使用这个注解中给出的提示，我可以创建一个包含字段名和它出现的（最后一个）页面的Map。

HashMap<String, Integer> formFieldPages = new HashMap<>();
for (int page_i = 0; page_i < pdf_document.getNumberOfPages(); page_i++) {
    List<PDAnnotation> annotations = pdf_document.getPage(page_i).getAnnotations(); //
    for (PDAnnotation annotation: annotations) {
        if (!(annotation instanceof PDAnnotationWidget)) {
            System.err.println("Unknown annotation type " + annotation.getClass().getName() + ": " + annotation.toString());
            continue;
        }
        String name = ((PDAnnotationWidget)annotation).getCOSObject().getString(COSName.T);
        if (name == null) {
            System.err.println("Unknown widget name: " + annotation.toString());
            continue;
        }
        // make sure the field does not exists in the map
        if (formFieldPages.containsKey(name)) {
            System.err.println("Duplicated widget name, overwriting previous page value " + formFieldPages.get(name) + " with newly found page " + page_i + ": " + annotation.toString());
        }
        formFieldPages.put(name, page_i);
    }
}

现在查找页面就像

int page = formFieldPages.get(docField.getPartialName());

请注意，如果由于某种原因该小部件不存在，这可能会引发nullpointerexception。
前面的答案如下。看来我的方法是错的，但我保留它作为参考。
我找到了 /P 元素，看起来像是页面：

int page = (int)currentField.getCOSObject().getCOSObject(COSName.P).getObjectNumber();
page = page - 5; // I couldn't figure out why it's off by 4, but tests showed that the actual PDF page 1 (index [0]) is represented by `\P {4, 0}`, page 2 ([1]) is called "5", page 3 ([2]) is "6", etc.