术：",{"type":17,"tag":25,"props":95,"children":96},{},[97],{"type":17,"tag":41,"props":98,"children":99},{},[100],{"type":23,"value":101},"关键技术一：数据与机理双驱动的波动方程智能求解方法",{"type":17,"tag":25,"props":103,"children":104},{},[105],{"type":23,"value":106},"基于神经算子构建物理引导正演模型框架，深度融合数值求解器数学机理，增强解释性与泛化能力。网络编码解析速度场、震源等输入参数，构造物理约束损失函数。创新采用“混合专家（MOE）预训练 +微调”的训练策略，大规模预训练分频段学习波场传播特征，增强对多种震源波形的适应能力。少量真实数据微调提升实际场景求解能力。推理实现“神经网络预测-传统算法修正”协同计算范式，保留求解器高精度的同时提升声波方程求解效率1个数量级。",{"type":17,"tag":25,"props":108,"children":109},{},[110],{"type":17,"tag":41,"props":111,"children":112},{},[113],{"type":23,"value":114},"关键技术二：基于自动微分的反演框架",{"type":17,"tag":25,"props":116,"children":117},{},[118],{"type":23,"value":119},"基于MindSpore自动微分框架，构建可微分正演流程，从而在反演中省去传统伴随方法波场反传过程，直接自动求取速度模型梯度修正量，减少了约50%计算量的同时大大简化了计算流程。支持灵活切换多种反演策略，并引入多种多参数优化相关优化策略，有助于跨过周期跳跃导致的局部最优，显著降低反演对准确初始速度模型的依赖。",{"type":17,"tag":25,"props":121,"children":122},{},[123],{"type":17,"tag":41,"props":124,"children":125},{},[126],{"type":23,"value":127},"关键技术三：基于扩散模型的三维地质数据生成技术",{"type":17,"tag":25,"props":129,"children":130},{},[131],{"type":23,"value":132},"首创基于扩散模型的技术，通过渐进式噪声添加与去噪机制捕捉复杂地质结构特征，解决三维地质速度模型数据稀缺的问题。采用大规模模拟数据预训练 + 少量真实场景数据微调的训练策略。成功生成包含断层、盐丘等复杂地质构造的速度模型，涵盖不同地质条件和反演中间状态。突破传统数据采集限制，极大丰富正演模型训练所需数据规模，为成功训练奠定基础。",{"type":17,"tag":25,"props":134,"children":135},{},[136],{"type":17,"tag":41,"props":137,"children":138},{},[139],{"type":23,"value":140},"# 02",{"type":17,"tag":25,"props":142,"children":143},{},[144],{"type":17,"tag":41,"props":145,"children":146},{},[147],{"type":23,"value":148},"价值收益",{"type":17,"tag":25,"props":150,"children":151},{},[152],{"type":17,"tag":80,"props":153,"children":155},{"alt":7,"src":154},"https://obs-mindspore-file.obs.cn-north-4.myhuaweicloud.com/file/2025/06/13/4407f36f99c94830a54962021e893cc2.png",[],{"type":17,"tag":25,"props":157,"children":158},{},[159],{"type":23,"value":160},"图2. a) 地震数据 (黑白) 与模拟数据 (红蓝) 精确叠合；b) 反演的初始速度模型；c) 反演结果用于成像显示出清晰的地层轮廓",{"type":17,"tag":25,"props":162,"children":163},{},[164],{"type":23,"value":165},"该解决方案在求解效率提升10倍以上，并在落地应用上得到验证：",{"type":17,"tag":167,"props":168,"children":169},"ol",{},[170,176],{"type":17,"tag":171,"props":172,"children":173},"li",{},[174],{"type":23,"value":175},"求解效率量级跃升：二维/三维声波方程正演计算效率提升4-6倍，全波形反演整体效率提升超10倍，初始模型依赖程度降低40%，为FWI技术的工业化应用扫清了关键障碍。未来有望将三维FWI处理从“数月级”缩短至“周级”，5分钟智能反演可完成传统集群数小时工作量。 使高精度成像计算更易获得，实现声波高精成像算力平权。",{"type":17,"tag":171,"props":177,"children":178},{},[179],{"type":23,"value":180},"工业化落地与赋能： 已完成二维实际资料验证及三维模型小试。成功实现某海洋拖缆资料6km测线二维反演生产验证，效率提升10倍以上，反演显示清晰浅层轮廓和高精度成像潜力，精度达标。计划于2025年在中石油2-3个生产项目试点。未来将集成端到端智能赋能的FWI应用模块到东方物探核心软件系统GeoEast，实现“智能化换芯”。",{"type":17,"tag":25,"props":182,"children":183},{},[184],{"type":23,"value":185},"未来，AI驱动的FWI新一代解决方案有望引领行业范式革新：不仅推动FWI发展，也为地质勘探、医学成像等领域的反问题求解提供新方法论，推动全产业链智能化升级。",{"title":7,"searchDepth":187,"depth":187,"links":188},4,[],"markdown","content:news:zh:3755.md","content","news/zh/3755.md","news/zh/3755","md",1776506088788]